<?xml version="1.0" encoding="UTF-8"?><rss xmlns:dc="http://purl.org/dc/elements/1.1/" xmlns:content="http://purl.org/rss/1.0/modules/content/" xmlns:atom="http://www.w3.org/2005/Atom" version="2.0" xmlns:itunes="http://www.itunes.com/dtds/podcast-1.0.dtd" xmlns:googleplay="http://www.google.com/schemas/play-podcasts/1.0"><channel><title><![CDATA[LeapNotes by Kader Diagne]]></title><description><![CDATA[Notes on AI, Africa, and the leaps that change the game]]></description><link>https://www.leapnotes.tech</link><image><url>https://substackcdn.com/image/fetch/$s_!hy2c!,w_256,c_limit,f_auto,q_auto:good,fl_progressive:steep/https%3A%2F%2Fbucketeer-e05bbc84-baa3-437e-9518-adb32be77984.s3.amazonaws.com%2Fpublic%2Fimages%2F16ce3919-10c1-4ed0-b46f-720759955c7f_147x140.png</url><title>LeapNotes by Kader Diagne</title><link>https://www.leapnotes.tech</link></image><generator>Substack</generator><lastBuildDate>Tue, 28 Jul 2026 21:32:20 GMT</lastBuildDate><atom:link href="https://www.leapnotes.tech/feed" rel="self" type="application/rss+xml"/><copyright><![CDATA[Kader Diagne]]></copyright><language><![CDATA[en]]></language><webMaster><![CDATA[kaderdiagne@substack.com]]></webMaster><itunes:owner><itunes:email><![CDATA[kaderdiagne@substack.com]]></itunes:email><itunes:name><![CDATA[Kader Diagne]]></itunes:name></itunes:owner><itunes:author><![CDATA[Kader Diagne]]></itunes:author><googleplay:owner><![CDATA[kaderdiagne@substack.com]]></googleplay:owner><googleplay:email><![CDATA[kaderdiagne@substack.com]]></googleplay:email><googleplay:author><![CDATA[Kader Diagne]]></googleplay:author><itunes:block><![CDATA[Yes]]></itunes:block><item><title><![CDATA[Ça n’a jamais été une course. C’est le moment pour l’Afrique de faire un grand bond en avant.]]></title><description><![CDATA[L'intelligence devient abondante. Pour l'Afrique, ce n'est pas un retard &#224; rattraper &#8212; c'est une mati&#232;re premi&#232;re. Pourquoi, et ce que nous b&#226;tissons]]></description><link>https://www.leapnotes.tech/p/ca-na-jamais-ete-une-course-cest</link><guid isPermaLink="false">https://www.leapnotes.tech/p/ca-na-jamais-ete-une-course-cest</guid><dc:creator><![CDATA[Kader Diagne]]></dc:creator><pubDate>Mon, 06 Jul 2026 14:58:58 GMT</pubDate><enclosure url="https://substackcdn.com/image/fetch/$s_!hy2c!,w_256,c_limit,f_auto,q_auto:good,fl_progressive:steep/https%3A%2F%2Fbucketeer-e05bbc84-baa3-437e-9518-adb32be77984.s3.amazonaws.com%2Fpublic%2Fimages%2F16ce3919-10c1-4ed0-b46f-720759955c7f_147x140.png" length="0" type="image/jpeg"/><content:encoded><![CDATA[<p><strong><span data-color="#ff0000" style="color: rgb(255, 0, 0);">Read this in English &#8594; [</span><a href="https://www.leapnotes.tech/p/it-was-never-a-race-this-is-africas?r=39ps0"><span data-color="#ff0000" style="color: rgb(255, 0, 0);">lien</span></a><span data-color="#ff0000" style="color: rgb(255, 0, 0);">]</span></strong></p><p><em>Le monde d&#233;bat sans fin de qui remporte la course &#224; l&#8217;IA. Je suis africain ; j&#8217;ai &#233;tudi&#233; et b&#226;ti l&#8217;IA en Allemagne &#8212; et, pour ma part, je n&#8217;y ai jamais vu de course. J&#8217;y vois une ouverture. Premier volet d&#8217;une s&#233;rie : pourquoi, et ce que nous b&#226;tissons.</em></p><div><hr></div><p>Quelque chose de rare est en train de se produire : l&#8217;intelligence devient abondante. Les savoirs, les mod&#232;les, les outils autrefois rares, co&#251;teux et sous cl&#233; s&#8217;ouvrent &#224; tous, et deviennent chaque mois moins chers et plus performants.</p><p>On en fait surtout une comp&#233;tition &#8212; qui b&#226;tira le mod&#232;le le plus puissant, les &#201;tats-Unis ou la Chine, l&#8217;Europe redoutant tout bas d&#8217;avoir d&#233;j&#224; perdu. Je laisse cette course &#224; ceux qui la courent. (Sangeet Paul Choudary d&#233;fend, de fa&#231;on convaincante, l&#8217;id&#233;e qu&#8217;ils courent peut-&#234;tre la mauvaise : &#224; mesure que l&#8217;intelligence devient abondante, l&#8217;avantage se d&#233;place de celui qui poss&#232;de le mod&#232;le le plus puissant vers celui qui le transforme en valeur r&#233;elle. &#192; lire.) Mais, &#224; mon sens, il n&#8217;y a jamais eu de course, tout simplement.</p><p>Pour l&#8217;Afrique, cette abondance n&#8217;est ni une menace &#224; contenir ni une ligne d&#8217;arriv&#233;e o&#249; elle serait &#224; la tra&#238;ne. C&#8217;est une mati&#232;re premi&#232;re. Quand la capacit&#233; devient bon march&#233; et ouverte, le terrain est pr&#234;t pour le geste que l&#8217;Afrique a d&#233;j&#224; accompli &#8212; un bond en avant &#8212; et cette fois sur plusieurs fronts &#224; la fois : sant&#233;, agriculture, finance, infrastructures, gouvernance. C&#8217;est le moment de l&#8217;Afrique. Et nous avions d&#233;j&#224; commenc&#233; &#224; b&#226;tir pour lui : j&#8217;ai lanc&#233; <strong>leapfrogging.africa</strong>, con&#231;u le mod&#232;le et r&#233;uni les premiers partenaires bien avant qu&#8217;on ne pr&#233;sente tout cela comme une course. Nous ne cherchions &#224; battre personne. Nous avons vu l&#8217;ouverture.</p><p class="button-wrapper" data-attrs="{&quot;url&quot;:&quot;https://www.leapnotes.tech/subscribe?&quot;,&quot;text&quot;:&quot;Subscribe now&quot;,&quot;action&quot;:null,&quot;class&quot;:null}" data-component-name="ButtonCreateButton"><a class="button primary" href="https://www.leapnotes.tech/subscribe?"><span>Subscribe now</span></a></p><p>Un mot sur mon parcours, car il &#233;claire tout ce qui suit. Je suis africain. Je suis venu en Allemagne &#233;tudier l&#8217;informatique et l&#8217;IA, j&#8217;ai pass&#233; des ann&#233;es comme chercheur au DFKI &#8212; le plus grand centre de recherche ind&#233;pendant en IA au monde, et un pionnier du domaine depuis 1988 &#8212; et j&#8217;en ai lanc&#233; une spin-off d&#8217;IA financ&#233;e par du capital-risque, ce qui m&#8217;a fait conna&#238;tre de l&#8217;int&#233;rieur la technologie comme le march&#233;. leapfrogging.africa est n&#233; directement de cette exp&#233;rience. Le DFKI est, par vocation, tourn&#233; vers l&#8217;application : universit&#233;s, industrie et pouvoirs publics s&#8217;y coordonnent pour transformer la recherche en applications et en entreprises r&#233;elles &#8212; sans le moindre capital-risque au d&#233;part. Le capital est venu apr&#232;s, vers les entreprises que ce travail a fait na&#238;tre, la mienne parmi elles. C&#8217;est cet ordre qu&#8217;il faut retenir : on b&#226;tit d&#8217;abord le moteur qui rend les entreprises dignes d&#8217;&#234;tre financ&#233;es, et le capital &#8212; abondant et mobile &#8212; finit par affluer. C&#8217;est ce moteur que leapfrogging.africa entend adapter et d&#233;ployer &#224; travers l&#8217;Afrique &#8212; IA appliqu&#233;e et entrepreneuriat, &#224; l&#8217;&#233;chelle du continent.</p><h2>Application et coordination &#8212; pas les mod&#232;les de fondation</h2><p>Voici le c&#339;ur du sujet &#8212; et c&#8217;est une opportunit&#233;, pas une comp&#233;tition. Quand l&#8217;intelligence est abondante, la ressource rare &#8212; et pr&#233;cieuse &#8212; n&#8217;est plus le mod&#232;le. Ce sont deux choses men&#233;es ensemble, et bien men&#233;es. La premi&#232;re, l&#8217;<strong>application</strong> : mettre cette intelligence au service de probl&#232;mes r&#233;els, dans les secteurs et les usages qui comptent vraiment, en l&#8217;adaptant aux r&#233;alit&#233;s, aux forces et aux besoins d&#8217;un lieu pr&#233;cis. La seconde, la <strong>coordination</strong> : orchestrer les personnes, les outils et les syst&#232;mes pour la d&#233;ployer de fa&#231;on fiable et &#224; grande &#233;chelle. Un mod&#232;le, seul, ne fait qu&#8217;&#233;num&#233;rer ce qu&#8217;on pourrait faire. La valeur et l&#8217;impact reviennent &#224; celui qui l&#8217;applique au bon probl&#232;me et le m&#232;ne &#224; bien.</p><p>Cela red&#233;finit toute la question de savoir qui est &#171; en avance &#187;. Construire le plus grand mod&#232;le de fondation n&#8217;allait jamais &#234;tre le terrain de l&#8217;Afrique, et le poursuivre lui co&#251;terait une d&#233;cennie. Mais ce n&#8217;est jamais l&#224; que r&#233;side la valeur durable. Elle r&#233;side dans l&#8217;application adapt&#233;e, soutenue par une coordination solide &#8212; et ce travail commence &#224; peine, pour tout le monde. C&#8217;est aussi pourquoi tout ce que nous faisons, y compris notre recherche, est guid&#233; par l&#8217;application, et non l&#8217;inverse.</p><h2>Pourquoi c&#8217;est le terrain de l&#8217;Afrique</h2><p>C&#8217;est ici que le r&#233;cit habituel s&#8217;inverse. Les conditions que l&#8217;on range d&#8217;ordinaire sous &#171; l&#8217;Afrique est en retard &#187; sont, &#224; bien y regarder, autant d&#8217;atouts pour ce travail pr&#233;cis.</p><p>L&#8217;Afrique n&#8217;est pas prisonni&#232;re d&#8217;infrastructures h&#233;rit&#233;es &#8212; aucun investissement ancien &#224; d&#233;fendre, aucune vieille mani&#232;re de faire &#224; prot&#233;ger. Elle a une longue exp&#233;rience des environnements complexes o&#249; se croisent de nombreux acteurs &#8212; pr&#233;cis&#233;ment ce qu&#8217;exige la coordination. Et ses contraintes de ressources &#8212; puissance de calcul limit&#233;e, &#233;lectricit&#233; intermittente, bande passante r&#233;duite, pression permanente sur les co&#251;ts &#8212; ne font pas que la freiner : elles imposent l&#8217;innovation frugale dont le reste du monde, gourmand en &#233;nergie et riv&#233; &#224; ses centres de donn&#233;es, aura bient&#244;t besoin &#224; grande &#233;chelle.</p><p>Et ce n&#8217;est pas un v&#339;u pieux : c&#8217;est ce qu&#8217;on observe depuis vingt ans. L&#8217;Afrique n&#8217;a jamais tir&#233; de lignes t&#233;l&#233;phoniques en cuivre jusqu&#8217;&#224; chaque village ; elle est pass&#233;e directement au mobile et a mis un t&#233;l&#233;phone dans presque toutes les mains. Elle n&#8217;a pas attendu que les agences bancaires et les cartes de paiement parviennent &#224; tous ; elle a saut&#233; au mobile money &#8212; et, loin de se contenter de l&#8217;adopter, elle l&#8217;a invent&#233; et r&#233;volutionn&#233; pour le reste du monde. &#192; chaque fois, s&#8217;affranchir de l&#8217;&#233;tape h&#233;rit&#233;e qui n&#8217;a jamais servi qu&#8217;une poign&#233;e n&#8217;a pas laiss&#233; l&#8217;Afrique &#224; la tra&#238;ne : cela l&#8217;a plac&#233;e en t&#234;te. L&#8217;IA est le prochain bond du m&#234;me genre. Ne refaisons pas le chemin d&#8217;un autre sur des d&#233;cennies ; allons droit l&#224; o&#249; va la valeur &#8212; IA appliqu&#233;e, edge computing, solutions frugales qui marchent dans les conditions africaines, et marchent donc partout.</p><p>Le monde compte d&#233;j&#224; bien assez de gens qui peaufinent des mod&#232;les dans des centres de donn&#233;es. Ce qui lui manque cruellement, ce sont des gens capables de faire fonctionner l&#8217;intelligence l&#224; o&#249; l&#8217;infrastructure ne les aidera pas &#8212; sur un appareil aliment&#233; au solaire, hors ligne, toute la journ&#233;e, de fa&#231;on fiable. Ce n&#8217;est pas une sp&#233;cialit&#233; hypoth&#233;tique. C&#8217;est le quotidien, ici. Et c&#8217;est une position d&#233;fendable &#224; l&#8217;&#233;chelle mondiale, pas un lot de consolation.</p><h2>Ce qu&#8217;est vraiment leapfrogging.africa</h2><p>Soyons clairs sur ce dont il s&#8217;agit, car c&#8217;est plus vaste qu&#8217;une &#233;cole ou un produit. <strong>leapfrogging.africa</strong> est un mouvement pour b&#226;tir et relier des &#233;cosyst&#232;mes d&#8217;IA &#224; travers le continent &#8212; d&#8217;un pays &#224; l&#8217;autre, d&#8217;un secteur &#224; l&#8217;autre. L&#8217;unit&#233; d&#8217;ambition n&#8217;est ni un dipl&#244;m&#233; ni une application ; c&#8217;est un &#233;cosyst&#232;me qui transforme l&#8217;intelligence en solutions d&#233;ploy&#233;es et d&#233;tenues par des Africains &#8212; en agriculture, sant&#233;, finance, mines, infrastructures et gouvernance &#8212; puis un r&#233;seau de ces &#233;cosyst&#232;mes, coordonn&#233; par-del&#224; les fronti&#232;res.</p><p>C&#8217;est une ambition continentale ; la m&#233;thode doit donc &#234;tre tout l&#8217;inverse du grandiose. On ne lance pas un continent. On commence petit et on le prouve : quelques pays pilotes, une poign&#233;e de partenaires engag&#233;s, un n&#339;ud qui fonctionne &#8212; r&#233;el et mesur&#233;. Puis on r&#233;plique. Le but n&#8217;est pas d&#8217;assigner un r&#244;le &#224; chaque pays, mais de cr&#233;er une incitation : un pays &#8212; ou une r&#233;gion &#8212; peut devenir un p&#244;le d&#8217;excellence dans les secteurs qui comptent pour lui, aussi loin qu&#8217;il choisit de s&#8217;engager. L&#8217;agritech l&#224; o&#249; l&#8217;agriculture porte l&#8217;&#233;conomie, l&#8217;IA de sant&#233; l&#224; o&#249; le besoin est le plus aigu, la fintech b&#226;tie sur l&#8217;avance du continent en mobile money, les mines, la logistique ou l&#8217;&#233;nergie l&#224; o&#249; ce sont des forces locales &#8212; une IA adapt&#233;e aux conditions r&#233;elles, pas g&#233;n&#233;rique, avec l&#8217;edge computing et la frugalit&#233; en ressources comme force technique partag&#233;e. Chaque n&#339;ud est ancr&#233; dans les universit&#233;s et l&#8217;industrie locales, le tout reli&#233; en un ensemble coordonn&#233;. Piloter, valider, passer &#224; l&#8217;&#233;chelle. Le r&#233;seau est le produit.</p><p>C&#8217;est ici que la th&#232;se de la coordination cesse d&#8217;&#234;tre abstraite. Un tel &#233;cosyst&#232;me ne tient que si la capacit&#233; est d&#233;tenue localement &#8212; on ne peut pas coordonner ce qu&#8217;on ne sait pas ex&#233;cuter &#8212; et si les universit&#233;s, l&#8217;industrie et les pouvoirs publics y si&#232;gent en copropri&#233;taires et partenaires, non en bailleurs. Le capital et les entrepreneurs viennent ensuite &#8212; non comme point de d&#233;part, mais comme ce qu&#8217;un &#233;cosyst&#232;me qui fonctionne attire et produit : le moteur d&#8217;abord, et le capital est ce qu&#8217;un moteur cr&#233;dible finit par attirer, en investisseur plut&#244;t qu&#8217;en bailleur ; les entrepreneurs transforment des personnes form&#233;es et de la recherche en entreprises, en produits et en emplois. (Comment attirer effectivement ce capital &#8212; et pourquoi le moteur est l&#8217;&#233;l&#233;ment d&#233;cisif &#8212; fera l&#8217;objet d&#8217;un article &#224; part.) Rien de tout cela n&#8217;a &#224; &#234;tre invent&#233; de z&#233;ro : un mod&#232;le r&#233;unissant tous ces acteurs a fait ses preuves depuis des d&#233;cennies &#8212; en Europe, aux &#201;tats-Unis et ailleurs. L&#8217;id&#233;e n&#8217;est pas de l&#8217;importer tel quel ; c&#8217;est que nous n&#8217;avons pas &#224; passer quarante ans &#224; red&#233;couvrir ce qui marche. Le m&#234;me instinct de leapfrogging vaut pour les institutions : nous partons de l&#224; o&#249; ce mod&#232;le est arriv&#233;, et nous l&#8217;adaptons sans m&#233;nagement &#224; l&#8217;Afrique et au d&#233;fi du moment. C&#8217;est l&#8217;&#233;chafaudage sur lequel l&#8217;&#233;cosyst&#232;me s&#8217;appuie &#8212; important, mais pas la destination.</p><h2>L&#224; o&#249; tout commence : celles et ceux qui b&#226;tissent</h2><p>Un &#233;cosyst&#232;me continental repose malgr&#233; tout sur une ressource rare : des personnes r&#233;ellement capables de d&#233;ployer l&#8217;IA dans les conditions africaines. Sans elles, chaque couche au-dessus se loue &#224; quelqu&#8217;un d&#8217;autre. Le premier pas fondateur &#8212; la graine dont na&#238;t tout le reste &#8212; est donc le capital humain.</p><p>C&#8217;est le r&#244;le de notre initiative phare, <strong>ALIT Africa</strong> (the Applied Learning Institute for Technology) : en partenariat avec des universit&#233;s africaines &#233;tablies, former des personnes &#224; d&#233;ployer l&#8217;IA dans des conditions r&#233;elles &#8212; appliqu&#233;e aux secteurs locaux, pens&#233;e pour l&#8217;edge computing et la frugalit&#233; en ressources, adapt&#233;e aux langues et aux march&#233;s africains. C&#8217;est aussi le premier endroit o&#249; nous appliquons &#224; nous-m&#234;mes cette logique du bond &#8212; jusque dans la mani&#232;re d&#8217;apprendre. La ma&#238;trise y prime sur le temps pass&#233; sur les bancs : c&#8217;est le niveau atteint, non le calendrier, qui d&#233;cide qu&#8217;on est qualifi&#233;. ALIT, c&#8217;est l&#224; que leapfrogging.africa commence &#8212; pas ce qu&#8217;il est. (Cela fera l&#8217;objet du prochain article : ce qu&#8217;est ALIT, comment il fonctionne, et pourquoi l&#8217;&#233;ducation elle-m&#234;me est l&#8217;un des domaines o&#249; l&#8217;Afrique peut faire le bond.)</p><p>Car le choix de fond est net, et il se joue maintenant. La transformation port&#233;e par l&#8217;IA en Afrique sera soit d&#233;tenue et op&#233;r&#233;e par des institutions africaines, soit conc&#233;d&#233;e sous licence par des fournisseurs &#233;trangers ; soit b&#226;tie sur une expertise africaine, soit d&#233;pendante de consultants ext&#233;rieurs ; soit con&#231;ue pour les contextes africains, soit adapt&#233;e &#8212; maladroitement &#8212; d&#8217;ailleurs. Je sais quelle version je veux aider &#224; b&#226;tir.</p><p>C&#8217;est le premier d&#8217;une s&#233;rie r&#233;guli&#232;re &#8212; la strat&#233;gie, le mod&#232;le d&#8217;&#233;cosyst&#232;me, les partenaires, les obstacles, et le lent travail concret de la construction, en public, &#224; mesure qu&#8217;elle avance.</p><p>La question n&#8217;a jamais &#233;t&#233; de savoir si l&#8217;Afrique prend part &#224; l&#8217;&#232;re de l&#8217;IA. C&#8217;est de savoir si nous transformons l&#8217;abondance en capacit&#233;, &#224; grande &#233;chelle &#8212; nos pays b&#226;tissant ensemble vers un m&#234;me but plut&#244;t que de se courir les uns apr&#232;s les autres.</p><p>Si cela vous parle, suivez la s&#233;rie. Et si vous pr&#233;f&#233;rez b&#226;tir plut&#244;t que lire &#8212; c&#8217;est la meilleure des invitations : &#233;crivez-moi &#224; alit@leapfrogging.africa.</p><p><strong>Leap. Build. Own.</strong></p><div><hr></div><p><em>Il faut le reconna&#238;tre : le cadrage de la &#171; mauvaise course &#187; &#201;tats-Unis&#8211;Chine que j&#8217;&#233;voque en ouverture s&#8217;inspire de l&#8217;essai de Sangeet Paul Choudary et de son livre</em> Reshuffle*, qui soutiennent que l&#8217;effet durable de l&#8217;IA est de rendre l&#8217;intelligence abondante et de d&#233;placer l&#8217;avantage vers la coordination et l&#8217;ex&#233;cution. J&#8217;ai braqu&#233; la m&#234;me lentille sur un continent dont il ne parlait pas &#8212; et pour lequel nous avions d&#233;j&#224; commenc&#233; &#224; b&#226;tir.*</p><p class="button-wrapper" data-attrs="{&quot;url&quot;:&quot;https://www.leapnotes.tech/subscribe?&quot;,&quot;text&quot;:&quot;Subscribe now&quot;,&quot;action&quot;:null,&quot;class&quot;:null}" data-component-name="ButtonCreateButton"><a class="button primary" href="https://www.leapnotes.tech/subscribe?"><span>Subscribe now</span></a></p><p></p>]]></content:encoded></item><item><title><![CDATA[Les emplois que l’ère de l’IA rend possibles - Partie 2/2]]></title><description><![CDATA[Read this in English &#8594; [lien]]]></description><link>https://www.leapnotes.tech/p/les-emplois-que-lere-de-lia-rend</link><guid isPermaLink="false">https://www.leapnotes.tech/p/les-emplois-que-lere-de-lia-rend</guid><dc:creator><![CDATA[Kader Diagne]]></dc:creator><pubDate>Mon, 06 Jul 2026 12:02:35 GMT</pubDate><enclosure url="https://substackcdn.com/image/fetch/$s_!hy2c!,w_256,c_limit,f_auto,q_auto:good,fl_progressive:steep/https%3A%2F%2Fbucketeer-e05bbc84-baa3-437e-9518-adb32be77984.s3.amazonaws.com%2Fpublic%2Fimages%2F16ce3919-10c1-4ed0-b46f-720759955c7f_147x140.png" length="0" type="image/jpeg"/><content:encoded><![CDATA[<p><strong><span data-color="#ff0000" style="color: rgb(255, 0, 0);">Read this in English &#8594; [</span><a href="https://open.substack.com/pub/kaderdiagne/p/the-jobs-the-ai-age-makes-possible?r=39ps0&amp;utm_campaign=post-expanded-share&amp;utm_medium=web"><span data-color="#ff0000" style="color: rgb(255, 0, 0);">lien</span></a><span data-color="#ff0000" style="color: rgb(255, 0, 0);">]</span></strong></p><p><em>La premi&#232;re partie l&#8217;a montr&#233; : en Afrique, la question de l&#8217;IA et de l&#8217;emploi n&#8217;est ni la protection ni la d&#233;rive, mais la cr&#233;ation d&#233;lib&#233;r&#233;e. Place au concret &#8212; les nouvelles cat&#233;gories de travail que fait &#233;merger l&#8217;intelligence abondante, et la strat&#233;gie pour les b&#226;tir. Deuxi&#232;me partie sur deux.</em></p><div><hr></div><p>La premi&#232;re partie d&#233;fendait une th&#232;se : dans la plupart des pays d&#8217;Afrique, l&#8217;IA est moins une menace pour les emplois formels existants &#8212; ils sont relativement peu nombreux &#8212; que la mati&#232;re premi&#232;re pour b&#226;tir les nombreux autres dont le continent a besoin. Et l&#8217;approche qui convient n&#8217;est ni celle des &#201;tats-Unis, qui laissent faire le march&#233;, ni celle de la Chine, qui d&#233;fend les emplois existants, mais la cr&#233;ation d&#233;lib&#233;r&#233;e de travail nouveau.</p><p>Mais une th&#232;se n&#8217;est pas un plan. &#171; L&#8217;IA va cr&#233;er des emplois &#187; n&#8217;est qu&#8217;un slogan tant qu&#8217;on ne peut pas nommer ces emplois ni les mesures qui les font na&#238;tre. Voici donc le concret : &#224; quoi ressemble r&#233;ellement ce travail, et ce que contiendrait une strat&#233;gie pour le construire.</p><p>Partons du point cl&#233; de la premi&#232;re partie. Dans une grande partie de l&#8217;Afrique, ce qui manque n&#8217;est pas le travail &#224; accomplir &#8212; ce sont des personnes capables de l&#8217;accomplir de fa&#231;on productive. L&#8217;objectif n&#8217;est donc pas de prot&#233;ger un effectif, ni de courir apr&#232;s les emplois que l&#8217;externalisation voudra bien offrir. Il est de maximiser le nombre de personnes que l&#8217;intelligence abondante rend productives. Ce simple changement de perspective change ce que l&#8217;on construit.</p><h2>Le travail nouveau que le bond en avant fait na&#238;tre</h2><p>L&#8217;erreur est d&#8217;imaginer que l&#8217;IA ne fait que supprimer des emplois. L&#8217;histoire est claire : les technologies &#224; usage g&#233;n&#233;ral cr&#233;ent aussi des cat&#233;gories de travail que personne n&#8217;aurait su nommer &#224; l&#8217;avance &#8212; et l&#8217;Afrique peut construire ces cat&#233;gories <em>d&#233;lib&#233;r&#233;ment</em>, au lieu d&#8217;attendre de voir lesquelles appara&#238;tront. Plusieurs se dessinent d&#233;j&#224; :</p><ul><li><p><strong>Int&#233;grateurs d&#8217;IA pour les petites entreprises</strong> &#8212; celles et ceux qui d&#233;ploient, adaptent et font tourner au quotidien les outils d&#8217;IA qui rendent les PME productives. La plupart des petites entreprises n&#8217;engageront jamais un grand cabinet de conseil ; quelqu&#8217;un de local fera le travail technique, le conseil et la maintenance &#8212; souvent la m&#234;me personne. Et il y en a &#233;norm&#233;ment &#224; faire.</p></li><li><p><strong>Sp&#233;cialistes des langues locales</strong> &#8212; celles et ceux qui adaptent, &#233;valuent et maintiennent les syst&#232;mes en wolof, bambara, yoruba, swahili, lingala, amharique, et les centaines de langues que les mod&#232;les mondiaux n&#233;gligent. Bien men&#233;e, c&#8217;est une activit&#233; qualifi&#233;e et durable &#8212; l&#8217;inverse de l&#8217;annotation ponctuelle et mal pay&#233;e contre laquelle la premi&#232;re partie mettait en garde. C&#8217;est la couche de conception et de qualit&#233;, celle qui se situe <em>au-dessus</em> de l&#8217;&#233;tiquetage brut des donn&#233;es, et non l&#8217;&#233;tiquetage lui-m&#234;me.</p></li><li><p><strong>Conseillers agricoles et op&#233;rateurs d&#8217;outils d&#8217;IA</strong> &#8212; celles et ceux qui apportent aux exploitations l&#8217;irrigation intelligente, les capteurs de culture et les robots de terrain, et qui aident les agriculteurs &#224; tirer parti de ce que ces outils r&#233;v&#232;lent. Avec cet appui, un seul agriculteur peut travailler une surface qui en exigeait auparavant plusieurs.</p></li><li><p><strong>Agents de sant&#233; assist&#233;s par l&#8217;IA</strong> &#8212; un nouveau profil de soignant pour un nouveau mod&#232;le de soins : des points de sant&#233; qui recueillent les donn&#233;es au plus pr&#232;s des habitants, reli&#233;s par la t&#233;l&#233;m&#233;decine et l&#8217;IA &#224; des sp&#233;cialistes parfois &#233;loign&#233;s. Ce mod&#232;le apporte un bon diagnostic et un bon suivi aux zones rurales sans construire de grands h&#244;pitaux partout &#8212; et il faut des personnes form&#233;es pour recueillir les donn&#233;es, faire fonctionner les appareils et accompagner les patients, ainsi que les techniciens qui font tenir l&#8217;ensemble.</p></li><li><p><strong>Gestionnaires de coop&#233;ratives de donn&#233;es</strong> &#8212; un r&#244;le plus r&#233;cent et moins &#233;prouv&#233; : aider les coop&#233;ratives, les municipalit&#233;s et les associations &#224; collecter, gouverner et exploiter leurs propres donn&#233;es au lieu de les c&#233;der. &#201;mergent plut&#244;t qu&#8217;&#233;tabli &#8212; mais le besoin est d&#233;j&#224; visible.</p></li></ul><p>Et sous-tendant toutes ces cat&#233;gories, la plus vaste de toutes : les <strong>micro-entrepreneurs outill&#233;s par l&#8217;IA</strong> &#8212; des personnes qui utilisent l&#8217;intelligence abondante pour proposer de la traduction, du design, de la comptabilit&#233;, du marketing, des services touristiques, de l&#8217;aide aux achats, et cent autres choses ; une seule personne pouvant d&#233;sormais accomplir le travail qui exigeait nagu&#232;re une petite &#233;quipe. Ce n&#8217;est pas la pr&#233;vision d&#8217;un chiffre pr&#233;cis. C&#8217;est la forme d&#8217;une opportunit&#233; &#8212; et son ampleur est &#224; la mesure du besoin.</p><p>Remarquez ce que la plupart ont en commun. Elles se situent au point de rencontre entre l&#8217;intelligence et une r&#233;alit&#233; locale &#224; laquelle il faut l&#8217;adapter &#8212; et ce point est difficile &#224; d&#233;localiser. Toutes les t&#226;ches ne restent pas locales : le travail de donn&#233;es brut file d&#233;j&#224; vers le pays le moins cher, comme le rappelait la premi&#232;re partie. Mais la couche de la <em>relation</em> &#8212; conna&#238;tre le secteur, la langue, la communaut&#233;, les conditions, et &#234;tre digne de confiance pour d&#233;ployer au sein de tout cela &#8212; est bien plus difficile &#224; piloter de loin. C&#8217;est l&#224; que se trouve le travail durable et mieux r&#233;mun&#233;r&#233;.</p><p>Une r&#233;serve : beaucoup de ces activit&#233;s commencent comme du travail &#224; la t&#226;che ou des gagne-pain informels, non comme des postes salari&#233;s &#8212; et l&#8217;avertissement de la premi&#232;re partie sur le bas extractif du march&#233; vaut ici aussi. Le r&#244;le d&#8217;une strat&#233;gie est d&#8217;aider ce travail &#224; progresser vers la stabilit&#233; et la propri&#233;t&#233;, non de se f&#233;liciter de son volume brut. C&#8217;est pr&#233;cis&#233;ment &#224; cela que sert une strat&#233;gie, par opposition &#224; un slogan.</p><h2>Industrialiser par l&#8217;intelligence &#8212; usines comprises</h2><p>Prenons du recul : tout cela d&#233;passe une simple liste d&#8217;emplois &#8212; c&#8217;est une autre mani&#232;re de se d&#233;velopper. L&#8217;Afrique a toujours besoin d&#8217;industrie : elle poss&#232;de des minerais et des r&#233;coltes qu&#8217;il faut transformer, et le faire sur place, au lieu d&#8217;exporter les mati&#232;res premi&#232;res pour importer les produits finis, c&#8217;est ainsi que la valeur reste sur le continent. Le bond en avant ne consiste donc pas &#224; se passer des usines. Il consiste &#224; ne pas refaire le lent chemin qui y m&#232;ne.</p><p>Cet ancien chemin suivait un ordre immuable : b&#226;tir d&#8217;abord l&#8217;industrie lourde et les grandes usines comme principal moteur d&#8217;emplois ; des d&#233;cennies plus tard, d&#233;velopper les services ; plus tard encore, une large vague d&#8217;entrepreneuriat &#8212; chaque &#233;tape attendant la pr&#233;c&#233;dente. L&#8217;intelligence abondante supprime la n&#233;cessit&#233; d&#8217;attendre. L&#8217;Afrique peut b&#226;tir des usines intelligentes qui transforment les ressources locales avec moins de personnel, mais plus qualifi&#233;, et d&#233;velopper en m&#234;me temps une vaste couche de services et de petites entreprises outill&#233;s par l&#8217;IA &#8212; au lieu de consid&#233;rer les grandes usines comme le seul lieu d&#8217;o&#249; vient le travail.</p><p>Sous tout cela se joue un basculement : du grand et centralis&#233; vers le petit et local. Longtemps, l&#8217;&#233;chelle a rim&#233; avec la taille &#8212; une usine g&#233;ante, un h&#244;pital central, une grande &#233;cole &#8212; tout circulant vers ce centre et depuis ce centre, &#224; grands frais et &#224; grand risque. L&#8217;IA, l&#8217;automatisation et la robotique changent la donne. Elles rendent les petites unit&#233;s locales assez efficaces pour valoir la peine d&#8217;&#234;tre construites : la production peut alors s&#8217;installer l&#224; o&#249; sont les ressources, et les services l&#224; o&#249; les gens en ont r&#233;ellement besoin, au lieu de tout transporter vers un centre et depuis lui. Pensez aux boulangeries : non pas une usine g&#233;ante qui cuit pour tout un pays et exp&#233;die le pain partout, mais une boulangerie dans chaque quartier, qui produit ce dont ce quartier a besoin. La m&#234;me chose devient possible pour les usines, les cliniques et les &#233;coles &#8212; un r&#233;seau de petites unit&#233;s locales plut&#244;t que quelques g&#233;ants.</p><p>Et petit ne veut pas dire isol&#233;. Les unit&#233;s sont connect&#233;es et partagent ce qu&#8217;elles apprennent : les robots d&#8217;une usine transmettent leur exp&#233;rience aux robots d&#8217;une autre ; des exploitations sous des climats diff&#233;rents &#233;changent ce qui marche ; des cliniques mettent en commun ce qu&#8217;elles observent. Un site central unique ne conna&#238;t jamais qu&#8217;une seule r&#233;alit&#233;. Un r&#233;seau de sites locaux apprend de plusieurs &#224; la fois &#8212; chaleur, froid, poussi&#232;re, sols et usages diff&#233;rents &#8212; et s&#8217;am&#233;liore d&#8217;autant plus vite. D&#233;centralis&#233; mais coordonn&#233; l&#8217;emporte sur grand et centralis&#233;. Cette coordination &#8212; de nombreuses unit&#233;s locales qui apprennent comme une seule &#8212; est la vraie force sur laquelle repose toute cette approche.</p><p>On retrouve ce mouvement dans tous les secteurs. Dans l&#8217;agriculture : irrigation intelligente, capteurs et robots sur de nombreuses petites exploitations plut&#244;t qu&#8217;une ferme g&#233;ante. Dans la sant&#233; : des points de collecte de donn&#233;es locaux, la t&#233;l&#233;m&#233;decine et des objets connect&#233;s qui accompagnent les gens au jour le jour &#8212; g&#233;rer les maladies chroniques, orienter la nutrition, rep&#233;rer t&#244;t les probl&#232;mes &#8212; au lieu de quelques h&#244;pitaux centraux co&#251;teux. &#192; chaque fois, la logique est la m&#234;me : appliquer l&#8217;intelligence maintenant, et placer la capacit&#233; l&#224; o&#249; sont les gens et les ressources, plut&#244;t que de b&#226;tir d&#8217;abord de co&#251;teuses infrastructures du XXe si&#232;cle et de tout centraliser autour d&#8217;elles. Bien men&#233;e, cette approche est &#224; la fois meilleure et moins ch&#232;re &#8212; et, dans la sant&#233;, elle maintient une main-d&#8217;&#339;uvre plus saine et plus productive tout en &#233;vitant l&#8217;explosion des co&#251;ts qui p&#232;se sur les &#233;conomies plus riches.</p><p>Voil&#224; ce que &#171; leapfrogging &#187; veut vraiment dire ici : non pas sauter l&#8217;industrie, mais la b&#226;tir sous une forme nouvelle &#8212; plus petite, locale, connect&#233;e, et tout &#224; la fois. L&#8217;Afrique peut atteindre une &#233;conomie moderne &#8212; usines, services et entrepreneuriat ensemble &#8212; sans passer des d&#233;cennies &#224; y grimper &#233;tape par &#233;tape. Cela s&#8217;appuie sur une force d&#233;j&#224; pr&#233;sente sur le continent, l&#8217;entrepreneuriat informel, au lieu de la contrarier ; tout le monde ne devient pas salari&#233;, et beaucoup deviennent entrepreneurs. L&#8217;occasion est de faire de l&#8217;intelligence abondante un chemin plus rapide et plus intelligent qu&#8217;aucun de ceux emprunt&#233;s jusqu&#8217;ici.</p><h2>La strat&#233;gie qui la construit</h2><p>Rien de tout cela n&#8217;est automatique &#8212; c&#8217;&#233;tait l&#8217;avertissement de la premi&#232;re partie. La cr&#233;ation d&#233;lib&#233;r&#233;e exige des mesures d&#233;lib&#233;r&#233;es. Une v&#233;ritable strat&#233;gie d&#8217;emploi par l&#8217;IA, par opposition &#224; une strat&#233;gie de protection, mobiliserait des leviers comme ceux-ci :</p><ul><li><p><strong>Programmes nationaux d&#8217;apprentissage et de formation professionnelle en IA</strong> &#8212; traiter les comp&#233;tences d&#8217;IA appliqu&#233;e comme les &#233;poques pass&#233;es traitaient les m&#233;tiers manuels.</p></li><li><p><strong>Services de vulgarisation de l&#8217;IA</strong> &#8212; calqu&#233;s sur les r&#233;seaux de vulgarisation agricole d&#233;j&#224; en place, pour mettre le conseil &#224; la port&#233;e des agriculteurs et des petites entreprises.</p></li><li><p><strong>Un corps de d&#233;ploiement de l&#8217;IA</strong> &#8212; une capacit&#233; organis&#233;e pour amener l&#8217;IA dans les services publics et les PME, cr&#233;ant au passage les emplois d&#8217;int&#233;grateurs &#233;voqu&#233;s plus haut.</p></li><li><p><strong>Une commande publique favorable aux startups d&#8217;IA locales</strong> &#8212; utiliser la demande de l&#8217;&#201;tat pour d&#233;velopper les capacit&#233;s nationales plut&#244;t que de les importer.</p></li><li><p><strong>Des fonds d&#8217;entrepreneuriat en IA</strong> &#8212; des capitaux dirig&#233;s vers la couche des micro et petites entreprises, l&#224; o&#249; se fera l&#8217;essentiel de la cr&#233;ation d&#8217;emplois.</p></li><li><p><strong>Des mod&#232;les de fondation et d&#8217;application en langues locales</strong> &#8212; trait&#233;s comme une infrastructure publique, car tout le reste en d&#233;pend.</p></li><li><p><strong>Une alphab&#233;tisation massive &#224; l&#8217;IA et par l&#8217;IA</strong> &#8212; la couche de base qui permet &#224; toute la population de participer, et non &#224; une mince &#233;lite.</p></li></ul><p>Rien d&#8217;exceptionnel dans tout cela. Ce sont les instruments ordinaires de la politique industrielle, dirig&#233;s vers une cible que d&#8217;autres ne visent pas &#8212; ni prot&#233;ger les emplois d&#8217;hier, ni attendre le march&#233;, mais b&#226;tir les conditions pour que le travail de demain existe.</p><h2>L&#224; o&#249; la strat&#233;gie prend racine</h2><p>Chaque &#233;l&#233;ment de cette liste repose sur le m&#234;me socle &#8212; celui de tout ce que fait <strong>leapfrogging.africa</strong> : des personnes r&#233;ellement capables de d&#233;ployer l&#8217;IA dans les conditions africaines. Sans elles, les emplois d&#8217;int&#233;grateurs, le corps de d&#233;ploiement, les mod&#232;les en langues locales et les micro-entreprises se louent tous ailleurs &#8212; et la strat&#233;gie retombe dans la simple consommation.</p><p>C&#8217;est le r&#244;le d&#8217;<strong>ALIT Africa</strong> &#224; l&#8217;extr&#233;mit&#233; appliqu&#233;e, et celui de l&#8217;&#233;ducation comme pilier &#224; part enti&#232;re de leapfrogging.africa en dessous &#8212; du primaire et du secondaire &#224; la formation professionnelle et continue, et con&#231;ue comme l&#8217;intelligence abondante le permet d&#233;sormais : adaptative, fond&#233;e sur la ma&#238;trise, o&#249; c&#8217;est le niveau atteint, et non le calendrier, qui d&#233;cide du moment o&#249; quelqu&#8217;un est qualifi&#233;. L&#8217;&#233;ducation n&#8217;est pas &#224; c&#244;t&#233; de la strat&#233;gie d&#8217;emploi. Elle <em>est</em>la strat&#233;gie d&#8217;emploi, vue depuis sa source. C&#8217;est le pilier &#171; emploi &#187; de tout l&#8217;argumentaire de leapfrogging.africa : si l&#8217;intelligence devient un bien courant, alors l&#8217;avantage comp&#233;titif de l&#8217;Afrique, c&#8217;est le nombre de personnes qu&#8217;elle peut rendre capables de l&#8217;appliquer &#8212; travailleurs, op&#233;rateurs, b&#226;tisseurs et entrepreneurs, &#224; grande &#233;chelle.</p><h2>Ce que personne n&#8217;a jamais fait</h2><p>Un dernier point, et c&#8217;est le plus important, car il est facile de le perdre de vue dans le quotidien.</p><p>Aucun pays dans l&#8217;histoire ne s&#8217;est industrialis&#233; en disposant d&#8217;une intelligence abondante d&#232;s le d&#233;part. Ni le Royaume-Uni. Ni l&#8217;Allemagne. Ni les &#201;tats-Unis. Ni la Chine. Chacun a d&#8217;abord b&#226;ti son &#233;conomie moderne, et n&#8217;a d&#251; y ajouter l&#8217;intelligence que plus tard &#8212; ce qui est pr&#233;cis&#233;ment ce qu&#8217;une grande partie du monde peine aujourd&#8217;hui &#224; faire.</p><p>L&#8217;Afrique pourrait &#234;tre la premi&#232;re &#224; b&#226;tir de larges pans d&#8217;une &#233;conomie moderne avec une intelligence abondante disponible d&#232;s le premier jour. Ce n&#8217;est pas un lot de consolation pour &#234;tre arriv&#233;e en retard. C&#8217;est une position de d&#233;part qu&#8217;aucune nation industrialis&#233;e avant elle n&#8217;a jamais connue.</p><p>D&#8217;autres cherchent comment prot&#233;ger la main-d&#8217;&#339;uvre d&#8217;hier, ou esp&#232;rent que le march&#233; tranchera la question &#224; leur place. L&#8217;Afrique a une t&#226;che plus rare, et une chance plus rare : concevoir l&#8217;avenir avant que l&#8217;ancien mod&#232;le ne s&#8217;installe &#8212; et faire en sorte que les emplois qu&#8217;elle cr&#233;e soient d&#233;tenus ici.</p><p><strong>Leap. Build. Own.</strong></p><p><strong>[<a href="https://www.leapnotes.tech/p/le-debat-sur-lia-et-lemploi-a-deux?r=39ps0">Lien Partie 1</a>]</strong></p>]]></content:encoded></item><item><title><![CDATA[Le débat sur l’IA et l’emploi a deux réponses. L’Afrique en a besoin d’une troisième - Partie 1/2]]></title><description><![CDATA[Pour l'Afrique, l'IA n'est pas une menace pour l'emploi. C'est le moyen d'en cr&#233;er.]]></description><link>https://www.leapnotes.tech/p/le-debat-sur-lia-et-lemploi-a-deux</link><guid isPermaLink="false">https://www.leapnotes.tech/p/le-debat-sur-lia-et-lemploi-a-deux</guid><dc:creator><![CDATA[Kader Diagne]]></dc:creator><pubDate>Mon, 06 Jul 2026 11:50:18 GMT</pubDate><enclosure url="https://substackcdn.com/image/fetch/$s_!hy2c!,w_256,c_limit,f_auto,q_auto:good,fl_progressive:steep/https%3A%2F%2Fbucketeer-e05bbc84-baa3-437e-9518-adb32be77984.s3.amazonaws.com%2Fpublic%2Fimages%2F16ce3919-10c1-4ed0-b46f-720759955c7f_147x140.png" length="0" type="image/jpeg"/><content:encoded><![CDATA[<p><strong><span data-color="#ff0000" style="color: rgb(255, 0, 0);">Read this in English &#8594; [</span><a href="https://open.substack.com/pub/kaderdiagne/p/the-ai-jobs-debate-has-two-answers?r=39ps0&amp;utm_campaign=post-expanded-share&amp;utm_medium=web"><span data-color="#ff0000" style="color: rgb(255, 0, 0);">link</span></a><span data-color="#ff0000" style="color: rgb(255, 0, 0);">]</span></strong></p><p><em>Les deux superpuissances de l&#8217;IA g&#232;rent son effet sur l&#8217;emploi de fa&#231;ons tr&#232;s diff&#233;rentes : l&#8217;une laisse largement faire le march&#233;, l&#8217;autre pilote depuis le sommet. Aucune des deux ne convient &#224; l&#8217;Afrique, car toutes deux supposent un vaste stock d&#8217;emplois formels qui, ici, n&#8217;a jamais &#233;t&#233; b&#226;ti &#224; la m&#234;me &#233;chelle. Ce n&#8217;est pas une raison pour laisser filer, ni pour se mettre sur la d&#233;fensive. C&#8217;est une raison d&#8217;avoir une strat&#233;gie. Premi&#232;re partie sur deux : comment faire de l&#8217;IA, non plus une menace pour l&#8217;emploi, mais ce qui finit enfin par en cr&#233;er &#8212; inspir&#233;e d&#8217;un r&#233;cent article de Katrin Bennhold dans le New York Times sur la Chine et l&#8217;emploi face &#224; l&#8217;IA.</em></p><div><hr></div><p>Tout est parti d&#8217;un article de Katrin Bennhold dans le New York Times. Elle y d&#233;crit deux mani&#232;res tr&#232;s diff&#233;rentes dont l&#8217;effet de l&#8217;IA sur l&#8217;emploi se d&#233;ploie chez les deux superpuissances de l&#8217;IA.</p><p>Aux &#201;tats-Unis, la r&#233;ponse consiste essentiellement &#224; ne pas intervenir. Les grands laboratoires courent apr&#232;s la superintelligence &#8212; des machines cens&#233;es d&#233;passer l&#8217;intelligence humaine, et pas seulement l&#8217;&#233;galer &#8212; pendant que l&#8217;&#201;tat reste en retrait, en supposant que le march&#233; absorbera le choc de lui-m&#234;me. Il n&#8217;y a pas vraiment de plan : laisser faire le march&#233; <em>est</em> la strat&#233;gie. Appelons cela le laissez-faire.</p><p>En Chine, c&#8217;est tout autre chose. P&#233;kin a inscrit la question de l&#8217;emploi dans son plan quinquennal et est pass&#233; &#224; l&#8217;acte : pression sur les entreprises pour &#233;viter les licenciements, salari&#233;s licenci&#233;s soutenus devant les tribunaux, formations financ&#233;es, et m&#234;me l&#8217;&#233;mergence d&#8217;un champ universitaire pour d&#233;terminer qui cr&#233;e la valeur une fois les machines arriv&#233;es. Appelons cela la protection : l&#8217;&#201;tat g&#232;re activement l&#8217;effet sur la main-d&#8217;&#339;uvre existante.</p><p>Deux approches tr&#232;s diff&#233;rentes &#8212; mais le m&#234;me postulat de d&#233;part. Toutes deux tiennent pour acquis un vaste stock d&#8217;emplois formels (cha&#238;nes de production, centres d&#8217;appels, postes de bureau) et se jouent autour de ce que l&#8217;IA lui fait subir. L&#8217;une laisse ce stock &#234;tre boulevers&#233;, l&#8217;autre le d&#233;fend. Dans les deux cas, c&#8217;est ce stock qui est au c&#339;ur du d&#233;bat.</p><p>La plupart des pays d&#8217;Afrique n&#8217;ont jamais b&#226;ti ce stock d&#8217;emplois formels &#224; une &#233;chelle comparable. Et cela change la question.</p><h2>L&#8217;Afrique part d&#8217;un autre point de d&#233;part</h2><p>Dans la plus grande partie du continent, le probl&#232;me central de l&#8217;emploi n&#8217;est pas <em>trop peu d&#8217;emplois apr&#232;s l&#8217;IA</em>. C&#8217;est <em>trop peu d&#8217;emplois, tout court</em> &#8212; et une population jeune qui arrive plus vite qu&#8217;aucune &#233;conomie formelle ne parvient &#224; l&#8217;absorber.</p><p>Pr&#232;s de 85 % de l&#8217;emploi en Afrique subsaharienne est informel &#8212; petite agriculture, commerce de rue, services de proximit&#233; &#8212; contre une moyenne mondiale d&#8217;environ 58 %. Chaque ann&#233;e, plus de 10 millions de jeunes entrent sur le march&#233; du travail africain, et la croissance actuelle ne cr&#233;e qu&#8217;environ 3 millions d&#8217;emplois formels pour les accueillir. Selon une estimation, la seule Afrique subsaharienne devra cr&#233;er de l&#8217;ordre de 15 millions d&#8217;emplois par an d&#8217;ici 2030 rien que pour suivre le rythme des nouveaux arrivants &#8212; et d&#8217;ici 2050, la r&#233;gion ajoutera plus de 620 millions de personnes &#224; sa population en &#226;ge de travailler, la plus forte expansion de ce type au monde.</p><p>Un d&#233;bat construit autour de ce que l&#8217;IA fait subir &#224; un stock d&#8217;emplois formels &#8212; le laisser &#234;tre boulevers&#233; ou le d&#233;fendre &#8212; est en r&#233;alit&#233; un d&#233;bat pour les grandes &#233;conomies industrialis&#233;es. Il r&#233;pond &#224; une question que ne se posent ni une grande partie de l&#8217;Afrique, ni d&#8217;autres &#233;conomies jeunes et largement informelles.</p><h2>Pourquoi aucune des deux r&#233;ponses ne convient</h2><p>Transposez chacune de ces approches &#224; l&#8217;Afrique, et le d&#233;calage saute aux yeux.</p><p>Le laissez-faire suppose un march&#233; d&#8217;emplois formels qui fonctionne, que l&#8217;IA viendrait rebattre. L&#224; o&#249; ce march&#233; est mince, &#171; laisser faire &#187; ne produit pas une redistribution certes chaotique mais dynamique. Cela produit une d&#233;rive : l&#8217;IA &#233;trang&#232;re est consomm&#233;e, la valeur qu&#8217;elle cr&#233;e est capt&#233;e ailleurs, et l&#8217;&#233;conomie locale h&#233;rite du bas extractif de la cha&#238;ne, sans le haut productif. Ici, la passivit&#233; n&#8217;est pas neutre. C&#8217;est un choix : celui de laisser passer l&#8217;occasion.</p><p>La protection &#233;choue pour la raison inverse : ici, il y a bien moins &#224; prot&#233;ger. D&#233;fendre le stock plus r&#233;duit d&#8217;emplois formels qui existent bel et bien passe &#224; c&#244;t&#233; de l&#8217;essentiel &#8212; de la m&#234;me mani&#232;re, et pour la m&#234;me raison, que &#171; poss&#233;der le mod&#232;le le plus puissant &#187; &#233;tait la mauvaise course. Le vrai d&#233;fi n&#8217;est pas d&#8217;amortir les emplois perdus &#224; cause de l&#8217;IA ; c&#8217;est de cr&#233;er les nombreux autres dont l&#8217;&#233;conomie a encore besoin.</p><p>Reste une troisi&#232;me posture, et c&#8217;est la bonne : ni la d&#233;rive, ni la d&#233;fense, mais la <strong>cr&#233;ation d&#233;lib&#233;r&#233;e</strong>. La question de l&#8217;IA et de l&#8217;emploi, en Afrique, n&#8217;est pas &#171; combien d&#8217;emplois vont survivre ? &#187;. C&#8217;est &#171; combien pouvons-nous en construire &#8212; et allons-nous le faire d&#233;lib&#233;r&#233;ment ? &#187;.</p><p>Ce mot &#8212; <em>d&#233;lib&#233;r&#233;ment</em> &#8212; change tout. La phrase la plus utile de l&#8217;article du Times ne porte pas sur les m&#233;thodes chinoises, dont l&#8217;essentiel ne se transpose pas de toute fa&#231;on. C&#8217;est sa conclusion : les d&#233;cideurs ont une marge de man&#339;uvre. La direction que prend cette technologie est un choix, pas la m&#233;t&#233;o. Et cette marge de man&#339;uvre va &#224; l&#8217;encontre des deux autres approches &#224; la fois &#8212; la d&#233;rive du laissez-faire comme le r&#233;flexe pass&#233;iste de la protection. Elle m&#232;ne &#224; la strat&#233;gie. Ce que contient cette strat&#233;gie fera l&#8217;objet de la deuxi&#232;me partie. La premi&#232;re &#233;tablit qu&#8217;il faut en avoir une.</p><p>Soyons clairs sur les limites de cette affirmation, car elles comptent. Je ne dis pas qu&#8217;<em>aucun</em> emploi africain n&#8217;est menac&#233; &#8212; certains le sont, et j&#8217;y viendrai franchement. Je dis o&#249; se situe l&#8217;essentiel. Pour un continent dont le d&#233;fi central est un <em>manque</em> d&#8217;emplois, et non leur destruction soudaine, la posture qui convient n&#8217;est ni passive ni d&#233;fensive. Elle est proactive.</p><h2>L&#8217;intelligence devient une infrastructure</h2><p>Commen&#231;ons par comprendre pourquoi cette cr&#233;ation est m&#234;me possible &#8212; pourquoi l&#8217;IA abondante est ici un levier plut&#244;t qu&#8217;une menace.</p><p>L&#8217;intention affich&#233;e de la Chine est d&#8217;utiliser l&#8217;IA pour <em>augmenter</em> ses travailleurs plut&#244;t que les remplacer, en rendant plus productifs ceux qui occupent d&#233;j&#224; un emploi formel. Reprenez cette intention et dirigez-la vers une autre cible : non pas les personnes en emploi, mais l&#8217;immense masse des <strong>sous-employ&#233;s</strong> &#8212; le petit exploitant agricole, le commer&#231;ant de march&#233;, le micro-entrepreneur, l&#8217;agent de sant&#233; communautaire, le petit dispensaire. Ils sont ici bien plus nombreux que les employ&#233;s de bureau, et c&#8217;est tout l&#8217;enjeu.</p><p>De m&#234;me que les routes, l&#8217;&#233;lectricit&#233; et l&#8217;internet sont chacun devenus des infrastructures sur lesquelles une &#233;conomie enti&#232;re pouvait s&#8217;appuyer, l&#8217;intelligence abondante devient une infrastructure <em>cognitive</em> &#8212; et elle arrive sur l&#8217;&#233;conomie informelle comme une capacit&#233; brute. Un diagnostic de maladie des cultures sur un simple smartphone transforme l&#8217;intuition d&#8217;un agriculteur en d&#233;cision, et permet &#224; un seul conseiller agricole d&#8217;atteindre bien plus de producteurs qu&#8217;auparavant. Un conseil en langue locale transforme un commer&#231;ant priv&#233; de donn&#233;es de march&#233; en quelqu&#8217;un qui peut fixer ses prix et planifier. Un agent de sant&#233; assist&#233; par l&#8217;IA peut prendre en charge, en toute s&#233;curit&#233;, des cas qui exigeaient auparavant d&#8217;orienter le patient &#224; des centaines de kilom&#232;tres. Rien de tout cela n&#8217;est un emploi salari&#233; dans une tour de bureaux. C&#8217;est un multiplicateur de productivit&#233; pos&#233; sur l&#8217;&#233;conomie qui fait d&#233;j&#224; vivre la majeure partie du continent &#8212; un plancher que l&#8217;on rel&#232;ve pour ceux qui s&#8217;y tiennent.</p><p>Cela r&#233;v&#232;le la nature r&#233;elle de la p&#233;nurie. Les gouvernements demandent d&#8217;ordinaire : <em>comment cr&#233;er des emplois ?</em> La bonne question est plut&#244;t : <em>comment accro&#238;tre la capacit&#233; productive ?</em> &#8212; car, dans une grande partie de l&#8217;Afrique, ce qui manque n&#8217;est pas le travail &#224; accomplir, mais des personnes capables de l&#8217;accomplir de fa&#231;on productive. Quand l&#8217;intelligence devient une infrastructure, des millions de gens jusque-l&#224; exclus du travail productif faute de formation, d&#8217;outils ou d&#8217;institutions peuvent soudain y acc&#233;der. La ressource rare n&#8217;est plus l&#8217;intelligence : c&#8217;est l&#8217;humain capable de l&#8217;appliquer.</p><h2>Le vrai bond en avant : s&#8217;affranchir des mod&#232;les d&#233;pass&#233;s, pas de la technologie</h2><p>C&#8217;est ici que le mot <em>leapfrogging</em> doit &#234;tre employ&#233; avec pr&#233;cision, car sa version courante est trop &#233;troite.</p><p>La plupart des gens entendent &#171; leapfrogging &#187; et pensent : sauter le t&#233;l&#233;phone fixe, sauter les agences bancaires, sauter l&#8217;ordinateur de bureau &#8212; passer directement &#224; la technologie la plus r&#233;cente. Vrai, jusqu&#8217;&#224; un certain point. Mais le vrai bond en avant ne consiste pas &#224; sauter une <em>technologie</em>. Il consiste &#224; s&#8217;affranchir d&#8217;un <em>mod&#232;le d&#8217;organisation d&#233;pass&#233;</em>.</p><p>Le monde d&#233;velopp&#233; a b&#226;ti ses &#233;conomies en cr&#233;ant d&#8217;abord des millions d&#8217;emplois de bureau, administratifs et de back-office &#8212; et il d&#233;pense aujourd&#8217;hui une &#233;nergie politique consid&#233;rable &#224; chercher comment les automatiser sans provoquer de rupture sociale. L&#8217;Afrique n&#8217;a pas &#224; entrer dans ce pi&#232;ge pour devoir ensuite en ressortir. Elle peut sauter toute l&#8217;&#233;tape qui consiste &#224; fabriquer des emplois administratifs &#224; faible valeur pour les d&#233;manteler une g&#233;n&#233;ration plus tard, et construire d&#8217;embl&#233;e un travail &#224; plus forte valeur, port&#233; par l&#8217;IA. C&#8217;est un saut bien plus grand que d&#8217;enjamber un c&#226;ble dans le sol. C&#8217;est sauter tout un d&#233;tour &#233;conomique.</p><h2>La couche humaine est le moteur d&#8217;emplois</h2><p>Et d&#233;ployer l&#8217;IA dans les conditions africaines ne se contente pas de rendre les travailleurs existants plus productifs. Cela <em>cr&#233;e des cat&#233;gories de travail enti&#232;rement nouvelles</em> &#8212; et le mod&#232;le existe d&#233;j&#224;.</p><p>Chaque saut africain pr&#233;c&#233;dent a fait exactement cela. Le mobile money en est l&#8217;exemple le plus clair. Il n&#8217;a pas seulement num&#233;ris&#233; les paiements ; il a b&#226;ti un vaste r&#233;seau humain pour les faire fonctionner. Le r&#233;seau d&#8217;agents M-Pesa, au seul Kenya, d&#233;passe aujourd&#8217;hui 300 000 points de service &#8212; plus de points de pr&#233;sence que toutes les banques du pays r&#233;unies. C&#8217;&#233;tait une cat&#233;gorie d&#8217;emplois qui n&#8217;existait pas avant le saut : des centaines de milliers de personnes gagnant leur vie &#224; l&#8217;interface humaine entre une nouvelle technologie et des populations qui ne pouvaient pas se servir seules. La technologie n&#8217;a pas retir&#233; l&#8217;humain de la boucle. Elle a cr&#233;&#233; la boucle et y a plac&#233; des humains.</p><p>Le d&#233;ploiement de l&#8217;IA a besoin de la m&#234;me couche, et il en a cruellement besoin &#8212; des gens pour adapter les mod&#232;les aux langues locales, les ajuster aux conditions r&#233;elles d&#8217;un secteur, les faire tourner l&#224; o&#249; l&#8217;infrastructure ne suivra pas, et livrer le service &#224; ceux qui ne toucheront jamais une API brute. Ce que sont concr&#232;tement ces emplois, et comment une strat&#233;gie les construit d&#233;lib&#233;r&#233;ment, fera l&#8217;objet de la deuxi&#232;me partie. Ce qu&#8217;il faut retenir pour l&#8217;instant, c&#8217;est que ce sont des emplois nouveaux que le saut engendre &#8212; et non d&#8217;anciens emplois qu&#8217;il d&#233;fend.</p><h2>Trois points dont cet argument doit tenir compte</h2><p>Un argument s&#233;rieux nomme ses propres faiblesses. Celui-ci en a trois.</p><ol><li><p><strong>Certains emplois africains sont bel et bien menac&#233;s.</strong> Le Kenya, l&#8217;Afrique du Sud, le Nigeria et l&#8217;&#201;gypte comptent de v&#233;ritables emplois d&#8217;externalisation (BPO), de centres d&#8217;appels et de back-office &#8212; exactement le type de travail que l&#8217;IA peut automatiser. Pire : la voie de l&#8217;externalisation de services, que l&#8217;Inde a gravie jusqu&#8217;&#224; la prosp&#233;rit&#233;, risque de dispara&#238;tre avant qu&#8217;une grande partie de l&#8217;Afrique ne l&#8217;atteigne. Si l&#8217;IA automatise l&#8217;externalisation d&#8217;entr&#233;e de gamme, cette premi&#232;re marche s&#8217;efface avant que la plupart du continent ne puisse y poser le pied. C&#8217;est r&#233;el, et cela renforce l&#8217;argument au lieu de l&#8217;affaiblir &#8212; c&#8217;est pr&#233;cis&#233;ment pourquoi la strat&#233;gie ne peut pas consister &#224; courir apr&#232;s le mod&#232;le d&#8217;externalisation que tout le monde est en train d&#8217;automatiser, mais &#224; construire la couche sup&#233;rieure: celle du d&#233;ploiement.</p></li><li><p><strong>&#171; l&#8217;IA va cr&#233;er des emplois &#187; n&#8217;est qu&#8217;un slogan tant que personne ne construit la couche qui les cr&#233;e.</strong> Rien de tout cela n&#8217;est automatique &#8212; et c&#8217;est exactement l&#224; que le laissez-faire &#233;choue en Afrique. Sans capacit&#233;s locales, le continent consomme l&#8217;IA &#233;trang&#232;re et ne capte rien du travail qui l&#8217;entoure. L&#8217;augmentation a lieu ; la valeur, elle, s&#8217;en va. L&#8217;opportunit&#233; est <em>conditionnelle</em>, et la condition, c&#8217;est le capital humain et la coordination. Ce n&#8217;est pas une faille dans l&#8217;argument. C&#8217;est la raison m&#234;me pour laquelle une strat&#233;gie d&#233;lib&#233;r&#233;e est indispensable.</p></li><li><p><strong>Le travail d&#8217;IA bas de gamme qui existe d&#233;j&#224; a &#233;t&#233; synonyme d&#8217;exploitation.</strong> Le travail de donn&#233;es que l&#8217;Afrique conna&#238;t d&#233;j&#224; est un avertissement, pas un mod&#232;le. Le magazine TIME a rapport&#233; que des travailleurs k&#233;nyans entra&#238;nant les filtres de s&#233;curit&#233; de ChatGPT gagnaient moins de deux dollars de l&#8217;heure, en triant des contenus violents et &#233;prouvants ; une grande plateforme s&#8217;est ensuite retir&#233;e du Kenya du jour au lendemain, laissant des milliers de personnes sans ressources. Voil&#224; &#224; quoi ressemblent les &#171; emplois de l&#8217;IA &#187; quand le travail est extractif et que la valeur se trouve ailleurs &#8212; et c&#8217;est la meilleure raison de viser plus haut, vers la propri&#233;t&#233;, plut&#244;t que de compter les volumes tout en bas.</p></li></ol><p>Nommer ces faiblesses n&#8217;affaiblit pas la th&#232;se de la cr&#233;ation. Cela d&#233;finit ce &#224; quoi doit ressembler une <em>bonne</em> cr&#233;ation : locale, d&#233;tenue localement, et qui monte en gamme.</p><h2>Qui cr&#233;e la valeur quand l&#8217;intelligence est bon march&#233; ?</h2><p>Ce qui nous ram&#232;ne &#224; la question la plus &#233;trange de l&#8217;article du Times &#8212; celle autour de laquelle des chercheurs chinois b&#226;tissent tout un champ d&#8217;&#233;tude : <em>qui cr&#233;e la valeur apr&#232;s l&#8217;IA ?</em></p><p>Quand l&#8217;intelligence elle-m&#234;me est abondante et presque gratuite, la valeur revient &#224; celui qui l&#8217;applique et l&#8217;adapte &#224; un probl&#232;me r&#233;el, dans un lieu r&#233;el. En Afrique, dans l&#8217;immense majorit&#233; des cas, celui qui l&#8217;applique est une personne &#8212; l&#8217;agent, l&#8217;op&#233;rateur, le b&#226;tisseur, le travailleur augment&#233;. La r&#233;ponse &#224; &#171; qui cr&#233;e la valeur apr&#232;s l&#8217;IA &#187; n&#8217;est donc ni une machine, ni un laboratoire. C&#8217;est un emploi. Toute la t&#226;che consiste &#224; faire en sorte qu&#8217;il y en ait assez, qu&#8217;ils montent en gamme, et qu&#8217;ils soient d&#233;tenus ici.</p><p>Voil&#224; pour la th&#232;se. Le plan d&#8217;action &#8212; les emplois pr&#233;cis que l&#8217;&#232;re de l&#8217;IA rend possibles, et la strat&#233;gie qui les construit d&#233;lib&#233;r&#233;ment &#8212; fera l&#8217;objet de la deuxi&#232;me partie.</p><p><strong>Leap. Build. Own.</strong></p><div><hr></div><p><em>Cet article part de la chronique de Katrin Bennhold dans le New York Times sur l&#8217;approche chinoise de l&#8217;IA et de l&#8217;emploi. Les chiffres sur l&#8217;emploi informel et le d&#233;ficit d&#8217;emplois proviennent de l&#8217;OIT et du rapport 2026 de la Fondation Mastercard sur l&#8217;emploi des jeunes en Afrique ; la projection de population active, de la Banque mondiale ; et les chiffres sur les agents de mobile money, des propres publications de Safaricom. Les &#233;l&#233;ments sur le travail de donn&#233;es sont tir&#233;s des reportages de TIME de 2023 et de la couverture ult&#233;rieure du secteur.</em></p><p></p>]]></content:encoded></item><item><title><![CDATA[The Jobs the AI Age Makes Possible - Part 2/2]]></title><description><![CDATA[The new kinds of work abundant intelligence opens up &#8212; and a smaller, local, connected way to build an economy]]></description><link>https://www.leapnotes.tech/p/the-jobs-the-ai-age-makes-possible</link><guid isPermaLink="false">https://www.leapnotes.tech/p/the-jobs-the-ai-age-makes-possible</guid><dc:creator><![CDATA[Kader Diagne]]></dc:creator><pubDate>Sun, 05 Jul 2026 18:45:20 GMT</pubDate><enclosure url="https://substackcdn.com/image/fetch/$s_!hy2c!,w_256,c_limit,f_auto,q_auto:good,fl_progressive:steep/https%3A%2F%2Fbucketeer-e05bbc84-baa3-437e-9518-adb32be77984.s3.amazonaws.com%2Fpublic%2Fimages%2F16ce3919-10c1-4ed0-b46f-720759955c7f_147x140.png" length="0" type="image/jpeg"/><content:encoded><![CDATA[<p><strong><span data-color="#ff0000" style="color: rgb(255, 0, 0);">Lire en fran&#231;ais &#8594; [</span><a href="https://open.substack.com/pub/kaderdiagne/p/les-emplois-que-lere-de-lia-rend?r=39ps0&amp;utm_campaign=post-expanded-share&amp;utm_medium=web"><span data-color="#ff0000" style="color: rgb(255, 0, 0);">lien</span></a><span data-color="#ff0000" style="color: rgb(255, 0, 0);">]</span></strong></p><p><em>Part one argued that Africa&#8217;s AI-and-jobs question is neither protection nor drift, but deliberate creation. This is the playbook: the new categories of work abundant intelligence opens up, and the strategy that builds them on purpose. <strong>Part two of two.</strong></em></p><div><hr></div><p><a href="https://open.substack.com/pub/kaderdiagne/p/the-ai-jobs-debate-has-two-answers?r=39ps0&amp;utm_campaign=post&amp;utm_medium=web&amp;showWelcomeOnShare=true">Part one</a> made a case: for most of Africa, AI isn&#8217;t a threat to a workforce that barely exists &#8212; it&#8217;s the raw material for building one. And the approach that fits is neither the American one of leaving it to the market nor the Chinese one of defending existing jobs, but the deliberate creation of new work.</p><p>A case, though, is not a plan. &#8220;AI will create jobs&#8221; is a slogan until you can name the jobs and name the moves that produce them. So here is the substance: what the work actually looks like, and what a strategy to build it would contain.</p><p>Start from the key point of part one. The scarcity in much of Africa isn&#8217;t work to be done &#8212; it&#8217;s people able to do it productively. Which means the goal isn&#8217;t to protect a headcount or to chase whatever jobs the outsourcing market happens to provide. It&#8217;s to maximize the number of people that abundant intelligence makes productive. That single change of focus changes what you build.</p><h2>New work the leap creates</h2><p>The mistake is to imagine AI only subtracts jobs. History is clear that general-purpose technologies also create categories of work nobody could have named beforehand &#8212; and Africa can build those categories <em>deliberately</em> rather than waiting to see which ones appear. Several are already visible:</p><ul><li><p><strong>AI integrators for small business</strong> &#8212; people who deploy, adapt, and keep running the AI tools that let SMEs become productive. Most small firms will never hire a global consultancy; someone local will do the technical work, the advisory work, and the maintenance &#8212; often the same person. There is an enormous amount of it.</p></li><li><p><strong>Local-language specialists</strong> &#8212; the people who adapt, evaluate, and maintain systems in Bambara, Wolof, Yoruba, Swahili, Lingala, Amharic, and the hundreds of languages the global models neglect. Done right, this is skilled, ongoing work &#8212; the opposite of the one-off, low-paid annotation part one warned about. It&#8217;s the design-and-quality layer that sits <em>above</em> raw data labeling, not the labeling itself.</p></li><li><p><strong>Agricultural AI advisors and tool operators</strong> &#8212; the people who bring smart irrigation, crop sensors, and field robots to farms, and help farmers act on what those tools show. With this support, one farmer can work land that used to need several.</p></li><li><p><strong>AI-supported health workers</strong> &#8212; a new profile of health worker for a new model of care: data-capturing health stations close to where people live, linked by telemedicine and AI to specialists who may be far away. It brings good diagnosis and monitoring to rural areas without building large hospitals everywhere &#8212; and it needs people trained to capture data, run the devices, and support patients, plus the technicians who keep it all working.</p></li><li><p><strong>Data-cooperative stewards</strong> &#8212; a newer, less proven role: helping cooperatives, municipalities, and associations gather, govern, and benefit from their own data instead of surrendering it. Emerging rather than established &#8212; but the need is already visible.</p></li></ul><p>And running underneath all of them, the largest category of all: <strong>AI-enabled micro-entrepreneurs</strong> &#8212; people using abundant intelligence to offer translation, design, bookkeeping, marketing, tourism services, procurement help, and a hundred other things, one person now able to do the work that once took a small team. This is not a forecast of a precise number. It&#8217;s the shape of an opportunity &#8212; and its scale matches the scale of the need.</p><p>Notice what most of these have in common. They sit at the point where intelligence meets a local reality it has to be adapted to &#8212; and that point is hard to move offshore. Not every task stays local: raw data work already flows to whichever country is cheapest, as part one&#8217;s warning showed. But the <em>relationship</em> layer &#8212; knowing the sector, the language, the community, the conditions, and being trusted to deploy inside them &#8212; is much harder to run from far away. That is where the durable, better-paid work sits.</p><p>One caveat: many of these begin as gig work or informal livelihoods, not salaried posts &#8212; and part one&#8217;s warning about the extractive bottom of the market applies here too. The task of a strategy is to help this work move toward stability and ownership, not to celebrate its raw numbers. Which is exactly what a strategy, as opposed to a slogan, is for.</p><h2>Industrialize through intelligence &#8212; factories included</h2><p>Step back, and this adds up to something bigger than a jobs list &#8212; a different way to develop. Africa still needs industry: it has minerals and crops that have to be processed and transformed, and doing that at home, instead of exporting raw materials and importing finished goods, is how value stays on the continent. So the leap is not about skipping factories. It&#8217;s about not repeating the slow path to them.</p><p>That old path ran in a fixed order: first build heavy industry and giant factories as the main engine of jobs; decades later, grow services; later still, a broad wave of entrepreneurship &#8212; each stage waiting for the one before. Abundant intelligence removes the need to wait. Africa can build smart factories that process local resources with fewer but more skilled people, and at the same time grow a large layer of AI-empowered services and small businesses &#8212; instead of treating giant factories as the only place work comes from.</p><p>Underneath this is a shift from giant and central to small and local. For a long time, scale meant size &#8212; one huge factory, one central hospital, one big school &#8212; with everything moved to and from it at high cost and risk. AI, automation, and robotics change that. They make small, local units efficient enough to be worth building, so production can sit where the resources are and services where people actually need them, instead of hauling everything to and from one center. Think of bakeries: not one giant plant baking for a whole country and shipping bread everywhere, but a bakery in each area, making what that area needs. The same now becomes possible for factories, clinics, and schools &#8212; a network of small, local units instead of a few giants.</p><p>And small doesn&#8217;t mean isolated. The units are connected, and they share what they learn: robots in one factory pass on their experience to robots in another; farms in different climates exchange what works; clinics pool what they see. A single central site only ever knows one reality. A network of local ones learns from many at once &#8212; different heat, cold, dust, soil, and daily use &#8212; and improves faster because of it. Decentralized but coordinated beats big and central. That coordination &#8212; many local units learning as one &#8212; is the real strength this whole approach is built on.</p><p>You can see the move across sectors. In farming, smart irrigation, sensors, and robots on many small holdings rather than one giant farm. In health, local data stations, telemedicine, and wearable devices that follow people day to day &#8212; managing chronic conditions, guiding nutrition, catching problems early &#8212; instead of a few costly central hospitals. Each time the logic is the same: apply intelligence now, and put the capability where the people and the resources are, rather than first building expensive 20th-century infrastructure and centralizing everything around it. Done well, the result is better and cheaper at once &#8212; and, in health, it keeps the workforce healthier and more productive while avoiding the runaway costs that strain richer economies.</p><p>That is what leapfrogging really means here: not skipping industry, but building it in a new shape &#8212; smaller, local, connected, and all at once. Africa can reach a modern economy &#8212; factories, services, and entrepreneurship together &#8212; without spending decades climbing there one stage at a time. It works with the continent&#8217;s existing strength in informal entrepreneurship rather than against it; not everyone becomes an employee, and many become entrepreneurs. The opportunity is to turn abundant intelligence into a faster, smarter path than anyone has taken before.</p><h2>The strategy that builds it</h2><p>None of this is automatic &#8212; that was the warning of part one. Deliberate creation requires deliberate moves. A serious AI-employment strategy would reach for tools like these:</p><ul><li><p><strong>National AI apprenticeship and vocational programs</strong> &#8212; treating applied AI skills the way earlier eras treated trades.</p></li><li><p><strong>AI extension services</strong> &#8212; modeled on the agricultural extension networks that already exist, putting advisory capability within reach of farmers and small businesses.</p></li><li><p><strong>An AI deployment corps</strong> &#8212; organized capacity to bring AI into public services and SMEs, creating the integrator jobs above in the process.</p></li><li><p><strong>Public procurement that favors local AI startups</strong> &#8212; using government demand to grow domestic capability rather than import it.</p></li><li><p><strong>AI entrepreneurship funds</strong> &#8212; capital aimed at the micro- and small-enterprise layer where most of the job creation will actually happen.</p></li><li><p><strong>Local-language foundation and application models</strong> &#8212; treated as public infrastructure, because everything else depends on them.</p></li><li><p><strong>Mass AI literacy</strong> &#8212; the base layer that lets the whole population participate rather than a thin elite.</p></li></ul><p>These are not unusual. They&#8217;re the ordinary instruments of industrial policy, pointed at a target others aren&#8217;t aiming for &#8212; not protecting yesterday&#8217;s jobs, not waiting on the market, but building the conditions for tomorrow&#8217;s work to exist.</p><h2>Where the strategy is anchored</h2><p>Every item on that list rests on the same foundation, which is why it&#8217;s the foundation of everything <strong>leapfrogging.africa</strong>does: people who can actually deploy AI under African conditions. Without them, the integrator jobs, the deployment corps, the local-language models, and the micro-enterprises all get rented from somewhere else &#8212; and the strategy collapses back into consumption.</p><p>That&#8217;s the role of <strong>ALIT Africa</strong> at the applied end, and of education as a full pillar of leapfrogging.africa beneath it &#8212; running from primary and secondary through vocational and professional training, and built the way abundant intelligence now allows: adaptive, mastery-based, with readiness rather than the calendar deciding when someone is qualified. Education isn&#8217;t adjacent to the jobs strategy. It <em>is</em> the jobs strategy, seen from its source. This is the employment pillar of the whole leapfrogging.africa argument: if intelligence is becoming a commodity, then Africa&#8217;s competitive advantage is the number of people it can make able to apply it &#8212; workers, operators, builders, and entrepreneurs, at scale.</p><h2>The thing no one has ever done</h2><p>One last point, and it&#8217;s the biggest, because it&#8217;s easy to miss in the day-to-day.</p><p>No country in history has industrialized with abundant intelligence available from the start. Britain didn&#8217;t. Germany didn&#8217;t. America didn&#8217;t. China didn&#8217;t. Every one of them built its modern economy first, and only later had to add intelligence into it &#8212; which is exactly what much of the world is now struggling to do.</p><p>Africa could be the first to build large parts of a modern economy with abundant intelligence available on day one. That&#8217;s not a consolation prize for arriving late. It&#8217;s a starting position no earlier industrializer ever had.</p><p>Others are working out how to protect yesterday&#8217;s workforce, or hoping the market resolves the question for them. Africa has the rarer task, and the rarer chance: to design the future before the old system is in place &#8212; and to make sure the jobs it builds are owned here.</p><p><strong>Leap. Build. Own.</strong></p><p><strong><a href="https://open.substack.com/pub/kaderdiagne/p/the-ai-jobs-debate-has-two-answers?r=39ps0&amp;utm_campaign=post&amp;utm_medium=web&amp;showWelcomeOnShare=true">Link to Part one</a></strong></p>]]></content:encoded></item><item><title><![CDATA[The AI Jobs Debate Has Two Answers. Africa Needs a Third - Part 1/2)]]></title><description><![CDATA[For Africa, AI isn't a threat to jobs. It's the way to create them.]]></description><link>https://www.leapnotes.tech/p/the-ai-jobs-debate-has-two-answers</link><guid isPermaLink="false">https://www.leapnotes.tech/p/the-ai-jobs-debate-has-two-answers</guid><dc:creator><![CDATA[Kader Diagne]]></dc:creator><pubDate>Sun, 05 Jul 2026 14:23:24 GMT</pubDate><enclosure url="https://substackcdn.com/image/fetch/$s_!hy2c!,w_256,c_limit,f_auto,q_auto:good,fl_progressive:steep/https%3A%2F%2Fbucketeer-e05bbc84-baa3-437e-9518-adb32be77984.s3.amazonaws.com%2Fpublic%2Fimages%2F16ce3919-10c1-4ed0-b46f-720759955c7f_147x140.png" length="0" type="image/jpeg"/><content:encoded><![CDATA[<p><strong><span data-color="#ff0000" style="color: rgb(255, 0, 0);">Lire en fran&#231;ais &#8594; [</span><a href="https://open.substack.com/pub/kaderdiagne/p/the-ai-jobs-debate-has-two-answers?r=39ps0&amp;utm_campaign=post&amp;utm_medium=web&amp;showWelcomeOnShare=true"><span data-color="#ff0000" style="color: rgb(255, 0, 0);">lien</span></a><span data-color="#ff0000" style="color: rgb(255, 0, 0);">]</span></strong></p><p><em>The world&#8217;s two AI superpowers are handling AI&#8217;s effect on jobs in very different ways &#8212; one largely leaves it to the market, the other manages it from the top. Neither fits Africa, because both assume a stock of jobs that here was mostly never built. That isn&#8217;t a reason to drift, and it isn&#8217;t a reason to defend. It&#8217;s a reason to have a strategy. Part one of two on turning AI from a threat to jobs into the thing that finally creates them &#8212; prompted by Katrin Bennhold's recent New York Times piece on China and AI jobs.</em></p><div><hr></div><p>This started with a New York Times article by Katrin Bennhold. In it, she describes two very different ways AI's effect on jobs is unfolding in the world's two AI superpowers.</p><div class="subscription-widget-wrap-editor" data-attrs="{&quot;url&quot;:&quot;https://www.leapnotes.tech/subscribe?&quot;,&quot;text&quot;:&quot;Subscribe&quot;,&quot;language&quot;:&quot;en&quot;}" data-component-name="SubscribeWidgetToDOM"><div class="subscription-widget show-subscribe"><div class="preamble"><p class="cta-caption"></p></div><form class="subscription-widget-subscribe"><input type="email" class="email-input" name="email" placeholder="Type your email&#8230;" tabindex="-1"><input type="submit" class="button primary" value="Subscribe"><div class="fake-input-wrapper"><div class="fake-input"></div><div class="fake-button"></div></div></form></div></div><p>In the United States, the response is essentially hands-off. The frontier labs are chasing superintelligence (machines meant to surpass human intelligence, not just match it) and the state has largely stood back, on the assumption that disruption sorts itself out. There isn&#8217;t really a plan; leaving it to the market <em>is</em> the approach. Call it laissez-faire.</p><p>In China, it looks very different. Beijing has written the employment problem into its five-year plan and started acting on it &#8212; leaning on companies against layoffs, backing dismissed workers in court, pushing retraining, even spawning an academic field to work out who creates value once the machines arrive. Call it protection: the state actively managing the effect on the existing workforce.</p><p>Two very different approaches but the same underlying premise. Both take as given a large stock of formal jobs (factory lines, white-collar desks, knowledge workers) and turn on what AI does to it. One lets that stock be disrupted; the other defends it. Either way, that stock is what the debate is about.</p><p>For most of Africa, that stock was mostly never built. And that changes the question.</p><h2> Africa starts from a different place</h2><p>The dominant labor problem across most of the continent isn&#8217;t <em>too few jobs left after AI</em>. It&#8217;s <em>too few jobs, full stop</em> &#8212; and a young population arriving faster than any formal economy is absorbing it.</p><p>Around 85% of employment in sub-Saharan Africa is informal &#8212; smallholder farming, street trading, small-scale services &#8212; against a world average near 58%. Every year, more than 10 million young people enter the African labor market, and current growth patterns produce roughly 3 million formal jobs to meet them. By one estimate, sub-Saharan Africa alone will need to create on the order of 15 million new jobs a year by 2030 just to keep pace with new entrants &#8212; and between now and 2050 the region will add more than 620 million people to its working-age population, the fastest such expansion anywhere in the world.</p><p>So a debate built around what AI does to a stock of formal jobs &#8212; whether to let it be disrupted or to defend it &#8212; is really a debate for the large industrialized economies. It answers a question that much of Africa, along with other young, largely informal economies, doesn&#8217;t have.</p><h2>Why neither answer fits</h2><p>Carry each approach to Africa, and the mismatch shows.</p><p>The laissez-faire approach assumes a functioning market of formal jobs for AI to reshuffle. Where that market is thin, &#8220;let it happen&#8221; doesn&#8217;t produce a messy-but-dynamic reallocation. It produces drift: foreign AI gets consumed, the value it generates is captured elsewhere, and the local economy inherits the extractive bottom of the supply chain without the productive top. Passivity isn&#8217;t neutral here. It&#8217;s a decision to miss the opening.</p><p>The protection approach fails for the opposite reason &#8212; there&#8217;s little to protect. Defending a stock of formal jobs that was never created is the wrong fight, in the same way, and for the same reason, that &#8220;who owns the smartest model&#8221; was the wrong race. You cannot cushion the loss of jobs that don&#8217;t exist.</p><p>Which leaves a third posture, and it&#8217;s the one that fits: not drift, not defense, but <strong>deliberate creation</strong>. Africa&#8217;s AI-and-jobs question isn&#8217;t &#8220;how many jobs survive?&#8221; It&#8217;s &#8220;how many can we build &#8212; and will we build them on purpose?&#8221;</p><p>That phrase &#8212; <em>on purpose</em> &#8212; is the whole thing. The most useful line in that Times article isn&#8217;t about China&#8217;s methods, most of which don&#8217;t transplant anyway. It&#8217;s the conclusion: policymakers have agency. The direction of this technology is a choice, not a weather system. And agency cuts against both other approaches at once &#8212; against the drift of laissez-faire and against the backward-looking reflex of protection. It points at strategy. What that strategy actually contains is part two. Part one is the case for having one at all.</p><p>Let me be clear about the limits of this claim, because they matter. This is not a claim that <em>no</em> African job is exposed &#8212; some are, and I&#8217;ll come to them squarely. It&#8217;s a claim about where the weight sits. For a continent whose defining challenge is a jobs <em>gap</em>, not a jobs <em>cliff</em>, the posture that fits is neither passive nor defensive. It&#8217;s proactive.</p><h2>Intelligence becomes infrastructure</h2><p>Start with why creation is even possible &#8212; why abundant AI is leverage here rather than a threat.</p><p>China&#8217;s stated instinct is to use AI to <em>augment</em> its workers rather than replace them, making people already in formal jobs more productive. Take that instinct and point it at a different target: not the employed, but the vast <strong>underemployed</strong> &#8212; the smallholder farmer, the market trader, the micro-entrepreneur, the community health worker, the small clinic. There are far more of them here than there are office workers, and that is the point.</p><p>Just as roads, electricity, and the internet each became infrastructure a whole economy could build on, abundant intelligence becomes <em>cognitive</em> infrastructure &#8212; and it lands on the informal economy as raw capability. A phone-based crop-disease diagnosis turns a farmer&#8217;s guesswork into a decision, letting one extension officer reach far more farmers than before. A local-language advisory turns a trader with no market data into one who can price and plan. An AI-supported health worker can safely handle cases that used to require a referral hundreds of kilometers away. None of this is a salaried job in a tower. It&#8217;s a productivity multiplier laid over the economy that already employs most of the continent &#8212; lifting the floor for the people standing on it.</p><p>This exposes what the shortage really is. Governments habitually ask <em>how do we create jobs?</em> The sharper question is <em>how do we increase productive capacity?</em> &#8212; because the binding scarcity in much of Africa isn&#8217;t work to be done, it&#8217;s people able to do it productively. When intelligence becomes infrastructure, millions who were locked out of productive work by lack of training, tools, or institutions can suddenly enter it. The scarce resource stops being the intelligence and becomes the human able to apply it.</p><h2>The real leap: skipping obsolete models, not technology</h2><p>Here&#8217;s where the word <em>leapfrogging</em> has to be used precisely, because the common version is too small.</p><p>Most people hear &#8220;leapfrogging&#8221; and think: skip landlines, skip bank branches, skip desktops &#8212; jump straight to the newer technology. True as far as it goes. But the deeper leap isn&#8217;t skipping a <em>technology</em>. It&#8217;s skipping an <em>obsolete organizational model</em>.</p><p>The developed world built its economies by first creating millions of clerical, administrative, and back-office jobs &#8212; and is now spending enormous political energy working out how to automate them without social rupture. Africa does not have to walk into that trap and back out of it. It can skip the entire stage of manufacturing low-value clerical work only to dismantle it a generation later, and build higher-value, AI-enabled work from the start. That is a far bigger leap than skipping a wire in the ground. It&#8217;s skipping a whole economic detour.</p><h2>The human layer is the job engine</h2><p>And deploying AI under African conditions doesn&#8217;t just make existing workers more productive. It <em>creates entirely new categories of work</em> &#8212; and we have the template.</p><p>Every prior African leap did exactly this. Mobile money is the clearest case. It didn&#8217;t only digitize payments; it built a vast human network to run them. M-Pesa&#8217;s agent footprint in Kenya alone now exceeds 300,000 outlets &#8212; more points of presence than every bank in the country combined. That was a job category that did not exist before the leap: hundreds of thousands of people earning a living at the human interface between a new technology and communities that couldn&#8217;t self-serve. The technology didn&#8217;t remove the human from the loop. It created the loop and put humans in it.</p><p>AI deployment needs the same layer, and needs it badly &#8212; people to localize models, adapt them to a sector&#8217;s real conditions, operate them where the infrastructure won&#8217;t help, and deliver the service to those who will never touch a raw API. What those jobs actually are, and how a strategy deliberately builds them, is the subject of part two. The point for now is that they are new jobs the leap generates &#8212; not old jobs it defends.</p><h2>Three things this argument has to account for</h2><p>A serious case names its own weak points. This one has three.</p><p><strong>First: Some African jobs really are exposed.</strong> Kenya, South Africa, Nigeria, and Egypt have real business-process-outsourcing, call-center, and back-office employment &#8212; exactly the kind of work AI can automate away. Worse, the services-outsourcing path that India climbed to prosperity may be pulled away before much of Africa reaches it: if AI automates entry-level outsourcing, that first step is gone before most of the continent can climb onto it. This is real, and it sharpens the argument rather than weakening it &#8212; it&#8217;s exactly why the strategy can&#8217;t be to chase the outsourcing model everyone else is automating, but to build the deployment layer above it.</p><p><strong>Second: &#8220;AI will create jobs&#8221; is just a slogan unless someone builds the layer that creates them.</strong> None of this is automatic &#8212; and this is exactly where laissez-faire fails Africa. Without local capability, the continent consumes foreign AI and captures none of the work around it. The augmentation happens; the value leaves. The opportunity is <em>conditional</em>, and the condition is human capital and coordination. That&#8217;s not a hole in the argument. It&#8217;s the reason a deliberate strategy is the whole point.</p><p><strong>Third: The low-end AI work that exists today has been exploitative.</strong> The AI data work Africa already has is a warning, not a model. TIME reported that Kenyan workers training ChatGPT&#8217;s safety filters earned under two dollars an hour, sorting through graphic and disturbing content; one major platform later pulled out of Kenya overnight, leaving thousands stranded. That is what &#8220;AI jobs&#8221; look like when the work is extractive and the value sits elsewhere &#8212; and it&#8217;s the strongest reason to aim higher, toward ownership, rather than to count volume at the bottom.</p><p>Naming these doesn&#8217;t weaken the case for creation. It defines what <em>good</em> creation has to look like: local, owned, and climbing.</p><h2>Who creates value when intelligence is cheap?</h2><p>Which brings us back to the strangest question in that Times article &#8212; the one Chinese scholars are building a whole field to answer: <em>who creates value after AI?</em></p><p>When intelligence itself is abundant and nearly free, value belongs to whoever applies and adapts it to a real problem in a real place. In Africa, overwhelmingly, that applier is a person &#8212; the agent, the operator, the builder, the augmented worker. So the answer to &#8220;who creates value after AI&#8221; isn&#8217;t a machine, and isn&#8217;t a lab. It&#8217;s a job. The task is to make sure there are enough of them, that they climb, and that they&#8217;re owned here.</p><p>That&#8217;s the case. The playbook &#8212; the specific jobs the AI age makes possible, and the strategy that builds them on purpose &#8212; is <a href="https://open.substack.com/pub/kaderdiagne/p/the-jobs-the-ai-age-makes-possible?r=39ps0&amp;utm_campaign=post&amp;utm_medium=web&amp;showWelcomeOnShare=true">part two</a>.</p><p><strong>Leap. Build. Own.</strong></p><p><strong><a href="https://www.leapnotes.tech/p/the-jobs-the-ai-age-makes-possible?r=39ps0">Link to Part two</a></strong></p><div><hr></div><p><em>This article takes off from Katrin Bennhold&#8217;s New York Times column on China&#8217;s approach to AI and employment. Figures on informal employment and the jobs gap are drawn from the ILO and the Mastercard Foundation&#8217;s 2026 Africa Youth Employment Outlook; the working-age projection from the World Bank; and the mobile-money agent figures from Safaricom&#8217;s own reporting. The data-work accounts are from TIME&#8217;s 2023 reporting and subsequent coverage of the sector.</em></p><div class="subscription-widget-wrap-editor" data-attrs="{&quot;url&quot;:&quot;https://www.leapnotes.tech/subscribe?&quot;,&quot;text&quot;:&quot;Subscribe&quot;,&quot;language&quot;:&quot;en&quot;}" data-component-name="SubscribeWidgetToDOM"><div class="subscription-widget show-subscribe"><div class="preamble"><p class="cta-caption">Thanks for reading LeapNotes by Kader Diagne! Subscribe for free to receive new posts and support my work.</p></div><form class="subscription-widget-subscribe"><input type="email" class="email-input" name="email" placeholder="Type your email&#8230;" tabindex="-1"><input type="submit" class="button primary" value="Subscribe"><div class="fake-input-wrapper"><div class="fake-input"></div><div class="fake-button"></div></div></form></div></div><p><a href="https://www.leapnotes.tech/p/the-jobs-the-ai-age-makes-possible?r=39ps0">Link to Part two</a></p>]]></content:encoded></item><item><title><![CDATA[Coordination Without a Center]]></title><description><![CDATA[A LeapNote on Stanford's DeLM &#8212; and what it suggests about how to build.]]></description><link>https://www.leapnotes.tech/p/coordination-without-a-center</link><guid isPermaLink="false">https://www.leapnotes.tech/p/coordination-without-a-center</guid><dc:creator><![CDATA[Kader Diagne]]></dc:creator><pubDate>Thu, 18 Jun 2026 15:55:28 GMT</pubDate><enclosure url="https://substackcdn.com/image/fetch/$s_!hy2c!,w_256,c_limit,f_auto,q_auto:good,fl_progressive:steep/https%3A%2F%2Fbucketeer-e05bbc84-baa3-437e-9518-adb32be77984.s3.amazonaws.com%2Fpublic%2Fimages%2F16ce3919-10c1-4ed0-b46f-720759955c7f_147x140.png" length="0" type="image/jpeg"/><content:encoded><![CDATA[<p>Most multi-agent AI systems run on a &#8220;boss&#8221;: a central orchestrator that hands out subtasks, collects every result, decides what to merge, and rebroadcasts it. It works &#8212; until it doesn&#8217;t. As the number of subtasks grows, that controller becomes the bottleneck, and the cost piles up.</p><p>Stanford&#8217;s <a href="https://arxiv.org/pdf/2606.10662">DeLM</a> &#8212; <em>Decentralized Multi-Agent Systems with Shared Context</em> &#8212; tries the opposite. There is no central orchestrator. Parallel agents claim subtasks from a shared queue, read the accumulated <em>verified</em> progress, reason locally, and write back compact verified updates to a shared context that everyone builds on. Coordination happens through the shared substrate, not through a controller.</p><div class="subscription-widget-wrap-editor" data-attrs="{&quot;url&quot;:&quot;https://www.leapnotes.tech/subscribe?&quot;,&quot;text&quot;:&quot;Subscribe&quot;,&quot;language&quot;:&quot;en&quot;}" data-component-name="SubscribeWidgetToDOM"><div class="subscription-widget show-subscribe"><div class="preamble"><p class="cta-caption"></p></div><form class="subscription-widget-subscribe"><input type="email" class="email-input" name="email" placeholder="Type your email&#8230;" tabindex="-1"><input type="submit" class="button primary" value="Subscribe"><div class="fake-input-wrapper"><div class="fake-input"></div><div class="fake-button"></div></div></form></div></div><p>And on the benchmarks, it holds up. According to the paper, DeLM beats the strongest baseline by up to 10.5 points on SWE-bench Verified &#8212; at roughly half the cost per task &#8212; and improves on LongBench-v2 Multi-Doc QA across four frontier model families.</p><p>I&#8217;m not reading this as &#8220;central control is bad.&#8221; The paper doesn&#8217;t claim that, and neither do I. But two things stay with me, and both sit at the center of what I write about here.</p><p>First: coordination doesn&#8217;t always need someone in charge. The reflex &#8212; in AI and in institutions alike &#8212; is that to align many actors you need a controller. DeLM is a clean reminder that a good shared, verified context plus simple rules can carry a lot of that load, and sometimes carry it better.</p><p>Second: the win here is efficiency, not a bigger model. Same models, coordinated better, at half the cost. That&#8217;s the applied layer &#8212; where durable value is migrating, and where builders who have <em>had</em> to be resource-efficient hold an edge.</p><p>And it rhymes with how I think the next ecosystems should be built: not one hub commanding everything, but a network of nodes building on a shared foundation, coordinating without a bottleneck. Decentralized coordination, done well, isn&#8217;t a compromise. On this evidence, it can be the better design.</p><p><em>Paper: <a href="https://arxiv.org/pdf/2606.10662">Decentralized Multi-Agent Systems with Shared Context</a> (Stanford, via arXiv).</em></p><div class="subscription-widget-wrap-editor" data-attrs="{&quot;url&quot;:&quot;https://www.leapnotes.tech/subscribe?&quot;,&quot;text&quot;:&quot;Subscribe&quot;,&quot;language&quot;:&quot;en&quot;}" data-component-name="SubscribeWidgetToDOM"><div class="subscription-widget show-subscribe"><div class="preamble"><p class="cta-caption"></p></div><form class="subscription-widget-subscribe"><input type="email" class="email-input" name="email" placeholder="Type your email&#8230;" tabindex="-1"><input type="submit" class="button primary" value="Subscribe"><div class="fake-input-wrapper"><div class="fake-input"></div><div class="fake-button"></div></div></form></div></div>]]></content:encoded></item><item><title><![CDATA[It Was Never A Race. This Is Africa’s Moment to Leap.]]></title><description><![CDATA[Intelligence is becoming abundant &#8212; and for Africa that's raw material, not a finish line we're behind on. Why, and what we're building.]]></description><link>https://www.leapnotes.tech/p/it-was-never-a-race-this-is-africas</link><guid isPermaLink="false">https://www.leapnotes.tech/p/it-was-never-a-race-this-is-africas</guid><dc:creator><![CDATA[Kader Diagne]]></dc:creator><pubDate>Sun, 14 Jun 2026 05:21:14 GMT</pubDate><enclosure url="https://substackcdn.com/image/fetch/$s_!hy2c!,w_256,c_limit,f_auto,q_auto:good,fl_progressive:steep/https%3A%2F%2Fbucketeer-e05bbc84-baa3-437e-9518-adb32be77984.s3.amazonaws.com%2Fpublic%2Fimages%2F16ce3919-10c1-4ed0-b46f-720759955c7f_147x140.png" length="0" type="image/jpeg"/><content:encoded><![CDATA[<p><em>The world is busy arguing over who&#8217;s winning the AI race. I&#8217;m an African who studied and built AI in Germany &#8212; and from where I stand, it was never a race. It&#8217;s an opening. This is the first in a series on why, and on what we&#8217;re building.</em></p><h4>Intelligence is becoming abundant &#8212; and for Africa that&#8217;s raw material, not a finish line we&#8217;re behind on. Why, and what we&#8217;re building.</h4><p>Something the world hasn&#8217;t seen in a long time is underway: intelligence is becoming abundant. The knowledge, the models, the tools that used to be scarce, costly, and gated are pouring into the open and getting cheaper and more capable by the month.</p><p>Most of the noise around this treats it as a contest &#8212; who builds the smartest model, the US or China, with Europe quietly fearing it has already lost. I&#8217;ll leave that race to the people running it. (Sangeet Paul Choudary argues, persuasively, that they may be running the wrong one &#8212; that as intelligence becomes abundant, advantage shifts from whoever owns the smartest model to whoever turns it into real-world value. It&#8217;s worth reading.) But from where I stand, it was never a race at all.</p><p>For Africa, this abundance isn&#8217;t a threat to defend against or a finish line it&#8217;s behind on. It&#8217;s raw material. When capability becomes cheap and open, the ground is set for exactly the move Africa has made before &#8212; a leap &#8212; and this time across many fronts at once: education, health, agriculture, finance, infrastructure, governance, etc. That is Africa&#8217;s moment. And we&#8217;d already begun building for it: I initiated <strong>leapfrogging.africa</strong>, architected the model, and brought in the first partners well before any of this was cast as a race. We weren&#8217;t trying to beat anyone. We saw the opening.</p><p>A word on where I&#8217;m standing, since it shapes all of this. I&#8217;m African. I came to Germany to study computer science and AI, spent years as a researcher at DFKI &#8212; the world&#8217;s largest independent AI research center, and a pioneer in the field since 1988 &#8212; and spun a venture-backed AI company out of it, which is how I came to know both the technology and the market from the inside. leapfrogging.africa grew directly out of that experience. DFKI is application-driven by design: universities, industry, and government coordinating to turn research into real applications and companies, with no venture capital at the table. Capital came afterward, to the companies that work produced, mine among them. That order is the part worth carrying: you build the engine that makes companies worth funding, and capital &#8212; abundant and mobile &#8212; comes looking. That&#8217;s the engine leapfrogging.africa is built to adapt and scale across Africa &#8212; applied AI and entrepreneurship at continental scale.</p><h3>Application and coordination &#8212; not foundation models</h3><p>Here is the part that matters, and it reads as an opportunity, not a contest. When intelligence is abundant, the scarce thing &#8212; and the valuable thing &#8212; is no longer the model. It&#8217;s two things done well together. The first is <strong>application</strong>: putting that intelligence to work on real problems, in the sectors and use cases that actually matter, adapted to the realities, strengths, and needs of a specific place. The second is <strong>coordination</strong>: orchestrating the people, tools, and systems to deploy it reliably and at scale. A model on its own just lists things that could be done. Value and impact belong to whoever applies it to the right problem and gets it done.</p><p>That reframes the whole question of who is &#8220;ahead.&#8221; Building the biggest foundation model was never going to be Africa&#8217;s game, and chasing it would waste a decade. But that was never where the lasting value sits. It sits in adapted application backed by strong coordination &#8212; and that work has barely begun, for anyone. It&#8217;s also why everything we do, including our research, is driven by application rather than the other way around.</p><p class="button-wrapper" data-attrs="{&quot;url&quot;:&quot;https://www.leapnotes.tech/subscribe?&quot;,&quot;text&quot;:&quot;Subscribe now&quot;,&quot;action&quot;:null,&quot;class&quot;:null}" data-component-name="ButtonCreateButton"><a class="button primary" href="https://www.leapnotes.tech/subscribe?"><span>Subscribe now</span></a></p><h3>Why this is Africa&#8217;s terrain</h3><p>This is where the conventional story inverts. The conditions usually filed under &#8220;Africa is behind&#8221; are, on closer inspection, advantages for exactly this kind of work.</p><p>Africa is unconstrained by legacy infrastructure &#8212; there is no sunk cost in the old way of doing things to defend. It is deeply experienced in navigating complex, multi-stakeholder environments, which is precisely what coordination demands. And its resource constraints &#8212; limited compute, intermittent power, thin bandwidth, relentless cost pressure &#8212; don&#8217;t merely hold it back; they force the efficiency innovation that the rest of an energy-hungry, datacenter-bound world is about to need at scale.</p><p>And this is not wishful thinking &#8212; it&#8217;s the pattern of the last twenty years. Africa never strung copper landlines to every village; it went straight to mobile and put a phone in nearly every hand. It never waited for bank branches and debit cards to reach everyone; it jumped to mobile money, and didn&#8217;t merely adopt the idea but pioneered and revolutionized it for the rest of the world. Each time, skipping the legacy stage that only ever served a few didn&#8217;t leave Africa behind &#8212; it put Africa in front. AI is the next leap of exactly that kind. Don&#8217;t replay decades of someone else&#8217;s path; skip to where the value is heading &#8212; applied AI, edge computing, resource-efficient solutions that work in African conditions and therefore work anywhere.</p><p>The world already has enough people fine-tuning models in datacenters. What it lacks, badly, is people who can make intelligence work where the infrastructure won&#8217;t help them &#8212; on a solar-powered device, offline, all day, reliably. That is not a hypothetical specialty. It is daily life here. And it is a defensible global position, not a consolation prize.</p><h3>What leapfrogging.africa actually is</h3><p>So let me be clear about what this is, because it&#8217;s bigger than any one school or product. <strong>leapfrogging.africa</strong> is a movement to build and connect AI ecosystems across the continent &#8212; across countries and across industries. The unit of ambition isn&#8217;t a graduate or an app; it&#8217;s an ecosystem that turns intelligence into deployed, African-owned solutions in agriculture, health, finance, mining, infrastructure, and governance &#8212; and then a network of those ecosystems, coordinated across borders.</p><p>That&#8217;s a continental claim, so the method has to be the opposite of grandiose. You don&#8217;t launch a continent. You start small and prove it: a few pilot countries, a handful of committed partners, one working node made real and measured. Then you replicate. The aim isn&#8217;t to assign each country a role but to create an incentive: a country &#8212; or a region &#8212; can grow into an excellence hub in the sectors that matter to it, as deeply as it chooses to engage. Agritech where farming drives the economy, health AI where the need is sharpest, fintech built on the continent&#8217;s mobile-money lead, mining, logistics, or energy where those are local strengths &#8212; AI adapted to real conditions, not generic, with edge computing and resource efficiency running across all of it as a shared technical strength. Each node is anchored to local universities and industry, all stitched into a coordinated whole. Pilot, validate, scale. The network is the product.</p><p>This is where the coordination thesis stops being abstract. An ecosystem like this only holds together if capability is owned locally &#8212; you cannot coordinate what you cannot execute &#8212; and if universities, industry, and government sit in it as co-owners and partners rather than donors. Capital and entrepreneurs come next &#8212; not as the starting point, but as what a working ecosystem attracts and produces: the engine comes first, and capital is what a credible engine eventually draws, as an investor rather than a donor; entrepreneurs turn trained people and research into companies, products, and jobs. (How that capital actually gets drawn &#8212; and why the engine is the decisive thing &#8212; is a post of its own).</p><p> None of this has to be invented from zero: a multi-stakeholder model along these lines has been proven over decades &#8212; in Europe, the US, and beyond. The point isn&#8217;t to import it wholesale; it&#8217;s that we don&#8217;t have to spend forty years rediscovering what works. The same leapfrog instinct applies to institutions: we start where that model arrived and adapt it, hard, for Africa and the challenge at hand. It&#8217;s the scaffolding the ecosystem stands on &#8212; important, but not the destination.</p><h3>Where it starts: the people who build</h3><p>A continental ecosystem still rests on one scarce resource: people who can actually deploy AI under African conditions. Without them, every layer above gets rented from someone else. So the foundational first step &#8212; the seed the rest grows from &#8212; is human capital.</p><p>That&#8217;s the role of our flagship initiative, <strong>ALIT Africa</strong> (the Applied Learning Institute for Technology): working with established African universities to train people to put AI to work under real conditions &#8212; applied to local sectors, built for edge computing and resource efficiency, adapted to African languages and markets. It&#8217;s also the first place we apply the leap to ourselves, including to how people learn &#8212; mastery over seat-time, with readiness rather than the calendar deciding when someone is qualified. ALIT is where leapfrogging.africa starts. It is not what leapfrogging.africa is. (It gets its own piece next: what it is, how it works, and why education itself is one of the areas Africa can leap.)</p><p>Because the underlying choice is stark, and it&#8217;s being made right now. AI-driven transformation in Africa will either be owned and operated by African institutions or licensed from foreign providers; built on African expertise or dependent on outside consultants; designed for African contexts or adapted, awkwardly, from elsewhere. I know which version I want to help build.</p><p>This is the first of a regular series &#8212; the strategy, the ecosystem model, the partners, the obstacles, and the slow real work of building it, in public, as it happens.</p><p>The question was never whether Africa takes part in the AI era. It&#8217;s whether we turn abundance into capability at scale &#8212; our countries building together toward the same goal rather than each racing the next.</p><p>If that resonates, follow along. And if you&#8217;d rather build it than read about it, that&#8217;s the better invitation &#8212; reach me at <a href="mailto:alit@leapfrogging.africa">alit@leapfrogging.africa</a>.</p><p class="button-wrapper" data-attrs="{&quot;url&quot;:&quot;https://www.leapnotes.tech/p/it-was-never-a-race-this-is-africas?utm_source=substack&utm_medium=email&utm_content=share&action=share&quot;,&quot;text&quot;:&quot;Share&quot;,&quot;action&quot;:null,&quot;class&quot;:null}" data-component-name="ButtonCreateButton"><a class="button primary" href="https://www.leapnotes.tech/p/it-was-never-a-race-this-is-africas?utm_source=substack&utm_medium=email&utm_content=share&action=share"><span>Share</span></a></p><div><hr></div><p><em>Credit where it&#8217;s due: the US&#8211;China &#8220;wrong race&#8221; framing I gesture at up top is drawn from Sangeet Paul Choudary&#8217;s essay and his book</em> Reshuffle*, which argue that AI&#8217;s lasting effect is to make intelligence abundant and to shift durable advantage toward coordination and execution. I&#8217;ve pointed the same lens at a continent he wasn&#8217;t writing about &#8212; and that we&#8217;d already been building for.*</p><div class="subscription-widget-wrap-editor" data-attrs="{&quot;url&quot;:&quot;https://www.leapnotes.tech/subscribe?&quot;,&quot;text&quot;:&quot;Subscribe&quot;,&quot;language&quot;:&quot;en&quot;}" data-component-name="SubscribeWidgetToDOM"><div class="subscription-widget show-subscribe"><div class="preamble"><p class="cta-caption">Thanks for reading LeapNotes! Subscribe for free to receive new posts and support my work.</p></div><form class="subscription-widget-subscribe"><input type="email" class="email-input" name="email" placeholder="Type your email&#8230;" tabindex="-1"><input type="submit" class="button primary" value="Subscribe"><div class="fake-input-wrapper"><div class="fake-input"></div><div class="fake-button"></div></div></form></div></div>]]></content:encoded></item></channel></rss>