Départ de Yann LeCun, frustrations de la nouvelle équipe : tout ce qu'il faut savoir sur le nouveau virage IA de Meta
Bonjour,
Aujourd’hui petite édition spéciale sur le bazar chez Meta/Facebook.
Allez bien jusqu’en bas de l’email, j’y présente une surprise en relation avec le sujet du jour ;)
Meta en perdition sur l’IA ?
Meta’s strategy shifted dramatically earlier this year after the company released Llama 4, an open-source model that disappointed Silicon Valley and Zuckerberg, Meta’s chief executive officer.
He sidelined some of the people who worked on that project and personally recruited top AI researchers and leaders, in some cases offering them hundreds of millions of dollars in multiyear pay packages, and some, like Alexandr Wang, who came in through a $14.3 billion investment deal.
Mmmmm… While Wang isn’t considered a technical AI researcher, he knows the industry through relationships he built at Scale and has impressed his new colleagues with his ability to get things done quickly and offer a big-picture vision, the people said.
Mr. Zuckerberg and Mr. Wang then went on a recruiting blitz for top A.I. researchers, offering pay packages worth hundreds of millions. Mark Chen, OpenAI’s chief research officer, said Mr. Zuckerberg had gone as far as delivering homemade soup to the doorsteps of some OpenAI employees to persuade them to join his company.
A month later, Meta restructured its A.I. division into 4 groups.
One was focused on A.I. research,
one on products
and one on infrastructure such as data centers and other hardware,
while TBD Lab (for “to be determined”) was focused on building superintelligence.
Collectively, the groups worked under an organization called Meta Superintelligence Labs. Mr. Wang would lead the entire effort.
The TBD Lab was placed in a siloed space next to Mr. Zuckerberg’s office at the center of Meta’s Silicon Valley headquarters, surrounded by glass panels and sequoia trees. Mr. Zuckerberg wanted to separate the new A.I. group from the bureaucracy of the company, which owns Facebook, Instagram and WhatsApp, two people with knowledge of the matter said.
Surprenant : The TBD group is using several third-party models as part of the training process for Avocado, distilling from rival models including Google’s Gemma, OpenAI’s gpt-oss and Qwen, a model from the Chinese tech giant Alibaba Group Holding Ltd., the people said.
Training the new model on Chinese technology signals a shift in tone for Zuckerberg. “China is well ahead — way ahead on open-source,” Nvidia Corp. CEO Jensen Huang said in early Dec 2025.
On an earnings call with investors in late July, Zuckerberg hinted that the company would pursue both open and close models moving forward.
Zuckerberg has pledged to spend $600 billion on US infrastructure projects over the next three years, most of them AI-related.
He’s also redistributing some of Meta’s investments internally, planning meaningful cuts to its virtual reality and metaverse efforts in favor of putting that spending into AI glasses and other related hardware.
Alexandr Wang has at times been frustrated by what he sees as micromanagement, according to people familiar with the matter. Zuckerberg has a long history of getting intensely hands-on with the company’s most important products.
Five months later, that divide has become more than physical.
In meetings this fall, Mr. Wang has privately told people that he disagreed with some of Mr. Zuckerberg’s longtime lieutenants, including Chris Cox, the chief product officer, and Andrew Bosworth, the chief technology officer, the people briefed on the conversations said.
TBD Lab’s researchers have come to view many Meta executives as interested only in improving the social media business, while the lab’s ambition is to create a godlike A.I. superintelligence, three of them said.
Some Meta employees have also disagreed over which division gets more computing power. Those working on the social media ranking algorithm have argued that more of the company’s new computing power should be used to improve their business rather than training A.I. models, people with knowledge of the matter said.
Roughly six months into Meta’s AI pivot, the team has been mostly heads-down without much to show publicly, yet. The news that has trickled out has skewed negative. Some new hires from Meta Superintelligence Labs have departed within weeks of arriving. In October, Meta eliminated 600 jobs from its AI unit, with deep cuts to its more academically focused unit, FAIR. LeCun left the company a month later, intent on doing his research elsewhere.
Wow: Yann LeCun, known as one of the godfathers of AI, recently left the company after years leading Meta’s long-term AI research group, in part because of frustrations that he couldn’t get enough resources, Bloomberg News reported.
Prior to his departure, some employees had been encouraged to keep LeCun, who was a big proponent of open-source technology, out of the spotlight, including at public speaking events, the people said. Meta no longer saw him as emblematic of the company’s AI strategy, and couldn’t trust that he’d stay on message, they added.
Also: Zuck made LeCun report to its head, Alexandr Wang, a new employee that is 28-year old!!
Sérieusement ?
The most high-profile product to come out of MSL has also received mixed reviews. To get ahead of OpenAI’s release of Sora 2, a video-generation tool, Meta rushed out its own product called Vibes using technology that it licensed from AI startup Midjourney Inc. While Meta insiders insist that Vibes is performing well in terms of usage, Sora’s launch just one week later quickly overshadowed Vibes in the press and online.
Ils n’ont que ça à faire ?
This summer, Meta leadership tasked data-focused employees with conducting market research on how the term would land with policymakers in the US and abroad. Working with external consultants, they found that “superintelligence” triggered fears of AI’s unchecked power, particularly in Europe, where regulators have already come after Meta’s AI offerings, the people said.
Et maintenant, voici une opportunité rare : la possibilité d’investir dans l’entreprise qu’est en train de monter Yann LeCun !
J’ai accès à une allocation que Yann LeCun a accordé à un ancien collègue français de chez Facebook et aujourd’hui dans la finance.
Yann LeCun (ancien Chief AI Scientist chez Meta, lauréat du prix Turing, et l’un des trois fondateurs de l’IA moderne) vient de quitter Meta pour créer sa propre boîte, AMI Labs. Elle devrait être fondée en France !
Il s’attaque à ce que beaucoup considèrent comme l’architecture qui va remplacer les LLMs : les world models. FT / BBC / WSJ en ont déjà parlé — c’est la story AI que tout le monde suit dans l’industrie.
Yann LeCun (source) :
“The goal of [my new] startup is to bring about the next big revolution in AI: systems that understand the physical world, have persistent memory, can reason, and can plan complex action sequences.”
La levée sera de plusieurs centaines de millions de dollars et devraient être le début d’une séquence rapide de gros financements — avec un horizon de liquidité possiblement bien plus rapide que du deep tech early-stage classique.
C’est un accès extrêmement rare à l’un des founders AI les plus importants au monde — à la formation — dans ce qui est largement considéré comme la prochaine catégorie AI à un trillion de dollars (robotique, autonomie, agents, transport, santé, intelligence du monde réel et automatisation industrielle).
L’investissement se fera via un SPV hébergé sur la plateforme Roundtable.
Si vous êtes intéressé(e) et pouvez investir minimum 100k€, vous pouvez répondre à cet email :)
Some other quick news
Are you surprised? Elon Musk’s xAI Is Reportedly Having Trouble Selling Grok to Enterprises (source)
Crazy stat: Nigeria, with 240m people, generates less electricity than the American state of Wyoming, which has 0.6m inhabitants.
Google CEO Sundar Pichai says we’re just a decade away from a new normal of extraterrestrial data centers (Fortune)
Why Disney is lending its characters to OpenAI’s Sora app (Ben Evans, WSJ)
Disney was once so protective of its intellectual property that it threatened to sue day-care owners for painting Mickey Mouse on the walls. But last week it announced it would allow people using OpenAI’s video generation app Sora to create videos with Pixar, Marvel, and Star Wars characters, and also that it was investing $1 billion in OpenAI. Disney sees it as a bet on the future of entertainment.
Comment créer une web app de recherche immo en quelques secondes avec Google AI Studio (JDN)
Parlons passé : sur 30 millions de Français au début du XIX siècle :
6 millions ne parlaient pas du tout français
6 millions le comprenaient très difficilement
3 millions seulement le parlaient parfaitement bien
Sergey Brin on the genius of Jeff Dean. (source)
Jeff Dean: Since 2018, he has been the lead of Google AI. He was appointed Google’s chief scientist in 2023 after the merger of DeepMind and Google Brain into Google DeepMind.
Google’s home-designed chips, the TPUs, were Jeff’s idea. He calculated that if users spoke to Google for just three minutes a day, Google would have to double its CPU data centers. Instead of buying more CPUs, Jeff decided to build a new chip for AI.
What’s fascinating is that when AI was not as big as today, when it can only “tell cats from dogs,” Larry and Sergey were like, “Cool, let’s make a custom chip for it.”
That’s a huge show of confidence in deep tech and top-tier technical talent. It makes Google the only AI company that has the top models, top chips, and top data centers.
Disappointing: Finland’s €560-per-month Universal Basic Income trial showed no change in crime and victimization (source)
The Data on Self-Driving Cars Is Clear. We Have to Change Course. (NYT)
The self-driving car company Waymo recently released data covering nearly 100 million driverless miles in four American cities through June 2025, the biggest trove of information released so far about safety.
I spent weeks analyzing the data. The results were impressive. When compared with human drivers on the same roads, Waymo’s self-driving cars were involved in 91% fewer serious-injury-or-worse crashes and 80% fewer crashes causing any injury. It showed a 96% lower rate of injury-causing crashes at intersections, which are some of the deadliest I encounter in the trauma bay.
If Waymo’s results are indicative of the broader future of autonomous vehicles, we may be on the path to eliminating traffic deaths as a leading cause of mortality in the United States. While many see this as a tech story, I view it as a public health breakthrough.
In medical research, there’s a practice of ending a study early when the results are too striking to ignore. We stop when there is unexpected harm. We also stop for overwhelming benefit, when a treatment is working so well that it would be unethical to continue giving anyone a placebo. When an intervention works this clearly, you change what you do.
There’s a public health imperative to quickly expand the adoption of autonomous vehicles. More than 39,000 Americans died in motor vehicle crashes last year, more than homicide, plane crashes and natural disasters combined.
Crashes are the No. 2 cause of death for children and young adults. But death is only part of the story. These crashes are also the leading cause of spinal cord injury. We surgeons see the aftermath of the 10,000 crash victims who come to emergency rooms every day. The combined economic and quality-of-life toll exceeds $1 trillion annually, more than the entire U.S. military or Medicare budget
(by Dr. Jonathan Slotkin is an executive and vice chair of neurosurgery at Geisinger Health System in Pennsylvania. He is a co-founder and general partner of Scrub Capital, a venture capital firm that invests in health care start-ups.)
Vous recevez la version gratuite de la newsletter Parlons Futur : une fois par semaine au plus, une sélection de news, mêlant sources anglophones et francophones, résumées en bullet points sur des sujets tech 🤖, science 🔬, éco 💰, géopolitique 🌏, défense ⚔️ et espace 🚀 pour mieux appréhender le futur 🔮.
Je m’appelle Thomas, co-fondateur de YeldaAI, qui développe des IA pour répondre au téléphone en langage humain pour les administrations et les entreprises. Plus d’infos sur moi en cliquant ici.
Mes derniers podcasts
Pourquoi le CEO de Nvidia investit 10 milliards $ dans Anthropic peu après avoir taclé son CEO (8 minutes, Youtube, Spotify, Apple)
Vrai ? Le CEO d’Anthropic AI : “Nous aurons 100 ans de progrès en biologie en 5 ans grâce à l’IA” (9 minutes, Youtube, Spotify, Apple)
Pourquoi Yann LeCun pense que l’Artificial General Intelligence (AGI) est un concept bidon (8 minutes, Youtube, Spotify, Apple)
Sommes-nous prêts à nous entourer de millions de robots humanoïdes ? (Youtube, Spotify, Apple)
🧠Contre-intuitif : pourquoi l’intelligence humaine baisse en fait depuis 10,000 ans (Youtube, Spotify, Apple)
🌍Entretien exclusif avec Philippe Bihouix, co-auteur de la BD à succès “Ressources : Un défi pour l’humanité” (Youtube, Spotify, Apple)
Pour vous inscrire et ne pas manquer les prochaines éditions, vous pouvez cliquer ici.
Merci, et à bientôt !
Thomas