Meta announces a broad AI reorganization: VP of AI Jerome Pesenti will leave later in June, the Responsible AI team will join the Social Impact team, and more

Axios 2022-06-03 Ina Fried

Context & Ripple Effects

Meta had already moved its AI group into Reality Labs in an earlier shift toward metaverse-focused AI work. Pesenti’s departure and the transfer of Responsible AI to Social Impact mark another change in where AI leadership and oversight sit inside the company.

Later coverage shows Meta repeatedly redrawing these boundaries: it formed a dedicated generative-AI product team and subsequently dispersed the Responsible AI team across generative AI and infrastructure. The 2022 move is an early step in that organizational churn.

First-order effects

Jerome Pesenti’s exit removes Meta’s VP-level AI leader as the company reallocates responsibility for Responsible AI to the Social Impact organization.
The Responsible AI team moves out of the core AI structure, giving Social Impact direct ownership of that function.

Second-order effects

Meta’s AI product and research leaders must operate with Responsible AI reporting through a separate internal organization, making cross-team coordination a more explicit management task.
The transfer establishes a precedent for Meta to treat AI safety and social-impact work as an organizational function that can be repositioned as product priorities change.

Third-order effects

Meta’s later creation of a generative-AI team and eventual split between AI products and AGI Foundations point to a durable pattern: AI capabilities, product delivery, and governance are being repeatedly separated and recombined rather than held in one stable unit.

The trend: Meta is institutionalizing AI by continually reshaping the reporting lines among research, product development, infrastructure, and responsible-use functions.

Discussion

@arlanwashere @arlanwashere on x
@boztank ... Oh, I should mention that Boz wrote to me in dm very upset that I'd dare tweet this. I was taken by surprise by his response bc neither Satya nor Sam Altman reacted in such a way when I asked this very straight forward question, even though I'm sure it was uncomforta…
@arlanwashere @arlanwashere on x
@boztank Hey, Boz, you're an investor in @Backstage_Cap and someone I consider friendly, and I will ask you the same I asked @sama and @satyanadella: show us the photos or stats of the people of color working on this in a a significant way. I have faith in you and look forward to…
@garymarcus Gary Marcus on x
some restructuring of AI at Meta, with the open-minded @an_open_mind departing. Joelle Pineau is assuming more responsibility. I am a fan of hers and hope that she can make her replicability checklist an industry standard: https://www.cs.mcgill.ca/... https://twitter.com/...
@an_open_mind Jerome Pesenti on x
AI Innovation is now happening everywhere at Meta. It is transforming the way we create and deliver all our products. To accelerate this transformation, we are moving on from a central AI model while preserving our bleeding-edge AI research. https://ai.facebook.com/...
@gregmepstein Greg M. Epstein on x
@timnitGebru ... Don't know anyone involved personally, but speaking rudely to @ArlanWasHere of all people, and based on something like this? I saw Arlan's 1st message as generous. Investor status or no, if I'd done something like it I hope I'd be apologizing profusely & consider…
@schrep Mike Schroepfer on x
Some big news here: FAIR, Meta's amazing AI research lab, is joining forces with Reality Labs Research, which is inventing the technologies needed for the future of computing. An awesome combination of fundamental and applied research: https://ai.facebook.com/...
@metaai @metaai on x
Our new structure for Meta AI will help us not only better pursue open, ground-breaking research, but also improve how we leverage AI in our products. Learn more about how the team at Meta AI is evolving: https://ai.facebook.com/... https://twitter.com/...
@chriswallace95 Chris Wallace on x
@CaseyNewton Scalability. It's harder to ensure that each individual product team truly acknowledges what it means for AI to be “responsible”. Easier to have a centralized group that could block launches that are not responsible
@arlanwashere @arlanwashere on x
@boztank ... 🤷🏾‍♀ ️Boz, I said “photos or stats” for this very reason. And please don't act like I'm some evil person for asking for photos. There are team photos on hundreds of thousands of company websites. Let's be intellectually honest. I'm asking to see what sort of represen…
@nikki_t Nicole Tinson on x
@timnitGebru The audacity, indeed. And for him to say this to anyone, but especially to someone whose fund he INVESTED in, makes it wilder. I was baffled.
@boztank Boz on x
We're announcing a new structure to how Meta AI operates within the company. Advancing our fundamental capability is of critical importance, which is why we're making this shift to embed AI deeply everywhere. More on that here: https://ai.facebook.com/...
@timnitgebru Timnit Gebru on x
This fucking guy is unbelievable https://twitter.com/...
@boztank Boz on x
@ArlanWasHere ... I think it is inappropriate to ask me for photos of employees, as I said.
@timnitgebru Timnit Gebru on x
Not only have they perpetuated all sorts of issues for those who are a global majority but incredibly underrepresented, they have no shame in lashing out at someone who asks a simple question about the representation we know is non existent. The audacity my blood boil.
@timnitgebru Timnit Gebru on x
@_trlwilliams ... Right. Entitled anger and defensiveness, like how dare you even ask me this?
@_trlwilliams Travell Williams on x
@timnitGebru ... Its the microaggression for me. Its easier to come out and say “we barely have any people of color working on this but we're currently working on improving” vs “don't ask me about the makeup of the folx creating this”
@timnitgebru Timnit Gebru on x
@boztank ... Your lack of any shame here shouldn't be surprising to many of us but then again here we are. The problem is not Arlan's question but your racist company.
@boztank Boz on x
@ArlanWasHere ... I'm not putting up pictures of employees, though they are welcome to do so themselves if they choose. We release a regular report on these statistics companywide here: https://about.fb.com/...
@arlanwashere @arlanwashere on x
@boztank ... That's fair! I have no problem with you thinking something is inappropriate. I have a problem with you lashing out and cursing at me. But all good. I wish you'd just responded normally, and that it didn't feel so weird right now. But we'll both live.
@mergesort @mergesort on x
@CaseyNewton I don't know about Facebook specifically but sometimes it's better to give a team autonomy, or maybe different management. I tend to not read too much meaning into corporate politics even when it looks messy or subpar on paper.
@moonalice Roger McNamee on x
.@TIME asked me to write about Sheryl Sandberg's decision to step down as chief operating officer of Meta. I believe Sheryl will be remembered most for choosing to maximize profit over democracy, public health, and users' right to self-determination. https://time.com/...
@sadlyitsbradley Brad Lynch on x
A lot of higher ups at Meta leaving. And weird things getting leaked freely Something doesn't feel right https://twitter.com/...
@inafried Ina Fried on x
Meta reorganization aims to decentralize Facebook's AI efforts; VP Jerome Pesenti to leave this month https://www.axios.com/...
@stevejobs_666 Steve on x
Sheryl Sandberg announced her departure yesterday, Meta VP of AI announced his departure today 🧐 Something bigger going on?
@unusual_whales @unusual_whales on x
Meta, $FB, says vice president for AI Jerome Pesenti will depart co in mid-June. (Another one leaves)
@caseynewton Casey Newton on x
Is there a good reason to separate the team in charge of making AI “responsible” from the team that is actually building products https://www.axios.com/... https://twitter.com/...
@schrep Mike Schroepfer on x
AI is becoming the foundation of pretty much everything that matters at Meta - better recommendations, platform safety, revenue, and increasingly building for the Metaverse - which will be the way most of us interact with AI every day... https://www.axios.com/...

Chronicles