Francois Chollet is leaving Google

If I were to speculate, I would guess he quit Google. 2 days ago, his $1+ million Artificial General Intelligence competition ended. Chollet is now judging the submissions and will announce the winners in a few weeks. The timing there can't be a coincidence.

More generally, there is unlimited opportunity in the AI space today, especially for someone of his stature, and staying tied to Google probably isn't as enticing. He can walk into any VC office and raise a hundred million dollars by the end of the day to build whatever he wants.

$100M isn't enough capital for an AI startup that's training foundation models, sadly.

A ton of folks of similar stature who raised that much burnt it within two years and took mediocre exits.

Google, in my experience, is a place where smart people go to retire. I have many brilliant friends who work there, but all of them have essentially stopped producing interesting work since the day they started. They all seem happy and comfortable, but not ambitious.

I'm sure the pay is great, but it's not a place for smart people who are interested in doing something. I've followed Francois (and had the chance to correspond with him a bit) for many years now, and I wouldn't be surprised if the desire to create something became more important than the comfort of Google.

I wonder how/if that mentality will shift over time. As it seems the market capture phase it over and the current big tech aren't simply keeping top talent around as a capture piece anymore.

Maybe they'll still do it, but basically only if it feels you can startup a billion dollar business. As opposed to a million dollar one.

It’s the other way around. Working at Google (or any other FAANG) for a time period past your personal “bullshit limit” will ensure you will never do anything ambitious with your life ever again.

I think we'll start to see a differentiation soon. The likes of Ilya will raise money to do whatever, including foundation models / new arch, while other startups will focus on post-training, scaling inference, domain adaptation and so on.

I don't think the idea of general foundational model from scratch is a good path for startups anymore. We're already seeing specialised verticals (cursor, codeium, both at ~100-200m funding rounds) and they're both focused on specific domains, not generalist. There's probably enough "foundation" models out there to start working on post-training stuff already, no need to reinvent the wheel.

Not really any different than what happened to IBM, Intel, Cisco, etc.

The people that want to build great things want the potential huge reward too, so they go to a startup to do it.

Except… it’s about leverage/impact factor. Google has very large impact, so if you do something big and central you’re instantly in the hands of hundreds of millions / billions of people. That’s a very different situation than IBM or Cisco.

Interesting, I think $100M is totally enough to train a SotA "foundation model". It's all in the use case. I'd love to hear explicit arguments against this.

Am I almost alone in having no interest working for a large firm like Google?

I've been in tech since the 90s. The only reason I'd go is to network and build a team to do a mass exodus with and that's literally it.

I don't actually care about working on a product I have exactly zero executive control over.

ambitious? man, i can barely pay rent and i work at a FAANG.

There's a bunch of failed AI companies who raised been $100M and $200M with the goal of training foundation models. What they discovered is that they were rapidly out paced by the large players, and didn't have any way to generate revenue.

You're right that it's enough to train one, but IMO you're wrong that it's enough to build a company around.

Why zero executive control ? I’d expect a company like google ( like most large orgs ) to have a very large amount of internal code for internal clients, sometimes developer themselves. My experience of large orgs tells me you can have control over what you build - it depends on who you’re building it for ( external or internal)

That's not what I mean. I've got a deep interest in how a product is used, fits in a market, designed, experienced AND built.

If I went to Google what I'd really want to do is gather up a bunch of people, rent out an away-from-Google office space and build say "search-next" - the response to the onslaught of entries currently successfully storming Google's castle.

Do this completely detached and unmoored from Google's existing product suite so that nobody can even tell it's a Google product. They've been responding shockingly poorly and it's time to make a discontinuous step.

And frankly I'd be more likely to waltz upon a winning lottery ticket than convincing Google execs this is necessary (and it absolutely is).

I imagine Black Forest Labs (Flux) is doing alright, at least for now. I still feel like they’re missing out on some hanging fruit financially though.

But yeah, you’re not going to make any money making yet another LLM unless it’s somehow special.

Chollet is a leading skeptic of the generality of LLMs (see arcprize.org). He surely isn't doing a startup to train another one.

you can say this about any Fortune 500 corporation, to be honest

can you please name names? I can't think of any (but am not an expert on the space).

tbh working at google has a lot of advantages that a lot of hackers don't appreciate until they start trying to doing their own thing.

For one thing as soon as you start doing your own thing you will quickly find your day eaten up by a trillion of small little admin (filling reports, chasing clients for payments, setting up a business address) things that you didn't know even exist. And that is not even taking into consideration the buisness development side of thing (going to marketing/sales meeting, arranging calls, finding product/market fit!, recruiting, setting up payroll....) At google you can have a career where 90% of the time you are basically just hacking.

I'm guessing you've never experience working at an early stage startup?

At a 3 < n < 100 employee start up you absolutely are not "eaten up by a trillion small admin" and at the same time you can visibly see your impact on the product and company in basically real time. I've had work I've finished on a Monday directly lead to a potential major contract by Friday. I've seen features I've implemented show up in a pitch deck that directly lead to the next round of funding. Every single person on the team can personally point to something that they've done that has lead to our team's success so far. It's immensely rewarding to see a company grow and realize that without you personally, that growth wouldn't have happened in the way it did.

"90% of the time you are basically just hacking" is sounds fun, but I personally find it much more rewarding to see each week's work making incremental but visible changes not only in the product but the company itself.

My point is that if you build internal products usually there’s a lot less convincing to do, and it’s much easier to get a lot of control ( no marketing, communication, etc ).

Now, if you want to ship a product to millions of people _and_ have full control over it, then a large org is indeed not the right place.

Full control? nope.

A system to consider honest input without regard for job titles or hierarchy? yes!

For instance, I am not a UX designer but I do keep abreast of consumer perception and preference in whatever field I'm working in - almost like a stalker.

If a designer designs an interface and the feedback is clearly and unanimously negative, I should be able to present this and affect actual change in the product - not have my concerns heard, not considered, but to force actual remedial action taken to fundamentally address the issue.

If a competitor rolls out a new feature that is leading to a mass exodus of our customers, I should be able to demonstrate this without the managers whiffing about some vision that nobody gives a shit about or sprint planning responding to it in 6-months or having days of endlessly yapping. If the ship's got a leak my brother, it should be quickly and swiftly addressed.

It'd be like driving to lunch and your car catches on fire, you ignore it, and think about what you're going to be getting for dessert.

People realize these urgencies in IT/devops but teams that don't want to rock the boat as you gently glide over a waterfall are a complete waste of time.

So control? No. But if someone waves their hands and shout danger, they shouldn't be patronizingly patted on the head and told everything's under control.

In conventional large companies, that's exactly what happens. You're on a team, get assigned tickets, attend meetings, everyone calmly plays their roles and if you notice something in someone else's lane, you're supposed to politely stay quiet and watch everybody crash.

This says more about the cost of rent than it does the compensation of FAANG.

Not really. Despite having a platform with lots of people. Most are out of date software and hardly use any features.

It’s like the claim that Microsoft teams has hundreds of millions of users just because it’s installed on Windows by default.

Understood. Based on my many years in a large org, what you’re describing depends on the large org, and more specifically on management.

I’ve seen both : bad managers who let the boat crash and wouldn’t listen, and very good ones ( leading thousands of people ) understanding there was a problem, owning it and fixing it.

There are large orgs which are like what you want ( I work in one of them and that’s why I’m not leaving). I suspect there are not many of them though !

Used to be called IBM :-)