[HN Gopher] Character.ai CEO Noam Shazeer Returns to Google
___________________________________________________________________
Character.ai CEO Noam Shazeer Returns to Google
Author : treesciencebot
Score : 85 points
Date : 2024-08-02 18:06 UTC (4 hours ago)
(HTM) web link (techcrunch.com)
(TXT) w3m dump (techcrunch.com)
| treesciencebot wrote:
| Was this an Inflection.ai style acquisition considering C.AI was
| profitable?
| fnbr wrote:
| do we know they were profitable? I doubt it, if they pulled
| this. I think they had high DAU/MAU but low paid users.
|
| this is basically Inflection 2.0.
| jsheard wrote:
| All public information about their finances only mentions
| their revenue, not their profits, which means they're almost
| certainly not profitable.
| beoberha wrote:
| I would be absolutely floored if any of these new crop of
| AI companies set profitable
| minimaxir wrote:
| In contrast to something like Adept.ai, this _appears_ just be
| the CEO + some key employees exiting, not a complete team
| poaching that leaves the company a shell of what it was.
|
| It doesn't necessairly imply Character.ai's business isn't
| doing well, but a CEO leaving is still indeed weird.
| tikkun wrote:
| They are likely unprofitable.
| tikkun wrote:
| I'm puzzled. Did Noam get $500mm+ from this?
|
| Character.ai could've been at $100mm+ ARR if they did a bit more
| of a monetization push based on my very rough estimates. If it
| was an acquisition I would've been imagining $3b+ price range.
|
| Huge get by Google! (Side note, Gemini 1.5's new alpha release
| from this week is now at the top of the lmsys leaderboard and
| sentiment on twitter for it is that it's strong, maybe as strong
| as sonnet 3.5, so it'll continue to be an interesting race
| between meta, openai, anthropic, gdm.)
|
| Edit: Okay perhaps it's - Noam keeps his C.ai stock, gets a big
| pay package from Google ($5mm-$15mm/year kinda range? not sure).
| Most of the value in C.ai remains, and he keeps his stock.
| riku_iki wrote:
| > I'm puzzled. Did Noam get $500mm+ from this?
|
| where did you get $500mm number?..
| tikkun wrote:
| I imagine he still had 15%+ of the company, so if there was a
| big payout, he would've gotten a big slice. Treat all my
| estimates with a grain of salt, though, please. I'm imagining
| he didn't get a huge payout and he actually kept most of his
| equity?
| riku_iki wrote:
| I think he just steps down as CEO, there is no company
| acquisition happening, so no one gets paid. I think company
| essentially failed to generate revenue and/or get new
| investment round.
| tikkun wrote:
| Yes that seems right. Though I think they could've gotten
| a new investment round if they wanted one, I'd be
| surprised if that was the cause.
| BoorishBears wrote:
| https://www.theinformation.com/articles/google-hires-
| charact...
|
| Investors are being bought at 2.5B valuation
| riku_iki wrote:
| I don't trust that site. They spread unconfirmable rumors
| behind paywall.
|
| Base on those numbers, $150M of series A with 1B
| valuation would be bought now for 375M(which is high for
| failed startup), my understanding is that such
| transaction has to be disclosed.
| kkakkx wrote:
| This is public info. 2.5b and $88 per share for all
| riku_iki wrote:
| info becomes public when it is officially announced by
| some of the transaction participants or disclosed in some
| fillings.
|
| I may be too skeptical, but it is just hard to believe
| someone (who?) throws 400M (to seed + A investors) plus
| significant amount to founders into essentially failed
| startup.
| jsemrau wrote:
| I suppose a lot of the current batch don't have a business
| model that is sustainable. Another candidate is stability ai.
| dekhn wrote:
| Noam is pre-IPO google with extensive work done on the most
| important infrastructure in the company's history; likely he
| has large monetary reserves as well as a pool of investors more
| than happy to give him money _simply to continue doing his
| research, regardless of future revenue_. You can assume that in
| returning to google, he will get some base pay which will be
| prodigious, but also extensive pay in the form of bonuses and
| stock. From what I know of him, the real motivation is to get
| more access to large TPUs.
| jchonphoenix wrote:
| This is another inflection style "acquisition." Highly unethical
| of the founders and screws over all your employees and investors
| who are left holding the bag.
|
| For those asking, c.ai has very high cost and looks like a
| typical consumer company that burns money for use, so they were
| decent on revenue but not near profitability.
| nextworddev wrote:
| don't forget Adept
| hankchinaski wrote:
| Everybody has a price
| alphabetting wrote:
| > Character's leaders told staff on Friday that investors would
| be bought out at a valuation of about $88 per share. That's
| about 2.5 times the value of shares in Character's 2023 Series
| A, which valued the company at $1 billion, they said.
|
| https://www.theinformation.com/articles/google-hires-charact...
| pton_xd wrote:
| > investors would be bought out at a valuation of about $88
| per share. That's about 2.5 times the value of shares in
| Character's 2023 Series A, which valued the company at $1
| billion
|
| Those investors almost certainly have a liquidation
| preference. How much did employee shareholders get? I'd guess
| zero.
|
| "I am confident that the funds from the non-exclusive Google
| licensing agreement, together with the incredible
| Character.AI team, positions Character.AI for continued
| success in the future," Shazeer said in a statement given to
| TechCrunch."
|
| That's a pretty hilarious statement from a Founder/CEO, given
| the circumstances.
| fngjdflmdflg wrote:
| >liquidation preference. How much did employee shareholders
| get? I'd guess zero.
|
| But they aren't filing for chapter 11? I assume all
| shareholders will be bought out, including the employees,
| and this will be paid for by Google who will license their
| models, presumably as a scheme to pay off of the investors
| as I doubt they actually need those models at all.
|
| (assuming the linked source is correct.)
| hobofan wrote:
| I'm also wondering how much money they spend on legal fees,
| given that they are copying then likenesses of many celebrities
| without their permission (that's the only way I've heard abot
| them before).
| mafuyu wrote:
| To do better by the employees, the CEO really should have
| fought harder to have the whole company get acqui-hired, even
| if Google would have shut down the service. Maybe there were
| some other considerations that I'm not seeing (ie. There are
| good reasons the company should be kept going, and there's a
| good path to success even without Noam. The article doesn't
| specify how many employees are going over, so it's hard to
| tell.) Landing the employees a relatively cushy Google SWE gig
| after helping build your company is the least you could do for
| them.
| simonw wrote:
| Key quote (from Character.AI):
|
| "Over the past two years, however, the landscape has shifted;
| many more pre-trained models are now available. Given these
| changes, we see an advantage in making greater use of third-party
| LLMs alongside our own. This allows us to devote even more
| resources to post-training and creating new product experiences
| for our growing user base."
|
| My interpretation is that Character.AI realized they don't
| actually need to train their own foundation models from scratch
| to support their product - they can build cheaper, faster and
| probably better if they use LLMs trained by other companies
| (could be GPT-4o/Claude/Gemini via APIs, could be Llama 3.1 self-
| hosted).
|
| If they're not training foundation models any more, the talents
| of people like Noam Shazeer aren't so important to them. They
| need to focus on product development instead.
| htrp wrote:
| I'd argue that their own foundational models are getting
| outperformed by the Llama finetunes on HF and at this point
| they're shifting cost structures (getting rid of training
| clusters in favor of hosted inference).
| zoogeny wrote:
| I think this highlights the winner-take-all stakes of
| intelligence. It also suggests that there is little to be
| gained by specialization. Building a brand might actually be
| more short-term profitable since you can swap in the latest AI
| models as they become available. In other words, if advancing
| the SOTA AI is your dream, a product company may not be the
| right place. And if building a product company is your dream,
| then building foundational AI might not be the best strategy.
| robrenaud wrote:
| > if advancing the SOTA AI is your dream, a product company
| may not be the right place.
|
| Does Meta get in the way of this?
|
| It's hard to compete with a company that is dead set on
| spending billions and seemingly wants to drive your SOTA AI
| product revenue to 0.
|
| If you are OpenAI or Anthropic right now, it seems like
| trying to run a great restaurant at a reasonable price right
| next to a good (great?) restaurant that is serving everyone
| for free.
| simonw wrote:
| Yes. Meta's current strategy is extremely disruptive to
| other companies that are trying to build a business in the
| foundation model space.
|
| Presumably this is because Meta desperately want to avoid
| becoming dependent on other companies in this new
| generative AI world. Mark Zuckerberg talks about not
| wanting a repeat of the Apple tax in his post about Llama
| 3.1 here: https://about.fb.com/news/2024/07/open-source-ai-
| is-the-path...
| zoogeny wrote:
| My take is that this has more to do with the coming years
| than the current climate.
|
| I think it is just a consequence of the cost of getting to
| the next level of AI. The estimates for training a GPT-5
| level foundational model are on the order of 1 billion. It
| isn't going to get cheaper from there. So even if your
| model is a bit better than the free models available today,
| unless you are spending that 1 billion+ today then you are
| going to look weak in 6 months to 1 year. And by then the
| GPT-6+ model training costs will be even higher, so you
| can't just wait and play catch up. You are probably right
| as well, in that there is a fear that a competitor based on
| an open source model gets close enough in capability to
| generate bad publicity.
|
| I imagine character.ai (like inflection) did calculations
| and realized that there was no clear path to recoup that
| magnitude of investment based on their current product
| lines. And when they brainstormed ways to increase return
| they found that none of the paths strictly required a
| proprietary foundational model. Just my speculation, of
| course.
| nickfromseattle wrote:
| > If they're not training foundation models any more, the
| talents of people like Noam Shazeer aren't so important to
| them.
|
| Why is the CEO important to model development regardless of
| talents? They've raised $150m+, have $15m+ ARR and ~200
| employees, etc. Shouldn't the CEO be CEOing?
|
| Edit: reading the comments below, it seems like maybe he
| thought the expected value of attempting to clear the hurdle of
| their valuation/liquidation preferences at a $250k/year salary
| as CEO was lower than a $5m+/year salary/RSUs from Google?
| htrp wrote:
| >Character.AI co-founder Daniel De Freitas is also joining Google
| with some other employees from the startup. Dominic Perella,
| Character.AI's General Counsel, is becoming an interim CEO at the
| startup.
|
| So basically leaving a shell of a company and the GC to try and
| run it / wind it down.
| bogwog wrote:
| Is this a cheat code for doing an acquisition without FTC
| scrutiny?
| tsunamifury wrote:
| And screw investors
| agnokapathetic wrote:
| per The Information -- investors don't get screwed, the
| board repurchased all shares at $88/share ($2.5B
| valuation).
| blackhawkC17 wrote:
| The money to repurchase shares must be coming from an
| external source?
|
| I assumed Character.AI wasn't profitable, like most AI
| startups.
| tsunamifury wrote:
| These two stories are hard to rectify unless the Google
| licensing deal is a lump sum of several billion dollars.
| Which feels borderline impossible. Even Microsoft didn't
| do that with open AI and the AI trade is cooking very
| fast.
| fnbr wrote:
| They raised ~$200M total, and $150M at the series A. So
| if they're just paying back investors, they'd "only" need
| $500M.
| ko_pivot wrote:
| Yes. Character.ai usage is way down from the peak and it is
| only getting easier for startups to compete in the space.
| However, the plausibility of an actual FAANG acquisition is
| almost zero, so this is the next best option... at least for
| certain stakeholders. Microsoft did this with Inflection.
| tikkun wrote:
| What was the peak? I see estimated 211 million visits in
| June 2024, vs 86 million in Feb 2023.
| tsunamifury wrote:
| Visits is an entirely meaningless metric
| crowcroft wrote:
| The new VC playbook could be.
|
| 1. Find an unhappy senior AI exec from who's published a few
| papers who's published a few papers. 2. Start a new org around
| them and hire a few key people with crazy salaries (which you can
| offer cause the time horizon for the company isn't that long). 3.
| Train a few models, release some good looking benchmarks (bonus
| points if big tech lend you their GPUs as part of some
| 'accelerator deal'). 4. Maybe find PMF and become incredibly
| rich. 4. If that fails, sign a massive but undisclosed licensing
| deal for your tech with big tech and give them your staff.
|
| Seems like a good way to take big bets in AI, while hedging most
| of the risk.
| m_ke wrote:
| In the CNN boom days after alexnet the playbook was: take your
| research lab, slap a c crop and new logo on it, keep doing your
| research and get acquired by FANG for high 8 figures.
|
| Only difference in this cycle is that you can't do real LLM
| research in academia these days so all of the top researchers
| are already at FANG.
| candiddevmike wrote:
| This isn't #4, there's no acquisition happening.
| beoberha wrote:
| AI is quickly becoming a commodity which is awesome for
| consumers. I think we're going to see a huge shakeout of
| companies who dazzled with the initial allure of LLMs being
| replaced by the "killer AI apps" where we really start to rethink
| modern computing.
| bentoboox wrote:
| This is likely an aqui-hire structured to not trigger an anti-
| trust probe.
|
| If there was any real intent to give c.ai a chance as a real
| business they would have hired a new ceo before the announcement.
| moralestapia wrote:
| >In a big move, Character.AI co-founder and CEO Noam Shazeer
| [...]
|
| It's like those rich guy/poor guy jokes.
|
| Poor guy leaves his job, tries bootstrap its own company and
| fails, comes back to its old job. "What a loser".
|
| Rich guy leaves his job, gets 150M to start a company in a blue
| ocean with a significant competitive advantage over 99.99% of
| humans alive. Still manages to fail and comes back to its old
| job. "What a bold move!"
| SXX wrote:
| > gets 150M to start a company
|
| Oh there is a place where rich guys can "get" $150M, but poor
| guys couldn't.
| philip1209 wrote:
| Seems like a way to do an "acquisition" while avoiding the brand
| risk of buying a mostly-porn company.
| atlasunshrugged wrote:
| And I'm sure Google is hopeful that it also carries low
| regulatory risk from the FTC/DOJ
| stephencoyner wrote:
| Here's their press release: https://blog.character.ai/our-next-
| phase-of-growth/
| dekhn wrote:
| Noam Shazeer is a long-time googler who worked on machine
| learning for quite some time (take a look at his patent history,
| for example https://scholar.google.com/citations?view_op=view_cit
| ation&h...)). He was a favorite of Jeff Dean and did some of the
| most impressive work on ML that I saw at Google. I think at some
| point in the past, Noam saw that google wasn't supporting his
| work very well (google research went through a dark time where
| many researchers with creative ideas were shut down, either for
| business or reputation reasons, see
| https://www.businessinsider.com/google-ai-characterai-ceo-no...)
| and I figure he made a startup because it was a more convenient
| position for him to do research, even if there wasn't a strong
| revenue model. Now he returns to Google in a position of deep
| strength and will be able to continue to pursue extremely
| ambitious ideas with far less restraint.
| tsunamifury wrote:
| I think your strongly strongly overestimate how fast Google
| Forgets about you. Deep strength is a very rosey way to Put it.
| underdeserver wrote:
| They forget about most people quick. Noam Shazeer is not most
| people.
| tsunamifury wrote:
| Have some familiarity with the players here. The founders are
| core model nerds and accidentally happen upon success as a
| majority sex chat product. They have little interest in that and
| investors are likely saying they won't support any more core
| model research. The Google deal lets the founders go back to
| doing core model work for Google and the company to focus on a
| consumer only product that uses third party models.
___________________________________________________________________
(page generated 2024-08-02 23:01 UTC)