Stability AI Head of Research Resigns from Startup
bloomberg.com
bloomberg.com
“I have Asperger’s and ADHD, and I have a very definitive view of the future,” he (Emad Mostaque, CEO) said on stage. “I think that shocks people because they can’t deal with the exponentials.”
What does this even mean?
Against this there is minimal risk in obtaining large scale supercompute which will be a scarce asset, funding will not peak for a few years and there is no way to scale to meet demand.
Almost nobody will also train their own models and open, auditable, models will be legally required for every government, regulated industries and more.
My role is to allocate the compute, build the partnerships & gather people who believe that this technology should be distributed globally which is why we have a very diverse team.
I am not very good with people though.
It's not about being good with people. Plenty of people are introverts - that's fine. Many are neurodiverse. However, it's pretty important in society to be honest and not-fraudulent. You've lost a lot of trust. It's not that you're not good with people. You're manipulative and people are not good with you anymore. https://www.forbes.com/sites/kenrickcai/2023/06/04/stable-di...
Even met the family and made fun of my work on my sons autism which shows what they are like.
How many companies preseed get a 4,000 A100 cluster dedicated built from AWS etc?
I wrote a reply anyway as this stuff makes me sad
Happy to reply to any issues you may have here too
This wording really doesn’t help your case. It sounds like a total exaggeration.
Link to the AWS case study that highlights this relationship, or absolutely anything than just repeating “we are one of the only companies to have AWS buy a million GPUs for us and let us use them and it’s one of the only ones in this part of AWS’s network. Source: trust me bro”
https://twitter.com/emostaque/status/1553120829336129537?s=4...
It was super interesting as an experience.
At the very least anyone who has used AWS for ML can tell you that getting a 40 A100 quota let alone 4,000 is a feat in itself ^_^
I doubt you yourself can clear this up (it's a he-said/she-said conundrum) - but perhaps Eric is around to confirm/deny this?
We missed one invoice and accounts asked him to send details so he could settle but instead he decided to a month later say that stuff to Forbes then said straight out he didn’t want it paid
https://twitter.com/erichallahan/status/1639695790724591618?...
https://twitter.com/emostaque/status/1665520024080068609?s=4...
Idk I hope he’s ok
https://twitter.com/emostaque/status/1671287589507063812?s=4...
He's a pretty bright guy in my interactions with him and his heart is in the right place. I'm willing to bet he lands on his feet elsewhere and does just fine in life.
And from where I stand, a person's paycheck is a metaphor for their livelihood, sometimes a very literal metaphor. Missing one payment might seem mundane, but it is entirely outside of the norm and is _definitely_ a situation which should cause employees to re-evaluate how much they trust their employer. It's probably not wise to imply anyone who does this is "overreacting as that will simply further erode the trust. Instead, you apologize, apologize, and apologize again and again. Which isn't me asking you to post five tweets where you apologized.
I agree Eric is bright and so wanted to support him through an internship where he had lots of freedom.
He didn’t manage to do it (as noted in his tweets) but I offered to pay anyway and also offered other support.
One invoice went to the wrong address (our fault) and as soon as he asked about it we apologised and asked him for details so we could settle it immediately as well as asking him for the third invoice he never sent (again for work not done).
We made mistakes and struggled as we scaled, faster than anyone could have imagined but always try to improve as folk should.
Power differentials are tough and complex but in this case we tried to be understanding and apologise when at fault and now he is on a crusade against the org per his tweets and it’s a bit tiring when half the org was hired from non traditional backgrounds and the community.
I honestly wish the best for him and hope that hate dies down, he will be happier that way.
Just thought I'd give you a bit a positivity :)
Sincerely a fellow person
>Pressures
>Literal livelyhood
Cool your jets, guy. They sent the payment to the wrong address and when he brings it up they reach out to him asking for for the new address and you're acting like some capitalist kingpin stole from some poor honest laborer. The terms you use make you come off as a Twitter outrage addict. They're red flags when used liberally to something minor like this.
Reading the interactions between Callahan and Mostaque, Callahan seems more interested in scoring outrage points on Twitter than the money, does that indicate a desperately broke person whose "literal livelyhood" is at stake? If I do a gig for a company and they mess up the check and asks me to confirm the address to send me a new 6k check, I don't go on Twitter and complain "Waaaah!!! I shouldn't have to do this! A PROPERLY RUN company wouldn't ask this! I can't even rn!", etc, I simply give them the updated info and the matter is closed. And the hilarity of Callahan whose was too mentally fragile and inexperienced to finish an internship (by his own admission) acting like he knows how well-run large companies function!
I'm gonna say what his daddy should have told him years ago: take some responsability, and stop being so weak. This goes for Callahan and to people like you who are blowing up this nothingburger.
In reality, Mostaque has a bachelor’s degree, not a master’s degree from Oxford.
Officially, the degree is not 'awarded' until you attend a graduation ceremony, or graduate 'in absentia' (i.e. have your name read out at a graduation ceremony).
You don’t actually get your bachelors or masters without either going to ceremony or asking for it in absentia so like I’m going to get them in a few weeks, do a photoshoot
“You make more money selling advice than following it. It’s one of the things we count on in the magazine business – along with the short memory of our readers.” — Steve Forbes
I'm with you on that one. And I'm not sure we are even close to the level of preparedness that we need (as a society) to weather that.
I don't want to see interviews about you quitting and saving the world, I wanted to see how you want to make money, and finally this last was honest. I guess you had to be quiet before, but I guess we had enough ,,altruisms'' and congress hearings with a few Sams already.
(I also have Aspergers though, so not representing most people)
I have written about our business model many times here is a thread
https://twitter.com/emostaque/status/1649152422634221593?s=4...
Also did a recent podcast on various AI business models with Sam Lessin that Nathan covers here
If the value is in better training data then everyone will train their own models. We should have a goal where we have domain-specific models that make us more efficient, but also constantly train a context-aware model based on individuals. Model composition is the future.
Models are like talented grads that occasionally go off their meds, proprietary black box models are like hiring them from McKinsey, fine tuning your own ones that you control is like hiring them
All govts, regulated industries etc will use open auditable models so much bigger TAM
How does having ADHD and/or Asperger's mean that? Are you implying that those give people high confidence?
The western world desperately needs a way to do resource allocation that is parallel to and augments financial markets. History did not end, and the "joker with the most cash is dictator of the org" model has become perhaps the largest existential threat to the western world.
This is highly efficient allocation as the code and models can be used widely
Nobody else is really doing this I do wonder at times if folk want us to stop
Please don't. You're doing humanity a service. AFAIK you're the only AI company who released something of this magnitude for free, without any restrictions.
The people who want you to stop are:
1. Bitter artists who see their livelyhood endangered and lash out at the most prominent example. We can sympathize but this isn't your fault, this was going to happen anyway. At least you democratized it.
2. People upset in general that some tech companies will be getting rich off AI while the rest of us get poorer, lose jobs, etc. You get the brunt of the hate because your free tools are usable by the average person, and therefore more people are exposed to it. They don't realize that if SD never existed, the only thing that would change is that this technology would still exist, except we'd all be paying some OpenAI-like company a monthly fee to use it.
3. Big NGO. Their raison d'etre is fighting against something, no matter what, because this is how they get donations/attention, which pays for the salaries of the trust fund kids who work at those places. Apparently if someone generates an image of a naked kid using SD, I'm supposed to care as if it was actual child abuse material? Somehow we don't extend that logic to the manufacturers of pen and paper which allow that same person to draw or write their disgusting fantasies.
4. Clueless people who fall for the propaganda from the least ethical members of the previous groups.
5. SD fans upset that you bent to the pressure from NGOs and Twittards and kneecaped SD in your rush to remove copyrighted art from the training data. Newer versions are objectively inferior from an end-user's perspective and deep down you know it, because you see what enthusiasts are using on /r/stablediffusion. But I don't hold this against you, even if it's disappointing. We will always have SD 1.5, which remains the best way for normal people (ie not rich VC corps) to do text-to-image. If you do nothing else for the rest of your life, you will be celebrated for this.
Stay up, playa'!
Something like a government that taxes the people and spends the tax revenue on stuff? In the USA, government spending accounts for 37% of GDP [1].
The fact that some individuals can gain control over quantities of capital which are large in absolute terms but small in comparison to the total size of the economy is a crucial feature of this system.
[1]: https://tradingeconomics.com/united-states/government-spendi...
First, and most importantly, this is a critique of all societal institutions. I do not think that the correct solution is necessarily taxation and redistribution. The investors in Stability AI didn't have to choose to misspend their money on an organization structured as a typical 20th century corporate dictatorship run by a big personality; they could have set up a much more effective structure. But they chose that corporate structure and leadership personality type/background because our society is sick on a fundamental level.
I would like to think financial markets are capable of breaking out of this insanity, but I'm extraordinarily skeptical.
Second, almost none of our alternative to the financial status quo of corporate-big-personality-dictatorship (aka government spending since we have no other alternative) goes to R&D/innovation. The vast bulk of that spending goes to entitlements and defense. The part we do spend on science is barely enough to support basic science and is still embarrassingly small compared to other outlays. For example, we spend twice as much on farm subsidies (not AG total, just subsidies) as we do on the National Science Foundation.
Third, of the part that does go toward R&D/innovation, almost none of it is spent on development and innovation. Almost all of our federal spending buckets on R&D is heavy on R, light on D, and "hackathon"/"pre-seed" levels of interest in innovation. We do not spend enough on R&D/innovation, and what we do spend is misallocated (mostly towards the goal of generating IP for Springer) because it is mostly allocated by folks at elite universities who are all friends with one another.
The public sector in the west is just as broken as the private, albiet the pathology presents differently.
I imagine there are alternatives beyond choosing between two horrendously broken systems.
Mutual cooperatives have all the pieces we need in my opinion for the core organization of human action. Unfortunately they are effectively impossible to start for anything with more than trivial startup costs, as capital is (somewhat understandably) allergic to promoting Democratic employee owned organizations.
Definitely need more of this. It's strange that, in the west, we're deeply committed to these basic values in all institutions except those which have become the most powerful and important in folks' every days lives (employers and service/goods providers), where we are instead deeply allergic to notions of democracy, equality, and checks and balances.
It’s absurd right? Democracy is great —- as long as it’s not in the workplace! That would be madness!
It’s a long hard road here FWIW so if you’ve got the stomach for it I suggest trying to build one.
This is exactly what I mean when I say that we are not at The End Of History. Western values need not necessarily win out. The belief in that inevitability is the source of the west's sclerotic apathy, and without correction will lead no where good.
4th time in days
Ren Ito - COO