Hacker News Board Hacker News, browsed like a classic message board
Live data from Hacker News

Emad Mostaque resigned as CEO of Stability AI

stability.ai

Post reply on HN
Print view View original

251–260 of 429 posts

Re: Emad Mostaque resigned as CEO of Stability AI

#251

Post by woadwarrior01 »

renewiltord wrote: post #193
llm_trw wrote: post #138

Earlier quoted context omitted.

The `big' AI models are trillion parameter models. The medium sized models like GPT3 and Grok are 185b and 314b respectively. There is no way for _anyone_ to run these on a sub $50k machine in 2024, and even if you can the token generation speed on CPU is under 0.1 tokens per second.

Mixtral 8x7b is better than both of those and runs on a top spec M3 Max wonderfully.

Indeed! Also, Mixtral 8x7b runs just as well on older M1 Max and M2 Max Macs, since LLM inference is memory bandwidth bound and memory bandwidth hasn't significantly changed between M1 and M3.

Re: Emad Mostaque resigned as CEO of Stability AI

#252

Post by l33tman »

mejutoco wrote: post #226
mirekrusin wrote: post #165

Earlier quoted context omitted.

SETI made sense because there is a lot of data where you download chunk and do expensive computation and return thin result. Model training is unlike that. It's large state that is constantly updated and updates require full, up to date state. This means you cannot distribute it efficiently over slow network with many smaller workers. That's why NVIDIA is providing scalable clusters with specialized connectivity so t…

That is the challenging part indeed. But if we think of mixture of experts models outperforming "monolithic" models, why not? Maybe instead of 8 you can do 1000 and that is easy to paralellize. It sounds worth exploring to me.

I think the MoE models are trained together just like any other network though, including the dispatcher layer that has to learn which "expert" route each token to. Perhaps you could do some kind of technically worse model architecture that is trained separately and then a more complex dispatcher that then learns to utilize the individually trained experts as best as it can?

Re: Emad Mostaque resigned as CEO of Stability AI

#253

Post by LunaSea »

throw23324 wrote: post #205
ShamelessC wrote: post #151

Earlier quoted context omitted.

[flagged]

It's true. Latin America already knows it, USA and Europe are yet to catch up. Most of crypto is bullshit, but at the very least Bitcoin is a massive, massive use case.

What can I currently buy with Bitcoin?

Re: Emad Mostaque resigned as CEO of Stability AI

#254

Post by Loughla »

vasco wrote: post #235
HlessClaudesman wrote: post #197

Earlier quoted context omitted.

Heres a Stable Diffusion buisness idea: sign up all the celebrities and artists who are cool with AI, and provide end users / fans with an AI image generation interface, trained on their exclusive likenesses / artwork (loras). You know, the old tried and true licensed merchandise model. Everybody gets paid.

> sign up all the celebrities and artists who are cool with AI "Cool with AI" and "sell my likeness so nobody ever needs to hire me again" are too close for comfort on this one.

And also, here's a way to just make so much pornography of me.

I'm betting the list of folks who would sign the AI license are pretty small, and mostly irrelevant.

Re: Emad Mostaque resigned as CEO of Stability AI

#255

Post by ben_w »

numpad0 wrote: post #223
HlessClaudesman wrote: post #197

Earlier quoted context omitted.

Heres a Stable Diffusion buisness idea: sign up all the celebrities and artists who are cool with AI, and provide end users / fans with an AI image generation interface, trained on their exclusive likenesses / artwork (loras). You know, the old tried and true licensed merchandise model. Everybody gets paid.

I think the following isn't said often enough: there must be a reason why there are extremely few celebrities and artists who are cool with AI, and it cannot be something abstract and bureaucratic as copyright concerns although those are problematic. It's just not there yet. GenAI outputs aren't something audiences wants to hang on a wall. It's something that evoke sense of distress. Otherwise everyone's tracing them…

Most people mix up all the different kinds of intellectual property basically all the time[0], so while people say it's about copyright, I (currently) think it's more likely to be a mixture of "moral rights" (the right to be named as the creator of a work) and trademarks (registered or otherwise), and in the case of celebrities, "personality rights": https://en.wikipedia.org/wiki/Personality_rights

> It's just not there yet. GenAI outputs aren't something audiences wants to hang on a wall.

People have a wide range of standards. Last summer I attended the We Are Developers event in Berlin, and there were huge posters that I could easily tell were from AI due to the eyes not matching; more recently, I've used (a better version) to convert a photo of a friend's dog into a renaissance oil painting, and it was beyond my skill to find the flaws with it… yet my friend noticed instantly.

Also, even with "real art", Der Kuss (by Klimt) is widely regarded as being good art, beautiful, romantic, etc. — yet to me, the man looks like he has a broken neck, while the woman looks like she's been decapitated at the shoulder then had her head rotated 90° and reattached via her ear.

[0] This is also why people look at a Google street view image with a ©2017 Google[1] tiled over on a blue sky and say "LOL, Google's trying to own the sky", or why people even on this very forum ask how some new company can trademark a descriptive term like "GPT"[2], seemingly surprised by this being possible even though there's already a very convenient example of e.g. Hasbro already having "Transformers".

[1] https://www.google.com/maps/@33.7319434,10.8655264,3a,77.2y,...

[2] https://news.ycombinator.com/item?id=35692476

Re: Emad Mostaque resigned as CEO of Stability AI

#256

Post by jacky2wong »

I think Stability is in an interesting situation. A few suggestions on its direction and current state: 1. Stability AI's loss of talent at the foundational research layer is worrying. They've lost an incredibly expensive moat and there's enough unsolved problems in the foundation layer (faster models, more energy efficient models, etc.) to ensure Stability provides differentiated offerings. Step 1 should be rectifying the core issues of employment and refocusing this more into the AI lab space. I have no doubt this will require a re-steering of the ship and re-focusing of the "mission". 2. Stability AI's "mission" of building models for every modality everywhere has caused the company to lose focus. Resources are spread thin. With $100M in funding, there should be a pointed focus in certain areas - such as imaging or video. Midjourney has shown there is sufficient value capture already in just 1 modality. E.g. StableLM seems like early revenue rush and a bad bet with poor differentiation. 3. There is sufficient competition on the API layer. Stability's commitment to being open-source will continue to entice researchers and developers but there should be a re-focus on improvements in the applied layer. Deep UX wrappers for image editing and video editing while owning the end to end stack for image generation or video generation would be a great focal point for Stability that separates itself from the competition. People don't pay for images, they pay for images that solves their problems.

Re: Emad Mostaque resigned as CEO of Stability AI

#257

Post by baobabKoodaa »

yreg wrote: post #40
echelon wrote: post #34

Earlier quoted context omitted.

No. There can be tremendous value in AI, you just won't find it here. The reason is that Stability.ai gave away everything for free . Until recently, they didn't even attempt to charge money for their models. I've heard the only reason they're not already closed up is that they're reselling all of the rented GPU quota they leased out years ago. Companies are subletting from Stability, which locked in lots of long ter…

> the only reason they're not already closed up is that they're reselling all of the rented GPU quota They have raised 110M in October and they say that training a particular model costs them hundreds of thousands $ in compute costs. They don't lack money.

They needed to take an emergency $50M loan after that, so it clearly didn't last long

Re: Emad Mostaque resigned as CEO of Stability AI

#258

Post by alfiedotwtf »

hiddencost wrote: post #185
alfiedotwtf wrote: post #173

Earlier quoted context omitted.

So you’re happy to continue forever with the system we have now - governments printing with no consequence?

No, but I'm not going to invest in a bunch of grifters and con artists holding up a technically bad solution because they want money and power.

Commercial banks make up most central banks, and if you believe commercial banks aren't grifters or con artists, you've obviously never heard of LIBOR for starters.

As for bad solutions, I'll give you incorrect economic forecasts on inflation and using interest rates as a lever.

Re: Emad Mostaque resigned as CEO of Stability AI

#259

Post by d-z-m »

rvz wrote: post #10

Looks like many AI startups are experiencing a bit of some turbulence and chaos in the recent months: First the OpenAI rebellion in November, then the Inflection AI acqui-hire from Microsoft not willing to pay the over-valued $4B and deflected that to $600M instead (after making $0 revenue) and now a bit of in-stability at Stability AI with the CEO resigning after many employees leaving. What does that say about the…

> OpenAI rebellion

This phrasing makes me feel like we're living in some techno-future where corporations are de-facto governance apparatuses, squashing rebellion and dissidents :^)

Re: Emad Mostaque resigned as CEO of Stability AI

#260

Post by baobabKoodaa »

Eisenstein wrote: post #204
throwaway11460 wrote: post #133

Earlier quoted context omitted.

Perhaps the hundreds of millions in VC investment and company value?

I'm not sure I understand. Are you saying that if a company gets a valuation because of an investment, that creates value?

They are saying it was not clear "buying branding rights to an image model" would lead to any investments, any kind of high valuation, or any other financial success. It is only clear in hindsight.
Post reply on HN