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

281–290 of 429 posts

Re: Emad Mostaque resigned as CEO of Stability AI

#281

Post by whywhywhywhy »

Hendrikto wrote: post #272
achow wrote: post #266

Earlier quoted context omitted.

> 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. Recently, during an interview [1], when questioned about OpenAI's Sora, Shantanu Narayen (Adobe CEO) gave an interesting persp…

> Not only Adobe perhaps knows better than anyone else what is need and workflow of a moviemaker I only ever heard creatives complain about Adobe and their UI/UX and how they don’t understand their customers. Never really used any of their products myself though. Maybe they still are best-in-class. I can’t tell.

Many years ago it was a good offering but it’s becoming increasingly clear with outsourcing and talent drain that the current teams working on the likes of Photoshop, After Effects and Premier do not actually understand how the core tool, both in its inner workings or even how it draws its own UI works at all and couldn’t either recreate it or even change its existing behavior.

Every major change in the last 6 years has either been weird window dressing changes to welcome panels or new document panels, in all cases building sluggish jank heavy interfaces, try navigating to a folder in the premier one and weep as clicks take actual seconds to recognize.

Or just silly floating tooltips like the ones in Photoshop that also take a second to visible draw in.

All tangible tool changes exist outside the interface or you jump to a web interface in a window and back with the results being passed between in a way that makes it very obvious the developers are trying to avoid touching the core tools code.

Very clear Narayens outsourcing and not being a product guy has lead to this

Re: Emad Mostaque resigned as CEO of Stability AI

#282

Post by whywhywhywhy »

bamboozled wrote: post #273
Hendrikto wrote: post #272

Earlier quoted context omitted.

> Not only Adobe perhaps knows better than anyone else what is need and workflow of a moviemaker I only ever heard creatives complain about Adobe and their UI/UX and how they don’t understand their customers. Never really used any of their products myself though. Maybe they still are best-in-class. I can’t tell.

People love illustrator

It was good in the 00s but now it’s rickety and antiquated and the GPU acceleration was never implemented correctly.

Figma could build an illustrator killer in 6 months if they wanted to and it would be obliterated.

If they actually tackled this task people would be kicking themselves for putting up with the shambles that is illustrator for this long.

Re: Emad Mostaque resigned as CEO of Stability AI

#283

Post by Zetobal »

achow wrote: post #266
jacky2wong wrote: post #256

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 rectifyi…

> 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. Recently, during an interview [1], when questioned about OpenAI's Sora, Shantanu Narayen (Adobe CEO) gave an interesting persp…

People don't want to buy a third of an inch drill, they want a third of an inch hole.

Re: Emad Mostaque resigned as CEO of Stability AI

#284

Post by abi »

Der_Einzige wrote: post #203
stale2002 wrote: post #12

Well RIP Stability AI. I love their models and I love how they have changed the entire open source AI ecosystem for the better, but the writing was always on the wall for them given how unprofitable they are. I don't think much of the AI startup scene or socials groups like e/acc would have existed if it weren't for the tech that they just gave away for free. Its interesting how Stability AI and their VC funding have…

Huggingface are also the real actual openAI

How? What models have they put out that are relevant?

Re: Emad Mostaque resigned as CEO of Stability AI

#285

Post by nsagent »

mirekrusin wrote: post #165
ctrw wrote: post #32

Earlier quoted context omitted.

Think seti at home. Instead of wasting all the compute on bitcoin we pretrain fully open models which can run on people's hardware. A 120b ternary model is the most interesting thing in the world. No one can train one now because you need a billion dollar super computer.

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…

You're right that parameter updates typically require full state, but during my PhD I've explored some possibilities to address this limitation (unfortunately, my ideas didn't pan out in the time I had). That said, there is research that has explored this topic and made some progress, such as this paper:

https://arxiv.org/abs/2303.14177

Re: Emad Mostaque resigned as CEO of Stability AI

#286

Post by nsagent »

nabakin wrote: post #262
mejutoco wrote: post #226

Earlier quoted context omitted.

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 don't think MoE allows for that either. You'd have to come up with a whole new architecture that allows parts to be trained independently and still somehow be merged together in the end.

This paper addresses that issue and allows fully independent training of the experts:

https://arxiv.org/abs/2303.14177

Re: Emad Mostaque resigned as CEO of Stability AI

#287

Post by tkgally »

nsagent wrote: post #286
nabakin wrote: post #262

Earlier quoted context omitted.

I don't think MoE allows for that either. You'd have to come up with a whole new architecture that allows parts to be trained independently and still somehow be merged together in the end.

This paper addresses that issue and allows fully independent training of the experts: https://arxiv.org/abs/2303.14177

This one, from a couple of days ago, might address that issue as well:

https://arxiv.org/abs/2403.13187

Re: Emad Mostaque resigned as CEO of Stability AI

#288

Post by brrrrrm »

FeepingCreature wrote: post #249
ImprobableTruth wrote: post #72

Earlier quoted context omitted.

Transmission speeds aren't fast enough for this, unless you crank up the batch size ridiculously high.

LoRA training/merging basically is "crank up the batch size ridiculously high" in a nutshell, right? What actually breaks when you do that?

Cranking up the batch size kills convergence.

Re: Emad Mostaque resigned as CEO of Stability AI

#289

Post by kamikaz1k »

andy99 wrote: post #275
Hendrikto wrote: post #272

Earlier quoted context omitted.

> Not only Adobe perhaps knows better than anyone else what is need and workflow of a moviemaker I only ever heard creatives complain about Adobe and their UI/UX and how they don’t understand their customers. Never really used any of their products myself though. Maybe they still are best-in-class. I can’t tell.

I don't know Adobe's business so could be wrong, but maybe "creatives" are not their key customers? If they're focusing on enterprise sales, they're selling to enterprise decision makers. Every user hates using microsoft products, and don't get me started on SAP. But these are gigantic companies with wildly successful products aimed at enterprise customers.

If they’re selling to enterprise decision makers, aren’t they also B2B? In which case they have the same deficiency they started OpenAI has.

Re: Emad Mostaque resigned as CEO of Stability AI

#290

Post by FeepingCreature »

brrrrrm wrote: post #288
FeepingCreature wrote: post #249

Earlier quoted context omitted.

LoRA training/merging basically is "crank up the batch size ridiculously high" in a nutshell, right? What actually breaks when you do that?

Cranking up the batch size kills convergence.

Wonder if that can be avoided by modifying the training approach. Ideas offhand: group by topic, train a subset of weights per node; figure out which layers have the most divergence and reduce lr on those only.
Post reply on HN