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

Leaked deck raises questions over Stability AI’s Series A pitch to investors

sifted.eu

Post reply on HN
Print view View original

61–70 of 72 posts

Re: Leaked deck raises questions over Stability AI’s Series A pitch to investors

#61

Post by esafak »

Jaymz87 wrote: post #14
ShamelessC wrote: post #7

Earlier quoted context omitted.

> I don't remember Stability claiming in any public forum that they solely developed Stable Diffusion This article makes no such claim. It is about whether or not investors knew.

Investors probably shouldn't be chucking $100m+ at startups without doing the bare minimum of DD if that's the point of contention here...

You can't pin it on investors, when founders deliberately schedule fundraising rounds tightly not to give investors time to think. This benefits the founder, but if they pull a fast one, investors will lose trust and be less lenient towards the next founder. Trust is a pillar of the venture capital ecosystem.

Don't say buyer beware.

Re: Leaked deck raises questions over Stability AI’s Series A pitch to investors

#62

Post by capableweb »

emadm wrote: post #59
capableweb wrote: post #53

Earlier quoted context omitted.

As someone who develops software using Stable Diffusion and gets sold for money, I'm well aware how simple Dreamstudio is. No professional would use it. Doesn't mean they are not developing it to be a real product, that they'll try to get people to buy for real money.

The strategy was not DreamStudio but variant open models for private data. We will be open sourcing DreamStudio shortly having refactored it.

That's exciting to hear! I wish you all the best with the release, the community will surely appreciate it. Thanks for the work you do with Stability AI, there are quite a few of us that really appreciate it, no matter of these hit-pieces :)

Re: Leaked deck raises questions over Stability AI’s Series A pitch to investors

#63

Post by emadm »

Havoc wrote: post #33

>what official affiliation there was between Stability and EleutherAI, they said: “None.” That's wild. Handing out a bunch of cash for nothing except vague association. No control? No equity? No IP? Reminds me of marketing. Cola puts logos on t-shirts of players and proclaims to have sponsored them, but I have yet to see cola claim to have "co-scored" a goal.

This is dumb.

We employed Eleuther team members as Stability AI employees/contractors and incubated them until the 501(c)3 was set up and we managed to bring in other funders too: https://techcrunch.com/2023/03/02/stability-ai-hugging-face-...

I am on the board and delighted to continue to support their work as an independent organisation for LM evaluation, alignment and interpretability which is much needed.

Indeed though our approach was handing out significant compute for no control, no equity, no IP.

Anyone who has received Stability AI grants will be able to attest to this with multiple breakthroughs as a result, for example funding https://github.com/BlinkDL/RWKV-LM, the work of https://github.com/lucidrains and others.

Similarly we funded the beta of MidJourney with a cash grant for compute without ever even floating asking for equity etc as it is a market-creating innovation.

At the time MidJourney was using cc12m_1, a model developed by one of our lead (employed) generative AI developers Katherine Crownson / RiversHaveWings (https://github.com/crowsonkb/v-diffusion-pytorch)

Our model is simply to take open innovation and create commercial variants of that (our stable series models) from scratch and on our own, plus variants of that for private data - https://twitter.com/EMostaque/status/1649152422634221593?s=2...

This means we can be hands off versus other funders and trust researchers and help them succeed, something others do not.

Re: Leaked deck raises questions over Stability AI’s Series A pitch to investors

#65

Post by emadm »

rvz wrote: post #58
capableweb wrote: post #41

Earlier quoted context omitted.

> that investors were mislead by the claims made by Stability.ai's IP Which claims, precisely? Could you quote these claims that were made by Stability/Emad that makes it clear they were misleading the investors? Because the article doesn't make those clear, these claims you talk about.

> Which claims, precisely? Assuming you have read the article (and its links), there is a video podcast which he claimed that 'his team' originally developed Stable Diffusion under Stability.ai without mentioning the original researchers that created the model, until they raised money which one of the researchers clearly mentions this in the article: > “Once he had this money he became a bit clearer who actually deve…

Most of the original researchers of latent diffusion are employed at Stability AI as is the co-lead of Stable Diffusion.

This leaked deck was before stable diffusion release and the number refers to other open source models built by stability ai employees.

Our model is not SaaS and has never been.

Re: Leaked deck raises questions over Stability AI’s Series A pitch to investors

#67

Post by dragonwriter »

capableweb wrote: post #39
anonylizard wrote: post #36

Earlier quoted context omitted.

Stability never had any revenue, nor any easy ways to measure their user base. Their 'growth' is basically dependant on them being able to release open source models that get so popular, unbeatable ecosystems get formed around them. On this front. SD1.4/1.5 is extremely successful, but there has been no followup. SD2.0 was a total disaster that shocked the community, SD2.1 is still no better than SD1.5. StableLM curr…

> Stability never had any revenue, nor any easy ways to measure their user base They do, for the products they have/will have. Right now, I think it's only DreamStudio, but like any SaaS product, they can and surely do measure the amount of users, and trivial for them to see the revenue. What you're talking about, the models they help release, are not products themselves, more like research output they give away for…

> Right now, I think it’s only DreamStudio

Clipdrop, also, at least.

Re: Leaked deck raises questions over Stability AI’s Series A pitch to investors

#68

Post by mountainriver »

emadm wrote: post #57
anonylizard wrote: post #36

Earlier quoted context omitted.

Stability never had any revenue, nor any easy ways to measure their user base. Their 'growth' is basically dependant on them being able to release open source models that get so popular, unbeatable ecosystems get formed around them. On this front. SD1.4/1.5 is extremely successful, but there has been no followup. SD2.0 was a total disaster that shocked the community, SD2.1 is still no better than SD1.5. StableLM curr…

Stability AI is 18 months old so far and has spent 100x less than OpenAI which was founded in 2015. I think it unreasonable to compare at this stage, we are still early as a company and organisation. We have revenue (to be announced) but release models differently as we are not proprietary. The numbers in that deck were as reported on models developed by team members (we chose not to assert IP rights) in use by vario…

That’s a fair point, and what Stability has given the ecosystem has been amazing, I’m rooting for them as they are one of the only orgs left that is still committed to open source

Re: Leaked deck raises questions over Stability AI’s Series A pitch to investors

#69

Post by ad404b8a372f2b9 »

emadm wrote: post #54
ad404b8a372f2b9 wrote: post #40

Earlier quoted context omitted.

Some of their behaviour over the past year makes me feel the company has a culture which doesn't hold honesty as a core value. The Runway takedown or something your refer to was Stability sending a takedown request to HuggingFace claiming an IP leak over the publication of version 1.5 by their academic partners Runway. That was sorted out, if you want to call it that, after massive public outcry and Emad claiming it…

I normally don't like engaging in these things but think should step in given increasing fud. Stability AI lawyers never contacted HuggingFace requesting an IP takedown. You may ask HuggingFace themselves. RunwayML agreed not to release 1.5 for safety reasons in writing and then chose to anyone presumably leading to confusion by our tech team who let HF know it couldn't be them. You can ask them why they did that, Ru…

Hugging Face themselves did say the takedown request was not a formal legal request from the get go:

"Company StabilityAI has requested a takedown of this published model characterizing it as a leak of their IP

While we are awaiting for a formal legal request, and even though Hugging Face is not knowledgeable of the IP agreements (if any) between this repo owner (RunwayML) and StabilityAI, we are flagging this repository as having potential/disputed IP rights. [...]"

They didn't hint at an account hijacking by a third party if that's what your team intended to convey to them however.

To add to that, your CIO addressed the community at the time and characterized RunwayML in the following terms: "[...] We also won't stand by quietly when other groups leak the model in order to draw some quick press to themselves while trying to wash their hands of responsibility."

Post reply on HN