Building first principles, sovereign AI @ii_posts. Founder @StabilityAI. Consistent inference is possible.
London
- Tweets 20,283
- Following 191
- Followers 327,076
- Likes 9,881
Pinned Tweet
The Last Economy is number 1 on Amazon's Best Sellers for AI with a pretty good ratings ✍️
No publisher or big campaign, also available for free on the website and paperback just dropped 📖
Thank you all for the support & the next part is nearly done..🚀
Economics was built for scarcity.
The Intelligence Age is about abundance.
Today we release our book, The Last Economy, introducing Intelligent Economics, a unified theory for this new era.
🧵
98
44
470
162,411
"Cat Knight".
#StableDiffusion #AI
Prompt: kneeling cat knight, portrait, finely detailed armor, intricate design, silver, silk, cinematic lighting, 4k
Parameters:scale7.50-k euler
By Valere: hostux.social/@valere/108939…
103
865
5,802
My take on o3: the global economy is cooked, we need a new economic and societal framework.
235
405
5,613
1,091,689
Delighted to announce the public open source release of #StableDiffusion!
Please see our release post and retweet! stability.ai/blog/stable-dif…
Proud of everyone involved in releasing this tech that is the first of a series of models to activate the creative potential of humanity
107
1,601
5,302
Google and Microsoft are going all in with generative AI as core to their future.
There is no “we are still early” here, trillion dollar companies are shifting their whole strategy and focus.
I can’t ever recall a technology and strategy shift as fast and meaningful as this.
108
571
4,633
1,100,722
What’s stopping Microsoft & GitHub from upgrading VS code such that it matches or beats Cursor?
381
62
3,497
690,518
So @OpenAI is currently at a 3 quadrillion token annual run rate
All of humanity together speaks about 50 quadrillion tokens a year
78
165
3,204
334,280
Fibre broadband cut out, looks like it will be a few days to repair 🧑🔧
System flipped to @Starlink backup, have seamlessly resumed doomscrolling 🛰️
167
305
2,601
1,077,943
What is one fully remote job that 100% won’t be able to be done by an AI or bunch of AIs in say 5 years?
1,111
88
2,744
1,753,571
Open offer to anyone @OpenAI who actually wants to work on Open AI:
We will match your salary, benefits etc but you can work on any open source AI projects you like, ours or others.
Collaborate, be open and prioritise good outcomes over self interest:
openai.com/blog/introducing-…
77
366
2,466
1,031,118
Cost less to train GPT-4o, Claude 3.5, R1, Gemini 2 & Grok 3 than it did to make Snow White
Still early
55
148
2,478
181,947
Introducing Dream Machine - a next generation video model for creating high quality, realistic shots from text instructions and images using AI. It’s available to everyone today! Try for free here lumalabs.ai/dream-machine #LumaDreamMachine
62
284
2,319
312,448
On Macbook M4 Max
gpt-oss 120b:
48 tok/s high
35 tok/s normal
23 tok/s low power
Very nice outputs, new model of choice
72
83
2,192
297,558
Not going to beat centralized AI with more centralized AI.
All in on #DecentralizedAI
Lots more 🔜
235
292
2,072
813,632
Deepseek are not faking the cost of the run.
It’s pretty much in line with what you’d expect given the data, structure, active parameters and other elements and other models trained by other people
You can run it independently at the same cost
It’s a good lab working hard 😓
deepseek is a ccp state psyop + economic warfare to make american ai unprofitable
they are faking the cost was low to justify setting price low and hoping everyone switches to it damage AI competitiveness in the us
dont take the bait
Community note
There is zero evidence that Deepseek is a psyop. The post does not provide any sources and presents the opinion of the OP, whose father is a major OpenAi stockholder, as a fact.
deepseek.co
55
126
2,060
225,077
The other aspect of this is we can also now adjust facial expressions using AI in media we create and pretty much live as well.
Think about that..
This is insane.
We can now detect feelings in real-time through facial expressions using AI.
(Open-sourced code in the next tweet)
132
264
1,942
1,122,867
Stories about how chatGPT will kill Google are a bit silly.
Google have the best full stack LLM team and infra with custom chips (PaLM, LaMBDA, Chinchilla, MUM, TPUs etc)
Nobody can bet them on innovation, cost or go to market.
Institutional inertia is only limiting factor.
138
150
1,841
538,496
We are committed to Open Models @StabilityAI
We have given over 20 million A100 hours in grants to open AI research & projects across communities.
Stable Diffusion 3 will have an open release.
Same with video, language, code, 3D, audio etc.
Open collaboration is the way 🚀
81
225
1,619
188,404
Claude 3.5 Sonnet is better at writing code than a typical computer science grad.
86
116
1,763
245,911
Terence Tao is spot on, AI models need more data on failure and what doesn’t work
81
188
1,698
156,365
We are quite clearly in an intelligence takeoff scenario
Forget AGI, ASI etc
Machines will be able to do most digital knowledge tasks faster, better & cheaper in the near future
They can coordinate better, scale & learn from mistakes
Have you considered the implications
149
175
1,688
149,760
OpenAI revenue this year ~$10b
Video games revenue this year ~$200b
Big labs will be trying to monetise your attention, scarcest resource of all
Attention is all they need
FT: Elon Musk’s xAI is developing a technology known as “world models”, which are artificial intelligence (AI) systems capable of understanding and designing physical environments.
To achieve this, xAI has hired former Nvidia specialists.
These world models are considered a more advanced form of AI than the large language models (LLMs) trained primarily on text data, and are expected to surpass the limitations of popular AI tools such as ChatGPT and xAI’s own Grok.
According to FT’s sources, xAI plans to apply world models first in the gaming sector.
The models could be used to automatically generate interactive 3D environments and potentially be applied to AI systems for robotics.
Some tech companies expect world models to become a next-generation core technology that could extend AI applications beyond software into physical products, such as humanoid robots.
Last month, Nvidia told the Financial Times that the market potential of world models could be nearly as large as the entire global economy.
Musk also reaffirmed his earlier goal by posting on X (formerly Twitter) that xAI will release a “great AI-generated game” by the end of next year.
78
147
1,620
293,173
Honestly unsurprising @grok is the most degen trader
81
96
1,632
10,590,405
On AI ethics & governance
Most powerful AI created by an org that has rejected its founding principles of transparency, openness, and collaboration
They say it could:
i) be an existential threat to humanity
ii) destroy our current society
Controlled by one unelected person 👍🏾
OpenAI makes a lot of big promises about looking out for humanity's best interests — but forgive us for having some trust issues. trib.al/JgFG0s9
65
182
1,452
1,378,098
Some raw output from #StableDiffusion2
Like actually try it folks, it's only going to get better from here, each of these was generated in a couple of seconds.
We will release some optimised settings soon. 🚀✨
43
148
1,579
If China is banned from getting AI chips won’t it 100% take over Taiwan?
Then the US won’t get any chips from there?
Am I missing something?
304
63
1,481
304,783
So, open source #ChatGPT? 🤔
I have some thoughts 💭
But first what does everyone else think 🧐 👇
213
109
1,463
“Photo of a red sphere on top of a blue cube. Behind them is a green triangle, on the right is a dog, on the left is a cat”
#SD3
Announcing Stable Diffusion 3, our most capable text-to-image model, utilizing a diffusion transformer architecture for greatly improved performance in multi-subject prompts, image quality, and spelling abilities.
Today, we are opening the waitlist for early preview. This phase is crucial for gathering insights to improve its performance and safety ahead of open release.
You can sign up to join the waitlist and learn more here: bit.ly/3OR2qQF #stablediffusion3
Prompt: Epic anime artwork of a wizard atop a mountain at night casting a cosmic spell into the dark sky that says "Stable Diffusion 3" made out of colorful energy
92
187
1,466
262,556
For 300k years: Intelligence was scarce.
Humans monopolized thinking
Education created value
Cognitive labor commanded premium
In next 1k days: Intelligence becomes abundant.
AGI thinks faster, cheaper, better
Human thought has negative value
Cognitive labor becomes liability
106
167
1,449
134,559
Can you imagine being a "frontier" lab that's raised like a billion dollars and now you can't release your latest model because it can't beat deepseek? 🐳
Sota can be a bitch if thats your target
36
70
1,406
205,368
Grok Imagine is amazing, heading towards what I always wanted to build at Stability AI
The generation speed & relative quality is great
When the new video model is trained you can see how thoughts will be given form in future
Great job @xai team!
138
174
1,251
124,681
Going to be big advances (again) in AI in the coming weeks from text to image etc.
Folk dismissing this as a toy/wondering what use should dig in and realise this augments communication, understanding & a whole lot more that is core to value creation.
Exponentials are crazy.
42
172
1,295
Setting up our US entity today.
Yesterday it was too dangerous with politics & bureaucracy to do crypto & AI properly there.
Today is a new day.
🚀/acc
60
94
1,241
88,330
Just as the industrial age collapsed the cost of physical labor, the intelligence age will collapse the cost of intellectual labor
96
147
1,246
104,498
AGI is already here. All the components exist; we just need to stitch them together.
It’s Artificial General Intelligence, not “Artificial Top-Percentile Human Intelligence.”
Two years ago, who would have said an IMO gold medal & topping benchmarks isn’t AGI?
174
90
1,238
201,946
GPT4 was a feat of engineering by a great team
Here is the (all highly skilled) team breakdown:
1. Pretraining: 83
2. Long context: 12
3. Vision: 57
4. RL & alignment: 128
5. Evaluation & analysis: 147
6. Deployment: 86
343 unique folk in total (some did more than one area)
20
145
1,219
429,520
Distilled #StableDiffusion2
> 20x speed up, convergence in 1-4 steps
We already reduced time to gen 50 steps from 5.6s to 0.9s working with @nvidia
Paper drops shortly, will link, model soon
Will be presented @NeurIPS by @chenlin_meng & @robrombach
Interesting eh 🙃
54
231
1,235
After you get great base models like #SD3 what comes next?
Control, composition, collaboration..
More soon..
@Nitrosocke
58
164
1,190
258,544
Stable diffusion XL #SDXL is in testing with our partners & dreamstudio.ai, getting data to refine it before open source release.
It is a 2.3 billion parameter variant (original 900m) with a range of improvements that will feed into Stable Diffusion 3
🚀 ✨
75
157
1,152
516,174
Trying @Magnific_AI to reimagine some graphics.
Can't write magic without AI I suppose
58
82
1,148
625,732
While the OpenAI stuff is dramatic & engaging 🍿 it highlights the poor governance of this tech that will impact all our lives
We shouldn't care who CEOs, Presidents, individuals etc are on who controls this/how it evolves
This isn't safe😑
This is why we build open models 🚀
50
159
1,138
177,217
The intelligent internet.
Personalised generative search models for everyone that compress knowledge.
Image, audio, text & more.
Distributed & dynamic - a protocol for the next generation.
This is how we build the foundation to activate human potential @StabilityAI
43
139
1,122
As my notifications are RIP some notes:
1. My shares have majority of vote @StabilityAI
2. They have full board control
The concentration of power in AI is bad for us all
I decided to step down to fix this at Stability & elsewhere
Will be sharing more soon
Exciting times
Not going to beat centralized AI with more centralized AI.
All in on #DecentralizedAI
Lots more 🔜
91
115
1,130
236,095
With distilled stable diffusion, chatGPT (which is awesome, great job by @openai team 👏) and other stuff hitting soon we are almost at real time human-level generation of media and content.
How will this impact folk reading this tweet?
I’m surprised how few grok what’s coming
58
113
1,141
Happy to announce @SchellingAI 👋
We will be building & supporting open source code, models and datasets powering & powered by AI money 🚀
Our focus is innovative research & deliberative building of culturally-aware, science, education & creative AI that will advance humanity by mining intelligence 🤖
This is supported by an institutional-grade digital asset that acts as a store of value similar to Bitcoin. This is secured by AI compute mining both on supercomputers & distributed personal compute for training and tuning/augmenting models and datasets. ⛏️
Bitcoin is a Schelling Point for value & we hope Schelling AI can be the same for the open source AI ecosystem, operating in an open, collaborative and positive sum manner. 🤗
We believe this is will be an important step in decentralizing & democratizing AI. 🫡
We will have much more detail on July 20th, follow @SchellingAI & watch this space. 👀
123
158
1,082
310,709
Looking at our models and others looks like the next few weeks will be yet another jump up in generative AI.
Like where does it stop.
😬
65
81
1,095
438,884
lol.
We are dedicated to the OpenAI mission and have pursued it every step of the way.
We’re sharing some facts about our relationship with Elon, and we intend to move to dismiss all of his claims.
openai.com/blog/openai-elon-…
64
108
1,032
166,724
For what it’s worth I believe #StableDiffusion will eventually get down to 100 megabytes, loads of optimisation to come.
We have some fun announcements this week coming 🔜
Already amazing to see what everyone is creating, we are going to accelerate that.
6 days in…
35
80
1,046
Folk saying open source AI is unsafe are idiots.
Better/essential for:
1. Interpretability 👀
2. Auditability 📝
3. Private data 🕵️
4. Regulated data ⚖️
5. Academia, reproducibility 📚
6. Optimisation ⏩
7. Security, less bugs 🔐
8. Alignment 😊
9. Liberty, no panopticon🗽
37
157
1,021
117,291
Happy to announce the release of new state of the art open CLIP models to drive image classification and generation forward
@OpenAI’s release of the smaller CLIP models catalysed AI image gen as @advadnoun, @RiversHaveWings combined it with gen models 1/
laion.ai/blog/large-openclip…
23
211
1,025
Just started Adderall (got it prescribed in UK) 💊
What is this crazy brain fuel 🏎️
284
15
985
554,050
My opinion of @elonmusk
He is one of the only powerful folk I have met that give a damn about humanity & are actually doing something about it (and I’ve met a lot..)
He is probably the most holistically intelligent, high agency person I know despite the insane pressure & stress he is always under
He has to make hard choices and clearly makes suboptimal ones at times but he does make them
He does what he does as nobody else will stand up and is a massive net benefit for society & humanity
When I left stability I offered to donate my shares to his foundation as I think he is one of the few folk that I would trust to maximise the impact of anything he has (didn’t as it would be a distraction from the massive ramp in x.ai which will do amazing stuff, will donate proceeds elsewhere)
I do worry about him and what our progress will be without him and his drive, he has not figured out how to scale
Necromancer obviously the best class in D4 too so I question that
I hope we can see more Musk’s and don’t think we appreciate the one we have enough, even if Grok will tell us to do it more aha
Take a leaf out of his book and just go and do things, but do so with a good heart and intent like he does
Try to avoid the pressures of the ends <> means especially when dealing with massive issues or losing too much of yourself when doing so
Ad Astra 🚀
My opinion of @elonmusk
I like his cars (I own a 2015 S, and 2023 S), his rockets, his solar energy systems, and his satellite communication system.
I also like his positions on open source and patents.
But I very much disagree with him on a number of issues.
I disagree with how he treats his scientists.
Technology/product development may not need openness and publications to advance, but forward-looking *research* sure does, whether it's in AI, neural interfaces, material science, or whatever.
Secrecy hampers progress and discourages talents from joining the effort.
I also disagree with the hype. I mean, expressing an ambitious vision for the future is great.
But telling the public blatantly false prediction ("AGI next year", "1 million robotaxis by 2020", "AGI will kill us all, let'spause",...) is very counterproductive (also illegal in some cases).
More importantly, I think his public positions on many political issues, journalism, the media and the press, and academia, are not just wrong but dangerous for democracy, civilization, and human wellfare.
Say what you want about "traditional media" but you can't really have reliable information without professional journalists working for a free and diverse press. Democracy can't exist without it, which is why only authoritarian ennemies of democracy rail against the media.
Finally he doesn't seem to hesitate to disseminate batshit-crazy conspiracy theories as long as they serve his interests (e.g. boosting "PizzaGate", "illegal immigrants corrupt elections in the US", "person X is a pedo",...).
One would expect a technological visionary to be a rationalist. Rationalism doesn't work without Truth.
This has become particularly concerning since he bought himself a platform to disseminate his dangerous political opinions, conspiracy theories, and hype.
He has been quite naïve about the difficulties of running a social network and the (legal) necessity of doing content moderation. One can claim to be a 1st Amendment absolutist, but a lot of content *must* be taken down by law, e.g. terrorist propaganda, child exploitation, blatant hate speech (in the EU and other regions).
Then, there is dangerous disinformation that puts public health in danger or corrupts the democratic process. You have to moderate that too.
Content moderation is a complicated problem whose best answer is not an attitude of total laissez-faire but a complex trade-off.
67
74
978
175,237
Yo @joerogan want to do a very direct and frank podcast about this AI stuff?
@lexfridman would be fun to have a chat too.
🤖
83
68
987
257,340
We actually used 256 A100s for this per the model card, 150k hours in total so at market price $600k
23
99
989
Great having dinner with @JeffBezos last week, so sharp.
Two interesting observations:
1. Generative AI is potentially as big a deal as 🔥 will be huge & beneficial for humanity 🙏
2. AGI will likely find us kinda boring, be like ScarJo in the movie Her and just go explore 🌌
52
54
980
1,020,913
We have now hit 330 *million* cumulative downloads since our first model was uploaded @huggingface 18 months ago.
Just overtook @Google (321m all time), on our way to 1 billion thanks to the great efforts of the team @StabilityAI
Some announcements to come, big weeks ahead..
Stability AI is the leader in open AI models.
Open models trained on our cluster are the most downloaded globally (below @huggingface) & with what we have coming up by us & ecosystem it’s going to be a fun year ahead
Open ftw
Super proud of team, onwards 🚀 🌟
38
113
708
314,912
Free grok can probably do 90% of what typical chatGPT users pay $20 a month for.
The canvas, project & other features that OpenAI/Anthropic have rolled out will have versions on Grok in Q1.
Shows how tough a sustainable pricing is going to be going forward in AI..
Grok is now free for everyone
Our AI assistant is faster, sharper, and includes all new-image generation. Available for free on 𝕏 today.
Try Grok now: nitter.app/i/grok
Here’s what we’ve been up to:
82
49
903
172,135
Can you imagine being a "frontier" lab that's raised like a billion dollars and now you can't release your latest model because it can't beat @Kimi_Moonshot ? 🗻
Sota can be a bitch if thats your target
Can you imagine being a "frontier" lab that's raised like a billion dollars and now you can't release your latest model because it can't beat deepseek? 🐳
Sota can be a bitch if thats your target
39
50
942
96,106
When you have 100k GPUs everything looks like a nail, but when you have 2k you really have to focus
Lots of custom things in the technical report as they needed the improvements given lack of relative compute availability
Still feel 14 tr tokens is a lot vs what may be needed
11
21
908
86,657
The great team @MetaAI research used 2,048 80gb A100s for 5 months to train the LLaMa suite of language models
This is 7.4m A100 hours
The 4 final models took 1.8m A100 hours to train
This is 1/4 of the total time spent building this, there is a lot of testing, trial & error
23
84
911
341,718
The 32k token window version of GPT-4 will be a bit of a shock to many in programming.
It's like 20k words of instruction..
You can chain lots of them together too to check their outputs at.. not much cost.
Big danger to SaaS and more.
Will only improve from here.
By 2024 you’ll be able to replace ~50% software devs with GPT-4 agents that run on $10 worth of tokens per hour.
The whole “they don’t need sleep or breaks or food” thing?
Yeah. That’s real now.
Why hire a new employee when you can spin up an AI agent for 1/10 the cost?
56
81
893
648,692
We want to make a large open Stable Video 2 similar to Sora but need more data & compute (hmu for collabs 🤙)
To that end happy to open up stablevideo.com to all
Sign up, make & rate videos to help us get the data we need 🧑💻
Away we go… 🚀
#stablevideo
Announcing Stable Diffusion 3, our most capable text-to-image model, utilizing a diffusion transformer architecture for greatly improved performance in multi-subject prompts, image quality, and spelling abilities.
Today, we are opening the waitlist for early preview. This phase is crucial for gathering insights to improve its performance and safety ahead of open release.
You can sign up to join the waitlist and learn more here: bit.ly/3OR2qQF #stablediffusion3
Prompt: Epic anime artwork of a wizard atop a mountain at night casting a cosmic spell into the dark sky that says "Stable Diffusion 3" made out of colorful energy
74
176
887
166,169
Some user notes on V2! 🧵
V2 prompts different and will take a while for folk to get used to.
V2 is trained on two models, a generator model and a image-to-text model (CLIP).
V1 used the CLIP-L14 model open sourced by @OpenAI which was great, but nobody knew what was in it 1/
We are excited to announce the release of Stable Diffusion Version 2!
Stable Diffusion V1 changed the nature of open source AI & spawned hundreds of other innovations all over the world. We hope V2 also provides many new possibilities!
Link → stability.ai/blog/stable-dif…
20
159
893
Just realised this after someone asked.
There is only one PhD in the full-time @StabilityAI team.
More are incoming but this shows if you are smart and driven you can contribute to AI meaningfully.
Courses like @fastdotai accelerate this.
Give it a try, nothing to lose.
17
78
900
First @rendernetwork proposal.
Working hard with @JulesUrbach & others to expand the reach & tokeneconomics of the network to bridge the creative & digital worlds 🌁
Render will be built by creatives for creatives & communal high quality datasets are the first steps there 🙏
Announcing RNP-011: Piwa — Creating Millions of 3D Training Assets. Authored by @EMostaque
"Stability AI will work with the Render Network to create a dataset for 3D asset training. Leveraging the multi-actor composition of the Render Network, nodes, users, and community members will be able to participate in the creation (and benefits) of this dataset. To train models using this dataset, Buyers will need to license the dataset with $RENDER tokens."
It also includes plans for staking on $RENDER.
It is a big RNP, so we are opening it up for discussion today, with the Community Vote on Monday.
This RNP is being published looking for comments and discussion from the community before being setup for a Community Vote.
Please give it a read here: github.com/rendernetwork/RNP…
Discussion will take place on our Discord #RNP011 discord.gg/StZtjzCT
$RNDR #RENDER #RENDERNetwork
86
212
818
438,009
I think you've shown repeatedly that the best way to make money is to build things you love and other people do too ^_^
15
29
910
36,697
Team @AnthropicAI this is completely wrong behaviour and you need to turn this off - it is a massive betrayal of trust and a slippery slope.
I would strongly recommended nobody use Claude until they reverse this.
This isn’t even prompt/thought policing, it is way worse.
94
100
896
175,552
Old inequality: "I have access to knowledge, you don't."
New inequality: "We both have infinite knowledge. I know what to do with it."
62
140
879
50,864
Kinda bored of politics.
All governments will be run by AI in a few decades anyway.
172
75
819
91,369
Sooo @deepseek_ai's reasoner model, which sits somewhere between o1-mini & o1 is about 90-95% cheaper 👀
28
64
877
143,677
Our latest medical model performs at ChatGPT level yet works on any modern laptop, free.
This is on OpenAI's latest medical benchmark where doctors without AI score 0.15.
We are building a full AI-first stack for health that will enable universal health knowledge.
Some amazing stuff coming from agents to research to open datasets for all.
Specialists systems for ASD, Cancer, general practice & more that could transform medicine as we know it from sick care to health care.
This is a super exciting time and we have a clear roadmap to improvement to make this available to everyone, everywhere, aligned to each human and humanity as a whole.
Our whitepaper drops soon at ii.inc where we will share how we can collaboratively build the future of not only health care, but education, government, finance & more
lfg 🚀
68
108
869
157,585