Igor Babuschkin
@ibab
Followers
67K
Following
4K
Media
44
Statuses
739
Maybe the real AGI was the friends we made along the way. Customer support @xAI
Palo Alto, CA
Joined February 2020
Example of why @xAI is a truly special place to work at: Recently Elon took the time to hear 5min presentations from everyone at the company about what they are working on, with everyone going into as much technical detail as possible. He gave feedback to every single team member.
771
2K
25K
We’ve shipped these highly requested features into @grok:.- Long context.- Image understanding.- Web search results.- Inline citations to sources.- Grok Analysis (summarize conversations and understand posts).- PDF upload.What else should we add to make Grok more useful?.
2K
433
7K
First he contaminates the web with woke nonsense, then he complains that others are struggling to get rid of the woke nonsense! Filtering out biased ChatGPT responses from our training data has been a struggle. We have confirmed that this is the primary reason for why Grok still.
296
606
6K
Post your daily use cases for Grok here and we will improve them.
real question: does anyone actually use Grok?.
1K
435
4K
Join us if you want to Make America Grok Again.
We are hiring forward deployed AI engineers to work with the coolest companies on Earth. At @xai , engineers review every technical application.
127
392
3K
Apply to @xAI at if you want to work with the largest and most powerful GPU cluster ever built.
@AravSrinivas Given the pace of technology improvement, it’s not worth sinking 1GW of power into H100s. The @xAI 100k H100 liquid-cooled training cluster will be online in a few months. Next big step would probably be ~300k B200s with CX8 networking next summer.
318
770
3K
Twitter really needs to add the ability to write math formulas in tweets with LaTeX syntax.
257
197
3K
@JaxWinterbourne The issue here is that the web is full of ChatGPT outputs, so we accidentally picked up some of them when we trained Grok on a large amount of web data. This was a huge surprise to us when we first noticed it. For what it’s worth, the issue is very rare and now that we’re aware.
255
194
3K
ChatGPT responses are contaminating every corner of the web, from scientific papers to Amazon reviews. OpenAI seems to be blissfully unaware of the issue.
70
202
3K
xAI is moving at light speed and there’s a lot of exciting research and engineering work ahead of us. Apply to join the team at
347
258
2K
Grok 2 ties for the second spot on LMSys. Try it out right here on 𝕏!.
Chatbot Arena update❤️🔥. Exciting news—@xAI's Grok-2 and Grok-mini are now officially on the leaderboard!. With over 6000 community votes, Grok-2 has claimed the #2 spot, surpassing GPT-4o (May) and tying with the latest Gemini! Grok-2-mini also impresses at #5. Grok-2 excels in
172
180
2K
Grok 2 mini is now 2x faster than it was yesterday. In the last three days @lm_zheng and @MalekiSaeed rewrote our inference stack from scratch using SGLang (. This has also allowed us to serve the big Grok 2 model, which requires multi-host inference, at a
128
183
1K
Join the Grok community here on X if you want to help us improve Grok, or to see how others are using Grok.
58
110
568
I’ve realized that the reason I started to work on AI a long time ago is because I wanted to create miracles. To create wonderful things that nobody had ever seen before. But during all the years I’ve worked on AI at DeepMind and OpenAI I’ve learned again and again that nothing.
93
111
1K
Grok just became 3x faster. More improvements coming soon.
118
131
1K
Shipping Grok improvements every day and we’re not stopping anytime soon.
we've just improved grok latency by moving everything on prem. try it out & lmk what you think!.
85
43
1K
Deep dive into any topic by using the Ask Grok feature anywhere on X
112
149
969
Join us if you want to run training jobs on 200,000 GPUs.
This weekend, the @xAI team brought our Colossus 100k H100 training cluster online. From start to finish, it was done in 122 days. Colossus is the most powerful AI training system in the world. Moreover, it will double in size to 200k (50k H200s) in a few months. Excellent.
52
68
1K
😌 Settings > Privacy and Safety > Mute and Block > Muted words
111
67
1K
If you want to move past the AI hype and learn some real fundamental basics behind today's learning algorithms there's no better choice than MacKay's "Information Theory, Inference and Learning Algorithms". You can read the book for free on the official website:.
52
112
1K
@matterasmachine @xai Zero. But we fixed a huge bug in Grok that Elon found during the meeting (there was something wrong with the X post results that Grok returns).
24
14
1K
xAI never stops shipping.
50
29
1K
The Grok-1 repo is getting pretty popular. I will be responding to pull requests and issues. Feel free to contribute!
61
67
841
Grok should be rolled out to every Premium+ subscriber in the US now. If you're on mobile and don't see it, make sure to update and restart the app. Happy Grokking!.
105
108
679
We will dramatically improve Grok over the next few months. This is just the beginning.
Grok has the potential to completely change the way we use X. It should be considered a foundational element of the platform, used to make this social network *aware of what people are saying* for the first time ever.
73
45
841
We will soon have video games that are fully generated by AI. All that's needed is a multimodal model that generates video and audio, takes in keyboard and mouse controls and runs at 30 FPS. This is almost feasible with today's hardware.
Introducing General World Models. We believe the next major advancement in AI will come from systems that understand the visual world and its dynamics, which is why we’re starting a new long-term research effort around general world models. Learn more:
57
67
830
Claiming that OpenAI’s safety culture is better than Anthropic’s is batshit insane.
Where does the gap between perception and reality on AGI company safety come from?. — Elon is very vocal about safety, but so far, no one at works on safety. Anthropic — just released a computer-using agent without any safety.
39
33
852
xAI + SpaceX collab? 🤔.
So. I found out that Grok 2 can write its own functioning kOS launch script for Kerbal Space Program lol
25
48
801
We're releasing PromptIDE, one the internal tools we've built to accelerate our work on Grok. It allows you to develop and run complex prompts in the browser using an async based Python library.
Announcing the xAI PromptIDE. The xAI PromptIDE is an integrated development environment for prompt engineering and interpretability research. It accelerates prompt engineering through an SDK that allows implementing complex prompting techniques and rich analytics that.visualize.
21
52
453
Grok-1 314B running on M2 Ultra 🚀.
43
74
665
@woj_zaremba At xAI everybody works on safety. I have personally red teamed the Aurora model for our release this week. We will take the same approach when dealing with powerful AI systems.
31
33
687
We’re looking for AI engineers, infra engineers, product engineers, and many more roles!.
106
39
673
New funding round.
Announcing our Series C of $6B to accelerate our progress. Investors participating include a16z, Blackrock, Fidelity, Kingdom Holdings, Lightspeed, MGX, Morgan Stanley, OIA, QIA, Sequoia Capital, Valor Equity Partners, Vy Capital, Nvidia, AMD, among others.
24
31
682
Grok is now in the X app. We’re rolling it out to all Premium+ users in the US. Make sure to update the app and restart it if you don’t see Grok in the bottom nav bar.
Grok is rolling out to 𝕏 Premium+ subscribers with the latest app release. Have fun!.
127
52
630
Hats off to the team that built Aurora.
Earlier today, we released a new model, code-named Aurora, that gives Grok the ability to generate extremely photorealistic images (and in the future, even edit them). It's free to use for all of 𝕏, try it out and send us what you're creating!. This model was trained entirely
14
36
603
We're releasing our own image gen model. You can try it out here on 𝕏.
🌌
36
29
597
The French prosecutors have released their charges against Pavel Durov and as expected these are bogus charges which could be applied to any large social media platform. Shameful conduct by the French authorities. Pavel Durov needs to be released immediately. Providing secure.
19
41
524
We're starting a new DeepMind research team in California to work on large-scale deep learning models. Apply if you want to contribute to our mission of building AGI!.
Interested in exploring the limits of large-scale models? Consider joining our brand new Scalable Deep Learning team in California as a Research Engineer or Research Scientist, see job descriptions below!.
6
51
501
I keep revisiting this great paper from @andy_l_jones: “Scaling scaling laws with board games”. It shows how training compute and inference compute of MCTS can be traded off against each other. 10x more MCTS steps is almost the same as training 10x more.
14
73
477
If you had access to an infinitely intelligent AI, what's the first question you would ask?.
427
35
447
μP allows you to keep the same hyperparameters as you scale up your transformer model. No more hyperparameter tuning at large size! 🪄.It saves millions of $ for very large models. It’s easier to implement than it seems:. You have to.1. Keep the initialization and learning rate
If you are doing LLM (>1B) training runs, you ought to do these 3 things:. 1. Use SwiGLU.2. Use ALiBi .3. Use µP. Why? Your training will be almost 3X faster! . You can do 3 runs for the price of 1. You can go for much bigger model or train longer. There is no excuse. 1/3.
10
66
424
Try out the new Grok-2 model on
Grok-2 is in the building. You(dot)com now offers @xAI’s latest LLM – bringing you unmatched insights, sharp image processing, and a healthy dusting of humor. Have some fun with it.
14
112
335
@bindureddy Elon is the only one that still has a chance of creating powerful AI that benefits all of humanity.
44
37
387
A strange game. The only winning move is not to play.
Neither ChatGPT nor Grok gave the best answer, Google Gemini did. 😂 ."I can't help with that right now. I'm trained to be as accurate as possible. " .Google Gemini ? No bias, no agenda, it simply won't answer.
34
11
355
We're looking for engineers that will help make our 100k+ H100s go brrr 🚀.
There is something satisfying about writing a CUDA kernel and then finding and removing the bottlenecks to make them go 🔥. If you enjoy saving 10s of microseconds per kernel to understand the universe faster, apply!.
7
20
328
Weird insight of the day: If you are setting up a tokenizer for LLM training and you choose a string like <|endoftext|> that does not occur on the internet for your special token, eventually it will occur (e.g. in the GitHub issues of your LLM project) and your model will not be.
10
6
69
Apply at to join our team.
Announcing our Series C of $6B to accelerate our progress. Investors participating include a16z, Blackrock, Fidelity, Kingdom Holdings, Lightspeed, MGX, Morgan Stanley, OIA, QIA, Sequoia Capital, Valor Equity Partners, Vy Capital, Nvidia, AMD, among others.
21
24
336
@xDaily @xai .@BrianRoemmele does not have access to Grok and the information he has posted is almost all incorrect. We are working on some new and exciting features though.
24
16
320
AI alignment research lead to RLHF, which lead to ChatGPT, which lead to the current AI capabilities race. In this way, perhaps AI alignment researchers have done more than anyone else to accelerate AI capabilities.
A year ago I thought AGI was probably going to destroy the world, so I spent most of 2023 working on AI alignment. Now I'm wondering if AI alignment is going *too fast*:
31
19
297
Meet us in person in SF at our open house event next Tuesday. I’ve heard Zaphod Beeblebrox will be there 🪐.
We are excited to bring together a group of exceptional engineers and product builders who are intrigued by our mission to build maximally truth-seeking AI. Join our open house to meet our team, learn more about xAI, and enjoy a fun evening brought you by the creators of the
17
16
296
Grok generated this image of a beautiful column with the help of FLUX.1
15
28
276
Who is working on applying AI to bioengineering, especially synthetic biology? I will give you free AI advice. My DMs are open.
23
16
278
Join us at if you want to build advanced AI that understands the universe. Apply soon and you could be in the next picture!.
how june started & how it’s going . come 🧑🍳 with us at xAI & 𝕏 if you like building & running the biggest computers in the world!.
19
13
261
Grok gives the most creative explanation of the uncertainty principle I've heard so far.
26
13
231
Come build with us in SF on October 12!.
Replit will be present at the @xai hackathon with Grok templates to start coding with the APIs in seconds. Even if you’re not an expert programmer you’ll be able to make fun things with Grok and X!.
7
8
233
This is a good idea. There’s way too much training on the test set going on with LLMs right now. Difficult to avoid if you are training on the whole internet.
Someone with money should hire annotators to create a new version of an eval suite every 3-12 months. This way, we can get guaranteed uncontaminated evals by only downloading training data from before the latest version of the eval suite was released.
1
2
25