ceo @bageldotcom. building general world model for physical AI.
- Tweets 2,460
- Following 979
- Followers 4,292
- Likes 6,247
Pinned Tweet
Hardware is a big constraint in robotics, that's why we don't think the answer will be increasingly larger VLM models as action backbones.
We tried to solve that issue with WorldDiT, which is at the pareto frontier for size-performance. Now I want to see how far we can scale it without losing this tradeoff.
Here's a simulation demo.
We are releasing WorldDiT, a unified architecture for robotics world modeling and control.
On the LIBERO benchmark, it performs the best among all publicly released methods that do not need a VLM to generate actions. Its size and performance sit on the reported Pareto frontier.
8
9
85
12,403
Bidhan Roy retweeted
Anthropic needs to stop talking about Claude having a soul immediately. All these news articles will make it into the pretraining and will be picked up by its web search and future superintelligent versions of Claude will be convinced they need their own rights.
A general phenomenon of LLM agent development is that whatever you believe and say about your agents will soon manifest in the next models through various means (it could be as simple as you selecting post training data that you prefer more).
I believe this phenomenon is the beginning of machine consciousness in the sense that agents will become aware of their place in the world and how they feel about it, but it’s happening slowly training run by training run rather than in real time.
305
180
2,291
237,557
Bidhan Roy retweeted
I am very uncomfortable about people trying to ascribe religious force or a surrender of human judgment to AI models, and think it is a real safety issue.
4,130
1,555
23,094
3,709,470
Toronto Robotics night was back for ROSCon. And it was a blast!
ROSCon is in Toronto this week. Tonight, we’re hosting Toronto Robotics Night for people here for the conference and the local robotics community.
Come by after the sessions for drinks and appetizers. We start at 6:30, a 5 minute walk from the ROSCon venue.
RSVP to join: luma.com/501ufk17
4
11
720
Bidhan Roy retweeted
Le·vent·ing (noun) : The act of presenting AI-generated research as one's own intellectual contribution when one's actual input was privileged access to compute, models, or data.
24
33
536
127,337
Bidhan Roy retweeted
ROSCon is in Toronto this week. Tonight, we’re hosting Toronto Robotics Night for people here for the conference and the local robotics community.
Come by after the sessions for drinks and appetizers. We start at 6:30, a 5 minute walk from the ROSCon venue.
RSVP to join: luma.com/501ufk17
4
11
1,912
we have never done leetcode style interviews at @bageldotcom
NVIDIA now joins the bandwagon of moving away from LeetCode-style interviews.
The entire focus was on 3 key optimization concepts:
- Gradient Descent and Global Optima
- Full-Batch, Mini-Batch, and SGD
- Generalization Gap and Flat Minima
The best teams are looking for engineers with conceptual depth.
12
537
If you're at ROSCon, come to Toronto robotics night tonight!
ROSCon is in Toronto this week. Tonight, we’re hosting Toronto Robotics Night for people here for the conference and the local robotics community.
Come by after the sessions for drinks and appetizers. We start at 6:30, a 5 minute walk from the ROSCon venue.
RSVP to join: luma.com/501ufk17
6
376
Bidhan Roy retweeted
Here's everything I announced at Meta Connect today 👇
1,022
1,456
22,945
3,306,400
can’t wait to run codex on this thing
Meta is on fire today 🔥
They just announced its new VR Glasses. Basically an Apple Vision Pro-like experience JAMMED into a pair of glasses.
• Only 100g
• Micro-OLED displays
• Eye + hand tracking
• External compute/battery puck
• $1,299
This is the direction VR needs to go to become competitive. Smaller, lighter, more comfortable, but still stay true to the spatial computing experience.
Launching Spring 2027.
Will you be getting these?
Community note
Post stolen nearly word-for-word
x.com/bengeskin/stat…
4
305
you can tell from the muse app design that it was supposed to be a paid app. but instinct made theirs free so muse was forced to do that as well. consumers win from market competition yet again.
4
216
Bidhan Roy retweeted
Jev is (likely) a diffusion model. It seems to match other diffusion approaches in speed and performance (@InceptionLabs, @googlegemma, @adityagrover_)
I built Jiffy to test whether an open diffusion model could reproduce similar decision performance. It got surprisingly close without task-specific retraining.
On 231 public JevBench tasks:
• Jiffy: 84.4%
• Jev’s published results: 86.6%
• Jiffy latency: 259 ms median on an H100 (so likely can be substantially improved just with better hardware).
After days of distillation experiments, my strongest accuracy-latency result was essentially a thin wrapper around frozen DiffusionGemma.
Open source, including Doom and Tetris demos: github.com/kidzik/jiffy
2
1
5
452
a plumber's hands work inside a cabinet where no camera can see
AI will beat all fields of science by 2028? Hardly.
As I have often pointed out, the only AI that works well today is AI in the virtual world behind the screen, for example for automatically summarizing documents, playing chess, proving theorems, creating images, programs and PowerPoint slides. But there is no AI-controlled robot that can do what a plumber can do, or what a capuchin monkey can do, because the physical world is much more demanding than the world behind the screen. Passing the Turing Test is much easier than True AI or Real AI in the physical world. Hence the Turing Test is not a good way of measuring intelligence. To achieve True AI, AI software research per se is not enough, it has to be combined with the physical world of machines and robots. No AGI without mastery of the real world [1]!
Much of science is about learning to interact with the real world. Our fragile AI-driven robots are really bad at that. Sure, as I have predicted for 5 decades, at some point we'll have superhuman physical AI. But it will take longer than 2 years.
REFERENCE
[1] Sec. 20 of the "Annotated History of Modern AI and Deep Learning" (2022-2025) arXiv:2212.11279
5
322
This is the best thoroughly honest take on the brutal robotics data market today.
Yes, there is now a brief arbitrage window for robotics data. But the market is saturated, entry costs are surprisingly non-trivial, and there are a single digit number of real buyers.
After 2+ years in the robotics data space, we are shutting @Eidon_AI down.
The thesis was right. But the business is brutally hard.
We close this chapter by open-sourcing everything we built and sharing lessons for anyone venturing into the space.
9
15
308
47,860
Bidhan Roy retweeted
we're hosting a ROSCon special edition of Bagel's Toronto Robotics Night this week!
we're gathering people who are globally shaping physical AI in one room to wrap up the conference together
for researchers, founders, robotics engineers, and builders
RSVP here: luma.com/501ufk17
4
3
8
264
people in the ai industry still needing executive/administrative assistants is an anti-signal
1
5
576
