Troy SK (@troysk704) profile picture

WonderJunkie

Foothills of Darjeeling
Joined April 2013
Photo by Troy SK (@troysk704)Photo by Troy SK (@troysk704)Photo by Troy SK (@troysk704)Photo by Troy SK (@troysk704)Photo by Troy SK (@troysk704)Photo by Troy SK (@troysk704)Photo by Troy SK (@troysk704)Photo by Troy SK (@troysk704)Photo by Troy SK (@troysk704)Photo by Troy SK (@troysk704)
It was more than 50, the watch died before my legs 🤪
1
2
729
Troy SK retweeted
Watch and weep. We live in an anti-science era under BJP RSS.
58
781
25
1,932
34,559
Troy SK retweeted
Everyone's been showing Fable and GPT-6 Astra driving robot arms. We trained an open 27B Qwen model to do it, purely from its own attempts in simulation, with zero demonstrations. And it can run locally at 220+ tok/s on an RTX 5090. On a real robot it had never touched, it now succeeds in 18 of 30 tests, up from 0 before training. It also copies a painting and designs its own chemistry experiment. It generalizes and follows instructions better than π0.5, and on LIBERO it comes close to Astra max. All without losing its general capabilities, like math and coding. Built on Qwen3.8-27B. We've open-sourced the model weights, code, and full training infra! Check out our blog at mit-han-lab.github.io/pear-b….
52
168
49
1,230
109,316
Troy SK retweeted
Post-Astra, robotics simulation has become more crucial than ever. But visual fidelity is not enough to make simulation useful for robotics: sim needs to simultaneously (1) closely reproduce reality's physical dynamics, and (2) cover a vast diversity of real-world scenes. At @VersorInc, we pioneer real2sim2real methods which accomplish both. 🧵
4
5
4
87
59,431
Today marks 10 years since #Demonetisation. It was the first time my friends chose the Government narrative over my arguments or common sense. I saw 100s of small businesses die out because of one decision, GST helped accelerate. As they say, elect a clown expect a circus!
42
This is fantastic!!!
É a primeira vez que eu vejo uma música eletrônica analógica
4
12
68
9,389
I redesigned Calculator. You're welcome.
932
5,299
492
59,066
1,726,783
Troy SK retweeted
Remember one thing when you read JEPA vs LLMs in the context of the platonic representation hypothesis. Hidden state of an LLM does not represent the full representation of the LLM, it carries only what is needed for the next token. It's not an encoder.
43
32
4
583
70,043
When in school, I had some classmates who were great memorizers/muggers. They lacked imagination and ended up with mediocre career/life because they never tried to understand things and just be a parrot. Be abstract and populate that latent space!
Replying to @heyshrutimishra
LLMs accumulate a lot of declarative knowledge. That's how they can answer a large number of questions. That's also why they require so many parameters and hence so much memory. But one can replace the accumulation of knowledge by *understanding* That what JEPA world models do. That's one reason they can be useful with considerably less memory and training data than LLMs. This paper merely shows that the underlying similarity structure of the data, whether text or image (e.g. co-occurence of words in text or objects in images) uniquely determines the shape of the representation point cloud, when the encoder is trained trained with a common anti-collapse criterion like SIGReg or VCReg (not by reconstruction).
68
Troy SK retweeted
She explains so many things in 2 mins. this will stay with me
She gets it. She just does
2
5
27
1,728
Troy SK retweeted
The great gift of agents is how they allow you to route around every insufferable gatekeeper in software.
264
247
40
5,447
167,946
Troy SK retweeted
Yann LeCun was right & nobody in this industry wants to say it out loud. 109 million parameters beat 72 billion. He has spent years saying scale is not the thing that makes a model understand the world. Meta bet the other way. He left in November and started his own lab in Paris. Last Thursday he published the proof. The theory everyone has been running on since 2024 is that if you train models big enough, on text, on images, on sound, they all end up with the same picture of reality inside them. Size does it. Nobody ever proved that part. Randall Balestriero and LeCun proved it. And the proof says size was never doing the work. What does the work is which two things you show a model together. A photo and the caption of that photo are two views of one thing. Models trained on pairs like that get there fast. Two paragraphs pulled from the same document are not two views of one thing. Models trained on those get there slowly and expensively. Next word prediction is the second kind. That is what every chatbot in the world runs on. In their tests the first method reached the same understanding after 21.6 million examples. Next word prediction needed 900 million. And a 109 million parameter text model ended up closer to an image model's view of the world than a 72 billion parameter language model was. 660 times smaller.
268
919
161
6,873
810,588
1. Direct conversion of human BLOOD cells into NEURAL stem cells drives near-complete epigenetic REJUVENATION, with no pluripotent intermediate (i.e. lower risk of cancer). Series of comments on the paper by Lea Berg and Oliver Brüstle. doi.org/10.1111/acel.70751 🧵
6
38
2
230
9,626
I have been building a new app every month to help the token companies to achieve maximum utilization 🙌🫡🤪
Sam Altman says its bad that everyone's personal computers are using the world's limited supply of memory chips. "It’s quite terrible to have everybody have a bunch of the world’s limited supply of DRAMs sitting in their own computer, and being used 5% of the time. Like, you want maximum utilization."
34
Troy SK retweeted
With @JayHe748646 @KehanWen170077 @leggedrobotics , we push humanoid mobility to next frontier. Achieved by a generalist contact planner + a generalist tracker + a generalized neural mapping, warmstarted from few human data. Website: leggedrobotics.github.io/ACP… Promo Video 👇
27
75
21
527
54,764
Troy SK retweeted
If you are going to Dharamshala International Film Festival, I made a little something for myself. Maybe you could use it as well. festplanner.app – at Dharmsala, India
5
8
1,357
Unitree Introducing: H2-X Kung Fu Legend Guardian🤖 Powerful performance to serve and guard you, with a maximum endurance of approximately 8 hours. Please use the robot safely and in a friendly manner.
1,959
3,821
3,068
28,765
18,146,051
Troy SK retweeted
A weekly jab in the belly is generating more revenue than the entire AI industry. Ozempic + Mounjaro: $71B in 2025. OpenAI + Anthropic: $29B. And they've barely started. ~2% of the 800 million eligible patients can currently access them. h/t @DrSamuelBHume
🤖 Made with AI
79
256
122
1,509
893,179
Troy SK retweeted
Never disturb a toddler in this moment! Don’t help. Don’t suggest. Observe. A capable problem solver is in training and development.
The stubbornness will save her life.
342
18,652
170
215,004
3,489,395
First in my bloodline to see a former fashion designer, who is Bengali, trying to pass off completely painted road surfaces in solid block colours as aalpona. The first is aalpona on road, and the second is what this government is trying to pass off as aalpona.
Kolkata, West Bengal: Minister Agnimitra Paul says, "This is not a permanent fixture. We decorated the roads with colors and painted 'Alpona' for Durga Puja. Alpona is a traditional heritage of us Bengalis. Moreover, painting roads isn't happening for the first time here; it exists in America, Iceland, and many other countries. Additionally, the paint we used is anti-skid paint..."
65
4
517
13,898