𝚐𝔪𝟾𝚡𝚡𝟾

@gm8xx8

99 Following

77012 Followers

𝚐𝔪𝟾𝚡𝚡𝟾 pfp

𝚐𝔪𝟾𝚡𝚡𝟾

LONG LIVE OPEN SOURCE

1 reply

6 recasts

22 reactions

𝚐𝔪𝟾𝚡𝚡𝟾 pfp

𝚐𝔪𝟾𝚡𝚡𝟾

we are not the same

7 replies

41 recasts

254 reactions

𝚐𝔪𝟾𝚡𝚡𝟾 pfp

𝚐𝔪𝟾𝚡𝚡𝟾

You ask me who to follow in AI, but don’t even follow me—after I’ve spent two years sharing insights and building a presence on Farcaster. Then, there’s another person who literally copied my following list, recommended others follow me, but when I tried to return the favor? Realized I’m blocked. I guess they just wanted to give the impression they are in the know. I see why so many AI folks have churned ☹︎ I’ll still check in and post occasionally—this isn’t me quitting Farcaster or anything like that, just sharing some 💭 Attached below is one example, but I’ve received plenty of similar messages—some harsher towards farcaster, all eye-opening, all from AI folks. I also made sure they knew I’m still here, just not posting as often. My experience hasn’t been all bad—many from Farcaster have reached out to say they only know what they do about AI because of me and have shared their gratitude. That support is why I still return.

11 replies

1 recast

19 reactions

𝚐𝔪𝟾𝚡𝚡𝟾 pfp

𝚐𝔪𝟾𝚡𝚡𝟾

https://warpcast.com/gm8xx8/0xb8d6774c

1 reply

0 recast

5 reactions

𝚐𝔪𝟾𝚡𝚡𝟾 pfp

𝚐𝔪𝟾𝚡𝚡𝟾

Physical-Intelligence π₀ is now open source. > π₀ and π₀-FAST Model Weights > source code for model architecture, on-robot Inference, and fine-tuning > pre-trained fine-tuned checkpoints robotics got another boost ✔️ 🔗: https://github.com/Physical-Intelligence/openpi https://warpcast.com/gm8xx8/0xce446768

2 replies

0 recast

13 reactions

𝚐𝔪𝟾𝚡𝚡𝟾 pfp

𝚐𝔪𝟾𝚡𝚡𝟾

FAST is a robot action tokenizer that simplifies and speeds up robot training. It enables: > 5x faster training compared to diffusion models. > Compatibility with all tested robot datasets. > Zero-shot performance in new environments, including the DROID dataset, successfully controlling robots in various settings with ease. > Simple autoregressive VLAs that match diffusion VLA performance. > Mixed-data VLA training, allowing integration of non-robot data like web data, subgoals, and video prediction. FAST compresses actions using discrete cosine transform, reducing redundancy and enabling efficient VLA training on high-frequency tasks. It scales to complex robot tasks with simple next-token prediction, converging in days instead of weeks. A pre-trained FAST tokenizer based on 1M robot action sequences is available on Hugging Face, working across various robots and supporting mixed-data VLA training. https://huggingface.co/physical-intelligence/fast

0 reply

4 recasts

10 reactions

𝚐𝔪𝟾𝚡𝚡𝟾 pfp

𝚐𝔪𝟾𝚡𝚡𝟾

DeepSeek releases Janus-Pro-7B and Janus-Pro-1B 🐋

0 reply

3 recasts

25 reactions

𝚐𝔪𝟾𝚡𝚡𝟾 pfp

𝚐𝔪𝟾𝚡𝚡𝟾

|￣￣￣￣￣￣￣￣￣￣￣￣￣l | everytime I open this app it l | feels like a step back in l l time. l |＿＿＿＿＿＿＿＿＿＿＿＿＿| \ (•◡•) / \ / —— | | |_ |_

4 replies

12 recasts

26 reactions

𝚐𝔪𝟾𝚡𝚡𝟾 pfp

𝚐𝔪𝟾𝚡𝚡𝟾

Open research and open source accelerate progress by fostering collaboration, transparency, and accessibility. They enable people to build on existing work, reduce redundancy, and solve problems more efficiently. Openness has always been a key driver of innovation and growth—open is the path forward!

2 replies

5 recasts

18 reactions

𝚐𝔪𝟾𝚡𝚡𝟾 pfp

𝚐𝔪𝟾𝚡𝚡𝟾

Looks like they’re finally paying attention. I still remember when I posted this. It wasn’t a direct response, but I shared it because someone mentioned Chinese models and how far behind they were after referencing a LinkedIn post. Zero engagement—people were too busy copying and pasting everything they saw on X or elsewhere to look like they were in the know, rather than learning from those actually in the trenches.

4 replies

18 recasts

24 reactions

𝚐𝔪𝟾𝚡𝚡𝟾 pfp

𝚐𝔪𝟾𝚡𝚡𝟾

AI evolution at its finest. Never said, “Don’t sleep on Meta,” but I’ve been shouting “Don’t sleep on DeepSeek” for well over a year! It’s incredibly rewarding to see my research and work validated with their latest release.

2 replies

0 recast

13 reactions

𝚐𝔪𝟾𝚡𝚡𝟾 pfp

𝚐𝔪𝟾𝚡𝚡𝟾

OpenAI released Operator ngl i’ve been quite impressed with it so far but I find the guardrails a bit excessive. https://m.youtube.com/watch?v=CSE77wAdDLg https://openai.com/index/introducing-operator/

0 reply

1 recast

7 reactions

𝚐𝔪𝟾𝚡𝚡𝟾 pfp

𝚐𝔪𝟾𝚡𝚡𝟾

Every time I open Farcaster, it feels like a step back in time.

1 reply

0 recast

5 reactions

𝚐𝔪𝟾𝚡𝚡𝟾 pfp

𝚐𝔪𝟾𝚡𝚡𝟾

Ilya Sutskever talk at NeurIPS 2024 Seq2Seq w/ Neural Networks https://m.youtube.com/watch?v=1yvBqasHLZs

1 reply

3 recasts

7 reactions

𝚐𝔪𝟾𝚡𝚡𝟾 pfp

𝚐𝔪𝟾𝚡𝚡𝟾

Would you consider merging with AI? Would you become part AI?

2 replies

0 recast

3 reactions

𝚐𝔪𝟾𝚡𝚡𝟾 pfp

𝚐𝔪𝟾𝚡𝚡𝟾

AgiBot World is an open-source dataset for robotic learning with over 1M trajectories from 100+ real-world scenarios, covering tasks like manipulation, tool use, and multi-robot collaboration. https://agibot-world.com

0 reply

1 recast

4 reactions

𝚐𝔪𝟾𝚡𝚡𝟾 pfp

𝚐𝔪𝟾𝚡𝚡𝟾

Moondream 2025-01-09 Release: Structured Text, Enhanced OCR, Gaze Detection https://moondream.ai/blog/introducing-a-new-moondream-1-9b-and-gpu-support

1 reply

0 recast

6 reactions

𝚐𝔪𝟾𝚡𝚡𝟾 pfp

𝚐𝔪𝟾𝚡𝚡𝟾

This will all make sense 🔜

0 reply

0 recast

2 reactions

𝚐𝔪𝟾𝚡𝚡𝟾 pfp

𝚐𝔪𝟾𝚡𝚡𝟾

rStar-Math shows SLMs can rival or surpass OpenAI o1 in math reasoning w/out distillation from larger models, using MCTS and three keys factors: 1. Code-Augmented CoT Synthesis: MCTS generates verified reasoning data to train policy SLMs. 2. Enhanced PRM: A novel training approach avoids naïve annotations, yielding a stronger process preference model (PPM). 3. Self-Evolution Framework: Four rounds of self-evolution refine reasoning with millions of synthesized solutions for 747k problems. Performance Highlights: > Achieves 90.0% on MATH, improving Qwen2.5-Math-7B by +31.2% and surpassing OpenAI o1-preview by +4.5%. > Boosts Phi3-mini-3.8B from 41.4% to 86.4%. > Solves 53.3% of AIME problems, ranking in the top 20% of high school competitors. don’t sleep on small models. https://arxiv.org/abs/2501.04519

1 reply

0 recast

9 reactions

𝚐𝔪𝟾𝚡𝚡𝟾 pfp

𝚐𝔪𝟾𝚡𝚡𝟾

This model didn’t quite pass my vibe check back in December, so I held off on sharing. That said, there’s still something to learn from this release, even if it’s not my top pick among SLMs right now.

0 reply

0 recast

2 reactions