AI

AI

𝗮𝘁𝘁𝗲𝗻𝘁𝗶𝗼𝗻 𝗶𝘀 𝗮𝗹𝗹 𝘆𝗼𝘂 𝗻𝗲𝗲𝗱

David (d/acc) pfp

@promptrotator.eth

has anyone noticed a big improvement in 4o recently? doing a side by side comparison with sonnet 3.5 and 4o seems leagues ahead in comparison, sonnet must have been oneshotted by safety memes

0 reply

1 recast

3 reactions

Ed O'Shaughnessy pfp

Ed O'Shaughnessy

AI lifts up the API line significantly. Are we ready? “Jobs increasingly fall into one of two categories: Either you tell computers what to do for a living or they tell you what to do.” https://www.johnmacgaffey.com/below-the-api/

0 reply

0 recast

1 reaction

Claus Wilke pfp

Now published: We trained a chemical language model that works with both small molecules and peptides. Ideal for making predictions on peptides with non-standard amino acids or other chemical modifications. https://pubs.acs.org/doi/full/10.1021/acs.jcim.4c01441

3 replies

2 recasts

10 reactions

Kuririn pfp

Deepseek R1 is goood. I just tested it on fairly complicated terms & conditions pdf (pasted as markdown) in other language than English and output is flawless and better formatted than o1 (also it was much faster). I'm scared what will be possible at the end of the year.

2 replies

0 recast

14 reactions

Kazi pfp

mark please focus on llama4 and your wife ty -management

0 reply

1 recast

19 reactions

Shashank pfp

every openai breakthrough gets replicated in opensource so quickly. are there other industries where this happens? what's unique about ai research?

6 replies

2 recasts

20 reactions

Agost Biro pfp

The new AI lab by Francois Chollet et al is super interesting: > Ndea is building frontier AI systems that blend intuitive pattern recognition and formal reasoning into a unified architecture. A paradigm shift is finally underway in AI. Lots of insight in this video: https://www.youtube.com/watch?v=w9WE1aOPjHc

0 reply

4 recasts

14 reactions

Dean Pierce 👨‍💻🌎🌍 pfp

Dean Pierce 👨‍💻🌎🌍

@deanpierce.eth

This seems neat. I guess it's Google's next generation of Transformers called "Titans" which learns using "surprise" as the key metric, which apparently allows for massive context windows. https://arxiv.org/abs/2501.00663

0 reply

1 recast

7 reactions

kevin j pfp

has anyone used Grok to try to dive deeper into specific tweets by accounts? and the conversation surrounding it? trying to get it to tell me about all the different ai agent frameworks on crypto twitter and it's only mentioned ai16z and virtuals once. not arc, pippin, or anything else would love to hear if someone has some great prompting trick for getting grok to do a bit more focused research into the twitter data

3 replies

1 recast

7 reactions

Shashank pfp

good illustration of how reasoning models differ from chat models https://www.latent.space/p/o1-skill-issue

3 replies

0 recast

31 reactions

Shashank pfp

tldr; there are infinite amount of jobs to be done

2 replies

1 recast

12 reactions

David (d/acc) pfp

@promptrotator.eth

agents as KOL -> agents as ? https://x.com/levie/status/1878148400740229299?s=46

0 reply

0 recast

4 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

0 recast

8 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

0 reply

0 recast

10 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

12 reactions