lucas @_lukaemon

my scratchpad for future agi to read lucasshen.org Seattle Joined July 2020

Tweets

6K
Followers

698
Following

5K
Likes

10K

lucas @_lukaemon

41 minutes ago

weak and empty people find meaning by controlling other people, setting unnecessary rules and enforce. creative people find meaning by appreciation, exploration and creation.

0 0 0 13 0

lucas @_lukaemon

22 hours ago

it's an interesting experiment. however, current web formats are pure legacy. link and website could be useful but not necessary ux elements in ai native info sphere.

Pietro Schirano @skirano

23 hours ago

it's an interesting experiment. however, current web formats are pure legacy. link and website could be useful but not necessary ux elements in ai native info sphere.

101 90 2K 192K 872

Download Video

0 0 1 122 0

In this game, agents could either selfishly harvest resources or cooperate to maintain a public good. When anonymous, they acted selfishly and the system collapsed. But when they could be identified and build reputations, they cooperated, even spontaneously inventing turn-taking…

18 32 225 17K 98

Download Image

Edward Z. Yang @ezyang

a week ago

Without further ado, The Parallelism Mesh Zoo

8 54 499 44K 457

Download Image

John Kraus @johnkrausphotos

2 weeks ago

Always worth the wait.

John Kraus @johnkrausphotos

2 weeks ago

Always worth the wait. https://t.co/Qnq91CaGzq

61 174 2K 322K 67

Download Image

542 1K 11K 13.9M 645

Download Image

lucas @_lukaemon

2 weeks ago

firework, beautiful

0 0 1 37 0

Download Image

lucas @_lukaemon

3 weeks ago

Lucas Beyer - Computer Vision in the Age of LLMs youtube.com/watch?v=kxO6AR… 'the grass is eating the horse'

1 0 0 74 0

Download Image

Jacob Austin @jacobaustin132

3 weeks ago

Today we're putting out an update to the JAX TPU book, this time on GPUs. How do GPUs work, especially compared to TPUs? How are they networked? And how does this affect LLM training? 1/n

38 531 3K 377K 5K

Download Image

Denny Zhou @denny_zhou

2 months ago

Slides for my lecture “LLM Reasoning” at Stanford CS 25: dennyzhou.github.io/LLM-Reasoning-… Key points: 1. Reasoning in LLMs simply means generating a sequence of intermediate tokens before producing the final answer. Whether this resembles human reasoning is irrelevant. The crucial…

41 460 3K 266K 4K

lucas @_lukaemon

4 weeks ago

meaning, even current multiturn tool calling agent is a temporary scaffold, useful one though.

Elon Musk @elonmusk

2 months ago

meaning, even current multiturn tool calling agent is a temporary scaffold, useful one though.

541 1K 7K 1.3M 716

1 0 0 72 0

Hattie Zhou @oh_that_hat

4 weeks ago

AI models “think” in two ways: - in the latent space over layers - in the token space over a sequence Latent space = natural talent, chain of thought = hard work. Just like for humans, hard work can get you far, but talent sets the ceiling. This is why pretraining can’t die.

29 52 619 46K 345

Costa Huang @vwxyzjn

4 weeks ago

Check out Jason’s new work on RL tricks. Lots of ablation studies comparing popular techniques like Clip Higher 🤩

Jason Liu @JasonLiu106968

4 weeks ago

Check out Jason’s new work on RL tricks. Lots of ablation studies comparing popular techniques like Clip Higher 🤩

8 27 154 30K 157

Download Image

1 6 80 7K 46

lucas @_lukaemon

4 weeks ago

45mm stack height is too much. running shoes are running out of control ...

1 0 0 60 0

Igor Babuschkin @ibab

4 weeks ago

Today was my last day at xAI, the company that I helped start with Elon Musk in 2023. I still remember the day I first met Elon, we talked for hours about AI and what the future might hold. We both felt that a new AI company with a different kind of mission was needed. Building…