Clemens Marschner

DeepSeek has released JanusFlow model.

Model: huggingface.co/deepseek-ai/...

The “small team“ had 100+ people and 1000s of GPUs to spare

Please explain - ROS2 has many best practices built in and the systems I have seen that didn’t use it were inferior in terms of speed or flexibility, respectively. What would you do better?

Stop it…

Foundations of Large Language Models by Tong Xiao, Jingbo Zhu

This is a book (231 pages) about large language models. It primarily focuses on foundational concepts rather than comprehensive coverage of all technologies. The book is structured into four main chapters, each exploring a key area:

Happy New Year everyone! Jim and I just put up our January 2025 release of Speech and Language Processing! Check it out here: web.stanford.edu/~jurafsky/sl...

Hard to recognize the contents compared to my version from 2003… after Manning/Schütze my favorite NLP book. MS was just beautiful with Cambridge University Press‘ typesetting. I loved it on a visceral level. But Jurafsky/Martin was also very accessible.

If you‘re in ML, consider robotics at this point. Especially if you‘re in Europe.

There are amazing challenges in the space of spatial intelligence, planning, understanding of the physical world, control to be solved with AI.

And if you want to turn it into products, contact me.

Just switching over from X for today.
Is there still sanity on this platform at least?

Going back and forth between 1 week and 1 year, 2 year, 5 year timelines and loving it!
There is no such thing as an architecture role. Every IC writes code, and that's how it should be.
It's just that more senior engineers should think strategically and shape where a company will be in the future.

On site, testing the #RobCo vision system 🦾🤖

Meanwhile, the @bsky.app developers…

YouTube video by Steven (MechaGodzi11a) Riptide Opening and Closing Credits and Theme Song

Def Riptide qhttps://youtu.be/bdhrYdWlxTw?si=zEMJ4JAAl7g7kkqk

I've spent the last two years scouring all available resources on RLHF specifically and post training broadly. Today, with the help of a totally cracked team, we bring you the fruits of that labor — Tülu 3, an entirely open frontier model post training recipe. We beat Llama 3.1 Instruct.

Thread.

That comment is on brand

I‘m turning this into my job account and will focus on robots and neural networks here. For architecture and city planning, see @cmarschnerde.bsky.social - like on X

I recently gave a tutorial on the DUSt3R paper (web: dust3r.europe.naverlabs.com, paper: tinyurl.com/5t2ks575, code: github.com/naver/dust3r) in a research group meeting. In case you missed it, didn’t understand it or would like to hear some perspectives on why it’s such a cool idea, read on… 1/23

With all those starter packs and the Xodus, ML Bluesky now feels like Twitter 2016. Finally, content

Also can we rename this to Twitter pls thx

Resilience is the art of keeping things working in the light of problems.
It requires slack. Slack is the enemy of efficiency.
A society that has gone too far optimizing for efficiency will constantly be on the verge of collapse