Thomas Lux (@tchlux) header
Thomas Lux (@tchlux) profile picture

Research scientist at Meta (MSL FAIR). PhD in Computer Science. Working on AI theory+algorithms in free time.

Philadelphia, PA
Joined March 2018
Photo by Thomas Lux (@tchlux)Photo by Thomas Lux (@tchlux)Photo by Thomas Lux (@tchlux)Photo by Thomas Lux (@tchlux)Photo by Thomas Lux (@tchlux)Photo by Thomas Lux (@tchlux)Photo by Thomas Lux (@tchlux)Photo by Thomas Lux (@tchlux)Photo by Thomas Lux (@tchlux)Photo by Thomas Lux (@tchlux)
Step 3: Do a rep of gratitude every day. No matter who you are, this is proven to help your mind and your body. And when it feels like there is absolutely nothing you can do, you can do this one thing. Write down one thing you’re grateful for, every day.
17
40
7
1,385
184,241
This week we speak with @tchlux about interpolation of sparse high dimensional data using deep learning and other techniques. We use geometry as the lens to discuss ML techniques. with @ecsquendor @DoctorDuggar youtube.com/watch?v=ZaOp1KNh…
3
6
1
18
A security guard tailed me on my walk home tonight. He demanded if I lived there because “you look suspicious.” I showed my keys & buzzed myself into my building. He left, no apology. This is the reality of black girls: One day you’re called an icon, the next day, a threat.
Wow, a fantastic @washingtonpost piece by @nnekamcguire . We live in a contradictory society that can celebrate a black girl poet & also pepper spray a 9 yr old. Yes see me, but also see all other black girls who've been made invisible. I can not, will not, rise alone.
3,647
21,726
2,374
90,423
Thomas Lux retweeted
Presenting DreamerV2, the first world model-based #ReinforcementLearning agent to achieve top-level performance on the Atari benchmark, learning general representations from images to discover successful behaviors in latent space. Read more at goo.gle/3ub8iZe
17
254
34
935
Thomas Lux retweeted
Introducing the World Models Library, an open-source, platform-agnostic suite of tasks and tools for examination of world model design and performance in visual model-based reinforcement learning. Learn more and grab the code at goo.gle/36EXY1t
7
291
16
1,009
Thomas Lux retweeted
Introducing Room-Across-Room (RxR) — the first multilingual dataset for vision-and-language navigation — and the associated RxR challenge, which encourages the ML community to train and evaluate their own instruction following agents. Learn more at goo.gle/395CUmA
9
158
15
532
Thomas Lux retweeted
Introducing ToTTo, an open domain table-to-text generation dataset, a challenging benchmark for high-precision text generation, and a controlled text generation task that can be used to assess model “hallucination”. Read more and grab the code at goo.gle/2XFMR3o
6
134
12
540
Our paper on training a single goal-conditioned policy 100% with asymmetric self-play to generalize to many unseen objects and tasks: arxiv.org/abs/2101.04882 and more cool videos are available at robotics-self-play.github.io… (The attached video is zero-shot)
3
57
5
342
While updating NeRF-in-the-Wild, we added 3 more scenes to the paper: Prague's Old Town Square, Hagia Sophia & Taj Mahal. They came out pretty well! I could not resist myself and created a demo reel / virtual world tour video showcasing the results at 720p60hz! Enjoy! 😃
2
17
5
143
Thomas Lux retweeted
10 months ago we shut down Atrium after raising $75m in venture capital. Anyone hearing that knows I made tons of mistakes along the way. Someone asked me today what my biggest lessons learned were. Here they are:
259
1,768
357
8,759
Thomas Lux retweeted
We’ve developed two neural networks which have learned by associating text and images. CLIP maps images into categories described in text, and DALL-E creates new images, like this, from text. A step toward systems with deeper understanding of the world. openai.com/multimodal/
90
899
328
2,872
Thomas Lux retweeted
Announcing a new dataset: the Pile! A free and publicly available 800GB dataset of diverse English text for language modeling! Download: pile.eleuther.ai/ Paper: pile.eleuther.ai/paper.pdf 1/7
7
259
32
1,140
Interested in offline RL? Handling image observations & continuous actions is important for offline RL in the real world. We introduce LOMPO to tackle this setting. arxiv.org/abs/2012.11547 with Rafael Rafailov, @TianheYu, @aravindr93 🧵👇(1/4)
3
43
1
251
First release of my code for ACM TOMS Alg 1012: DelaunaySparse -- code for piecewise linear interpolation in arbitrary dimension: vtopt.github.io/DelaunaySpar… Interfaces for Python 3.6+, Fortran, C/C++, and command line Read the paper here: doi.org/10.1145/3422818
2
4
From the International Space Station: I voted today — Kate Rubins
309
2,978
1,273
23,413
Announcing DQN Zoo: a collection of reference implementations of reinforcement learning agents developed at DeepMind & based on the Deep Q-Network (DQN) agent: https://nitter.gq/t.co/4WS6ngkpvL Work by @johnquan_ and @georgostrovski, with support from many others!
101
4
314
Time lapse of this morning’s stunning pre-dawn Falcon 9 launch from Cape Canaveral. 🎥: Me for @SuperclusterHQ
62
936
126
3,512
Research finds a new way to hack Siri and #GoogleAssistant with #ultrasonicwaves techxplore.com/news/2020-03-…
14
21
Thomas Lux retweeted
Tesla’s R&D productivity is completely insane. It is out of this world. 1/5 $TSLA @elonmusk @Tesla
42
525
62
3,551