EncodeThis's profile picture. AI + Neuro @DeepMind | Ph.D @PrincetonNeuro | ex @ptoncompmemlab. Opinions my own.

Luis Piloto

@EncodeThis

AI + Neuro @DeepMind | Ph.D @PrincetonNeuro | ex @ptoncompmemlab. Opinions my own.

Ghim

I made the mistake of doing a Shrek impression for my toddler. Now she won't stop asking to "talk to Shrek." First thing in the morning. Right up to bedtime. And, folks, my Shrek impression is so bad. Painfully bad. Help.


Luis Piloto đã đăng lại

Introducing Gemini 2.5 Pro Experimental! 🎉 Our newest Gemini model has stellar performance across math and science benchmarks. It’s an incredible model for coding and complex reasoning, and it’s #1 on the @lmarena_ai leaderboard by a drastic 40 ELO margin. Only a handful of…


Luis Piloto đã đăng lại

Our paper, Recursive sequence generation in crows (or Re-crow-sion, my preferred title) is out! Thanks to an awesome team (@KatharinaBrecht, @MillieJohnston_ , A_Nieder), we found that crows join primates in being able to grasp center-embedded sequences. science.org/doi/10.1126/sc…


This is great work in this space, congrats to the authors! I do wish I was a reviewer on this as there are some nuances worth unpacking but overall this is really compelling and important work! Wishlist: 1) try object-based models (OP3?) 2) eval models trained on SAYCam data

(New paper!) Do popular video understanding and embodied models have infant-level physical reasoning abilities? Not as far as we can tell Paper: openreview.net/pdf?id=9NjqD9i… Site: allenai.org/project/inflev…

LucaWeihs's tweet image. (New paper!) Do popular video understanding and embodied models have infant-level physical reasoning abilities?

Not as far as we can tell

Paper: openreview.net/pdf?id=9NjqD9i…
Site: allenai.org/project/inflev…


Trying some code out in Julia and let me tell you: importing (or rather including) into the global namespace is the fastest way to make your code unreadable. Where is this function defined?!? Is this really how people write packages in Julia or am I missing something?


Can you help me identify this robot? I haven't been able to figure it out just from this silhouette. I need to give a presentation to my daughter about the robots on my t-shirt. Thanks!

EncodeThis's tweet image. Can you help me identify this robot? I haven't been able to figure it out just from this silhouette. I need to give a presentation to my daughter about the robots on my t-shirt. Thanks!

I'm really proud of this work. I'm mega, ultra, jumbo grateful for the support of my collaborators, family, and colleagues to make this project a reality. I hope researchers use our dataset directly and/or engage with more nuanced, specific and psych-inspired benchmarks.

Read more about PLATO in @NatureHumBehav: dpmd.ai/nhb-plato and check out the data set via GitHub: dpmd.ai/github-plato. Work by @EncodeThis, Ari Weinstein, @PeterWBattaglia, and Matt Botvinick. 2/2



My 3yo has reinvented the filibuster: at bedtime she asks to sing the first song and then makes up a neverending song. It is too adorable to be frustrating and she does eventually run out of steam.


10 months ago: Galaxy watch 4 released w/o flagship feature 2 weeks ago: Update finally adds flagship feature but causes watch to constantly need factory resetting if enabled. No fix yet. Now: Galaxy watch 5 will have blah blah blah The gadget upgrade treadmill is disgusting.

EncodeThis's tweet image. 10 months ago: Galaxy watch 4 released w/o flagship feature

2 weeks ago: Update finally adds flagship feature but causes watch to constantly need factory resetting if enabled. No fix yet.

Now: Galaxy watch 5 will have blah blah blah

The gadget upgrade treadmill is disgusting.

Using mock objects in unit tests feels a lot like giving students a permission slip for their parents to sign...but then just asking the student if the parent signed it or not. You checked something...just maybe not what you actually needed to check.


This astral projection analogy is what was missing from my ssh lecture slides

EncodeThis's tweet image. This astral projection analogy is what was missing from my ssh lecture slides

I want a full textbook of computer concepts analogized to magic like this

swagitda_'s tweet image. I want a full textbook of computer concepts analogized to magic like this


Luis Piloto đã đăng lại

Excellent, persistent interviewing from @Stone_SkyNews. This is how it's done.

Từ Sky News

Luis Piloto đã đăng lại

We're hiring! The arXiv Technical Director will lead the effort to update arXiv’s technical design and implementation, in order to meet short- and long-term strategic goals cornell.wd1.myworkdayjobs.com/en-US/CornellC… @cornell_tech #OpenScience #Hiring


It's almost surely unintentional that this mistake is in the author's favor. Nonetheless, it further erodes my confidence in the current ML publishing environment.

EncodeThis's tweet image. It's almost surely unintentional that this mistake is in the author's favor. Nonetheless, it further erodes my confidence in the current ML publishing environment.

Luis Piloto đã đăng lại

In new @DeepMind work led by @ToniCreswell, in collaboration with Irina Higgins, we present "Selection-Inference: Exploiting Large Language Models for Interpretable Logical Reasoning" - arxiv.org/abs/2205.09712 1/3


This time last year: I defended my thesis, I lost my mother, and our second daughter was born. All within 10 days. I don't think a year will go by that I won't reflect on the intensity of that time and relive the clash of emotions. I miss you, mom. Your granddaughter is wonderful


task failed successfully 😂

EncodeThis's tweet image. task failed successfully 😂

Luis Piloto đã đăng lại

!!!NEW PREPRINT!!! When do adolescents reach adult levels of executive function? We used FOUR independent datasets (N>10,000), behavioral data from 17 distinct EF tasks, and nonlinear modeling techniques to address this and related questions. 🧵 psyarxiv.com/73yfv

tervoclemmensb's tweet image. !!!NEW PREPRINT!!!
 
When do adolescents reach adult levels of executive function?
 
We used FOUR independent datasets (N>10,000),  behavioral data from 17 distinct EF tasks, and nonlinear modeling techniques to address this and related questions. 🧵
 
psyarxiv.com/73yfv

Luis Piloto đã đăng lại

Intriguingly, transformers can achieve few-shot learning (FSL) without being explicitly trained for it. Very excited to share our new work, showing that FSL emerges in transformers only when the training data is distributed in particular ways! arxiv.org/abs/2205.05055 🧵👇

scychan_brains's tweet image. Intriguingly, transformers can achieve few-shot learning (FSL) without being explicitly trained for it.

Very excited to share our new work, showing that FSL emerges in transformers only when the training data is distributed in particular ways!
arxiv.org/abs/2205.05055  🧵👇

Loading...

Something went wrong.


Something went wrong.