Akarsh Kumar (@akarshkumar0101) / X

Akarsh Kumar

518 posts

Akarsh Kumar

@akarshkumar0101

PhD Student @MIT_CSAIL RS Intern @SakanaAILabs RL, Meta-Learning, Emergence, Open-Endedness, ALife

Joined February 2018

Pinned
Akarsh Kumar
@akarshkumar0101
Jun 7
We never really knew how to train nonlinear RNNs well… BPTT struggled with vanishing grads (no long-range memory) and sequential rollout (hard to parallelizable). What if instead an oracle told us the optimal memory state m_t at each step? Then the RNN could do one-step
181K
Akarsh Kumar
@akarshkumar0101
Apr 1, 2022
I'm very grateful to share that I'll be joining @MIT_CSAIL in the Fall to do a PhD in AI research!!! Super excited to work with @phillip_isola on AI-GAs (meta-learning) and Open-Endedness (evolutionary computation) for creating intelligent embodied agents!!!
Akarsh Kumar
@akarshkumar0101
May 20, 2025
Excited to share our position paper on the Fractured Entangled Representation (FER) Hypothesis! We hypothesize that the standard paradigm of training networks today — while producing impressive benchmark results — is still failing to create a well-organized internal
Kenneth Stanley
@kenneth0stanley
May 20, 2025
Could a major opportunity to improve representation in deep learning be hiding in plain sight? Check out our new position paper: Questioning Representational Optimism in Deep Learning: The Fractured Entangled Representation Hypothesis. The idea stems from a little-known
36K
Akarsh Kumar
@akarshkumar0101
Dec 24, 2024
Very excited to share ASAL! Artificial Life aims to recreate natural evolution, but is severely bottlenecked by hand-designed simulations. We propose using CLIP to automatically discover the interesting ALife simulations!
00:16
Sakana AI
@SakanaAILabs
Dec 24, 2024
Introducing ASAL: Automating the Search for Artificial Life with Foundation Models sakana.ai/asal/ Artificial Life (ALife) research holds key insights that can transform and accelerate progress in AI. By speeding up ALife discovery with AI, we accelerate our
46K
Akarsh Kumar
@akarshkumar0101
Mar 10, 2025
I recently gave a talk at the MIT Embodied Intelligence Seminar on Automating the Search for Artificial Life with Foundation Models, check it out!
Akarsh Kumar - Automating the Search for Artificial Life with Foundation Models
From youtube.com
49K
Akarsh Kumar
@akarshkumar0101
Jan 9, 2025
This video shows my grand vision for ALife: evolving alien intelligences from scratch. We're still far from achieving it. If you are interested in making it happen (through ALife or RL), please reach out!
00:00
14K
Akarsh Kumar
@akarshkumar0101
May 27, 2025
Replying to @skalskip92 and @NBA
Awesome! Reminds me of a (preliminary) player 3d pose reconstruction project I did a few years back: youtube.com/watch?v=OUCRef…
6.6K
Akarsh Kumar
@akarshkumar0101
Sep 9, 2022
Just finished the first day of classes as a PhD student. Still feels surreal… Can’t wait to dive into research!
Akarsh Kumar
@akarshkumar0101
Jul 5, 2025
Check out this episode on @MLStreetTalk to learn about our Fractured Entangled Representations work! Very entertaining video and does a great job summarizing our findings!
Machine Learning Street Talk
@MLStreetTalk
Jul 4, 2025
AI is so smart, why are its internals 'spaghetti'? We spoke with @kenneth0stanley and @akarshkumar0101 (MIT) about their new paper: Questioning Representational Optimism in Deep Learning: The Fractured Entangled Representation Hypothesis. Co-authors: @jeffclune @joelbot3000
00:00
3.6K
Akarsh Kumar
@akarshkumar0101
Dec 7, 2023
I'll be at #NeurIPS2023! Let's chat! My interests: - Open-Ended Meta-RL - Exploration/Curriculum/Environment Design - Evolutionary ALife/Multi-Agent Emergence - Foundation Models for Decision Making - 2024 internships!
4.8K
Akarsh Kumar
@akarshkumar0101
Jan 10, 2025
Check out the dataset of lifeforms we found and play with them yourselves!
Sakana AI
@SakanaAILabs
Jan 10, 2025
Replying to @SakanaAILabs
We are releasing a dataset of the lifeforms found by ASAL from our large scale searches! You can view the datasets here: pub.sakana.ai/asal/data/illu… pub.sakana.ai/asal/data/illu… To load them up and play around with them visit the ASAL github repo or this colab notebook:
00:00
1.8K
Akarsh Kumar
@akarshkumar0101
Nov 27, 2022
I'll be at #NeurIPS2022 all week! Would love to chat with anyone doing work in Open-Endedness, AIGAs (meta-learning), exploration, curriculum learning, or RL foundation models!
Akarsh Kumar
@akarshkumar0101
Feb 8, 2022
Subbed into the game right after this picture @JCHartzell
Akarsh Kumar
@akarshkumar0101
Jul 16, 2025
Out of curiosity, are there any examples in the AI literature showing that Occam's razor is not the right fundamental principle to follow in certain natural (not contrived) circumstances? Not sure exactly what I am looking for, but maybe something along the lines of "compression
Feynman: Knowing versus Understanding
From youtube.com
1.8K