Sandesh Katakam

BS-MS'25 (Math Major + CS Minor) @IISER_Berhampur sometimes I care too much about the interpretability of Machine learning models.😅Generative Models, RL, MLSys

[Aaron Roth @Aaroth]

People used to be able to impress and intimidate reviewers with complicated proofs. This will change. In the age of AI inscrutable proofs are cheap. It is understandable proofs that are valuable. Opaque complexity is now it is a sign of laziness or lack of insight.

[Sandesh Katakam @sandeshkatakam]

@iScienceLuvr I’d love to join! Let me know if there is one/if you are creating one :)

[Andrew Gordon Wilson @andrewgwils]

Now and then I miss the days of doing Gaussian process research. ML/AI was a simpler world, driven by clear ideas, principles, and goals. People were there out of passion, not because it's what everyone is doing, not as a ticket to a glamorous job. Less restlessness, less fomo.

[Sandesh Katakam @sandeshkatakam]

@siddharthb_ @TurboML Interested in ML Engineer role

[Sandesh Katakam @sandeshkatakam]

@Kacccha_aam Sandesh Katakam

[Pi School @picampusschool]

Session 15 of the School of AI started yesterday! We are delighted to announce the start of Session 15 of the School of AI, our flagship programme designed to foster cutting-edge innovation and address transformative challenges in artificial intelligence. This session will feature 5 complex challenges tackled by 12 distinguished Fellows from diverse backgrounds worldwide. It is with great pride that we introduce the Fellows of Session 15: Sandesh Katakam, Prakhar Rathi (marking a significant milestone as the first Fellow to attend the School of AI for a second time), Cristiana Di Tullio, Tahreem Rasul, Hamid Rasul, Vishwa Ramkumar, Sheikh Ayatur Rahman, Giovanni Panella, Daniel Falbo, Zead O. Yousef, Atufa Shiree,n Sameer Ahmed. Our exceptional coaching team will support this cohort alongside our esteemed current coaches, @denocris, @Marcello_AI, and @thisisvij98. We welcome two new AI Coaches: Emanuele Fratocchi and Carlo Metta. This blend of experience and fresh perspectives aims to push the boundaries of AI research and application. Pi School AI Director @Seb_Bratieres and our CEO @isabelleandrieu will support the entire team. We look forward to this session's impactful contributions and invite you to follow their progress as they tackle some of AI's most compelling challenges. #PiSchoolOfAIS15 #ArtificialIntelligence #MachineLearning #DeepLearning #AIResearch

[Sathvik Redrouthu @_sathvikr]

Our @ycombinator cofounder video!

In 6 weeks, we built state-of-the-art (SOTA) chip design agents, and caught 5 bugs in a certain company's next AI chip (saving an estimated $5M)

Check us out: https://t.co/mG73u7oVQZ

[Sandesh Katakam @sandeshkatakam]

@santiviquez @rasbt would love to join! Count me in!

[Garry Tan @garrytan]

One of the craziest O1 YC hackathon entries was a service that would give you a Jupyter notebook with runnable examples for ML papers tailored to your own personal understanding of ML

You can definitely see hyper personalized instruction is inevitable yet likely pushed back decades because of those three factors

[The Nobel Prize @NobelPrize]

BREAKING NEWS
The Royal Swedish Academy of Sciences has decided to award the 2024 #NobelPrize in Physics to John J. Hopfield and Geoffrey E. Hinton “for foundational discoveries and inventions that enable machine learning with artificial neural networks.”

[Sahar Abdelnabi 🕊 @sahar_abdelnabi]

NeurIPS reviewer: solution is not new and was heavily studied.
@NeurIPSConf #NeurIPS24

[Sandesh Katakam @sandeshkatakam]

@christinetyip @gensynai @icmlconf @ESFoMo I'm interested! DMs disabled for non subscribers on twitter

[Ilya Sutskever @ilyasut]

After almost a decade, I have made the decision to leave OpenAI.  The company’s trajectory has been nothing short of miraculous, and I’m confident that OpenAI will build AGI that is both safe and beneficial under the leadership of @sama, @gdb, @miramurati and now, under the excellent research leadership of @merettm.  It was an honor and a privilege to have worked together, and I will miss everyone dearly.   So long, and thanks for everything. I am excited for what comes next — a project that is very personally meaningful to me about which I will share details in due time.

[Sandesh Katakam @sandeshkatakam]

Excited to attend Oxford Machine Learning summer School! Kicked off with MLx fundamentals online today, and can't wait for the in-person sessions on MLx Rep. learning&Gen AI, and MLx Health/Bio at Oxford University in July. Looking forward to meeting fellow participants! #OxML

[Andrej Karpathy @karpathy]

Clearly LLMs must one day run in Space

Step 1 we harden llm.c to pass the NASA code standards and style guides, certifying that the code is super safe, safe enough to run in Space.
https://t.co/n9bc9HH79a (see the linked PDF)
LLM training/inference in principle should be super safe - it is just one fixed array of floats, and a single, bounded, well-defined loop of dynamics over it. There is no need for memory to grow or shrink in undefined ways, for recursion, or anything like that.

Step 2 we've already sent messages out to Space, for possible consumption by aliens, e.g. see:

Arecibo message, beamed to space:
https://t.co/UIyOh45jfg
Voyager golden record, attached to probe:
https://t.co/fuwD59oKF6
The Three Body problem (ok bad example)

But instead of sending any fixed data, we could send the weights of an LLM packaged in the llm.c binary, with instructions for the machine code. The LLM would then "wake up" and interact with the aliens on behalf of the human race. Maybe one day we'll ourselves find LLMs of aliens out there, instead of them directly. Maybe the LLMs will find each other. We'd have to make sure the code is really good, otherwise that would be kind of embarrassing.

:) Step 2 is clearly not a serious proposal it's just fun to think about. Step 1 is a serious proposal as, clearly, LLMs must one day run in Space.

[Nicholas Mancuso @nmancuso_]

I've been toying around with stochastic trace estimation using linear operators, and ended up collecting things into an easy-to-use Python module, traceax.

It's built on top of lineax, which is a wonderful linear op library built on top JAX primitives 👇
https://t.co/vAjlXB48Wp

[Daniel Johnson @_ddjohnson]

Excited to share Penzai, a JAX research toolkit from @GoogleDeepMind for building, editing, and visualizing neural networks! Penzai makes it easy to see model internals and lets you inject custom logic anywhere.

Check it out on GitHub: https://t.co/mas2uiMqj9

[MIT CSAIL @MIT_CSAIL]

19 years ago today MIT researchers got a computer-generated gibberish paper accepted to a predatory journal: https://t.co/kEQY9GJI3Z

Generate your own here: https://t.co/mP4LW13kAk

[Peter J. Liu @peterjliu]

What do you call the disparity between GPU-rich and GPU-poor?

Jensen's inequality