Alex Lewandowski

PhD candidate studying learning algorithms @uAlberta, @AmiiThinks, @rlai_lab

[Bryan Chan @chanpyb]

Why do we assume an RL agent can always compute the correct action immediately? Every policy is a program and thus resource bounded. In this blog, I argue why computation should be part of decision making, illustrated with some toy examples.

🔗https://t.co/tBBRc6K0CX

[Louis Kirsch @LouisKirschAI]

After automating AI research with @SchmidhuberAI and building AI Scientists at DeepMind, now comes the real experiment: the institution itself.

Excited to co-found @inherent_labs: the recursively self-improving lab for scientific AI.
https://t.co/SQjUduaG3D

[Richard Sutton @RichardSSutton]

I am definitely going to this...

[RL in Big Worlds @rlc_bigworlds]

RL in Big Worlds is a workshop at @RL_Conference about ideas that enable agents to achieve goals in environments vastly more complex than themselves

This requires giving agents the ability to learn continually and use approximate value functions, models and policies effectively

[Alex Lewandowski @axlewandowski]

@DimitrisPapail Love this work! One way to avoid training: LMs are already universal computers if you prompt them to simulate Lag systems: https://t.co/LQYeD0uvgr

My followup shows even untrained LMs can do this, suggesting programmability is key to language computers: https://t.co/YhEKzyQYwo

[sorina @robot_in_space2]

We organized an RL competition during the first Openmind Research Institute Winter School in Malaysia. The participants were able to implement SARSA and SAC in just 2 days onboard our Embodied MuJoCo Ant! 🎉

[Michał Bortkiewicz @m_bortkiewicz]

Great week at #NeurIPSanDiego packed, intense, and genuinely inspiring. Grateful for all the discussions and feedback. Now looking forward to some quieter days and cooking up new stuff 🚀

[Erin Grant @ermgrant]

Thrilled to announce I'll start in 2026 as faculty in Psych & CS @UAlberta + @AmiiThinks Fellow!! 🥳 Recruiting students to develop theories of cognition in natural and artificial systems 🤖💭🧠. Find me at #NeurIPS2025 workshops (talk at @CogInterp & organising @DataOnBrainMind)

[Alex Lewandowski @axlewandowski]

At NeurIPS and interested in continual learning? Stop by our spotlight poster this Friday @ 11am (#508).

Our work provides a computational approach to the big world hypothesis through embedded agents.

Feel free to reach out if you want to meet up while in San Diego!

[Alex Lewandowski @axlewandowski]

As a reviewer: I spent multiple days replying to Authors' rebuttals. I cannot overstate how eerily similar some of it felt to pointing out bugs to an LLM. I proceeded on the assumption that my critique served the scientific record; it is disheartening to hear otherwise.

[Delip Rao e/σ @deliprao]

Hey @iclr_conf, reverting scores is unnecessary punishment for the majority of the authors who had nothing to do with this incident and had successful rebuttals. Instead of detecting collusions on your end (you have a ton of metadata) why is this everyone’s burden to bear?

[Alex Lewandowski @axlewandowski]

Language models can make convenient writing partners, if you know how to use them. But one disadvantage of their increased use in peer-review is a collapse in thought diversity. Instead of hearing from thousands of different minds, we get thousands of samples from a few.

[Alex Lewandowski @axlewandowski]

The link to your paper belongs in 1/n, not n/n.

The paper is the product. Algorithms be damned.

[Alex Lewandowski @axlewandowski]

@tw_killian I have the opposite problem: everything is RL

[Alex Lewandowski @axlewandowski]

@jsuarez Yeah, it's not RL. It's offline RL. Calling it supervised learning is equally misleading: the targets in offline RL provide only feedback on a chosen action. In supervised learning, the target provides full feedback. Also, interaction is needed for evaluation in offline RL.

[Prabhat Nagarajan @prabhatmn]

Have people seen this prescient 2001 post by @RichardSSutton on self-verification?
"An AI system can create and maintain knowledge only to the extent that it can verify that knowledge itself".
This sentiment underpins much LLM reasoning research today.
https://t.co/gK3DOwqYm8

[Finding The Frame Workshop @RLFrameWorkshop]

Thanks to everyone who joined us for another great workshop! 🥳
This year we once again asked our panelists to share a paper or book that heavily influenced their perspective 📚 check out their recommendations here! https://t.co/ZVfbTWk0lg

[Mateusz Ostaszewski @MatOstasze]

🚀 Excited to announce our paper "Balancing Expressivity and Robustness: Constrained Rational Activations for RL" will be an *oral* at #CoLLAs2025!

We study how trainable rational activations boost expressivity in RL but can also harm stability:

[Alex Lewandowski @axlewandowski]

I will be presenting a poster at RLDM on Wednesday @ 4:30pm. We show that embedding an agent implicitly constrains both it and the environment. We use this constraint to characterize continual adaptation. https://t.co/3eS6qDCD2T

Feel free to reach out if you're attending RLDM!

[Finding The Frame Workshop @RLFrameWorkshop]

🚨 We extended Finding the Frame's submission deadline to June 15 AoE! 🚨

✨We're looking for bold ideas that rethink the foundations of RL: goals, values, rewards, formalisms, and beyond🚀

More details: https://t.co/mNy0SGJtRu

See you @RL_Conference !

[Finding The Frame Workshop @RLFrameWorkshop]

🚨 Reminder! Submissions for @RL_Conference's Finding the Frame are due May 30 (AoE)!
We're looking for bold ideas that rethink the foundations of RL: goals, values, rewards, formalisms, and beyond.
🧠Philosophy, theory, critique welcome!
🔗More details: https://t.co/mNy0SGJtRu