Lossfunk

Foundational questions on artificial and biological intelligences

[Lossfunk @lossfunk]

🦠🐵🙋🤖

[Lossfunk @lossfunk]

@Avg_sapient @paraschopra @AWSCloudIndia @kvfrans @danijarh @svlevine @pabbeel @jo_brandstetter @phillip_lippe @tungnd_13 10/ We release:

Paper - https://t.co/e6UoCcsjJQ

Website - https://t.co/TpRNHJFqTG

Code - https://t.co/SNIGudv6Xh

Dataset - https://t.co/QR0SUuHwS1

[Lossfunk @lossfunk]

🚨 New Paper

Hybrid Neural World Models.

We build neural world models for discontinuous environments and our method discovers when they're making error, with no ground truth.

Our neural models run physics fast, leaping many simulator steps in a single pass.

How we find points where not to trust it: ask the model for the same future two ways, and the disagreement between the two answers marks where it's failing without needing labels.

The same recipe holds on three systems with nothing in common: a chemical wave, compressible gas with shocks, rigid bodies colliding in 3D. 👇

[Lossfunk @lossfunk]

9/ Hybrid Neural World Models.

by @Avg_sapient and @paraschopra at @lossfunk.

Thanks @awscloudindia for generous credits for Lossfunk which made this research possible

Pinging @kvfrans, @danijarh, @svlevine, @pabbeel, @jo_brandstetter, @phillip_lippe, @tungnd_13, @adityagrover_, @AnimaAnandkumar,@zongyili_nyu , @ml_angelopoulos, and @msalbergo since this might be interesting/relevant to your work.

[Lossfunk @lossfunk]

3. Days. Remaining.

Let's go 🤖

[Dhruv Trehan @dhruvtrehan9]

⏰ 3 days left until the deadline for the Conference for AI Scientists (CAISc) 2026.

With awards incl. model and compute credits over $20,000, this is a great chance to get feedback on ongoing research.

Our ask - use an LLM or AI system in some part of your research workflow.

[Lossfunk @lossfunk]

Watch the session if you missed it 👀

[Paras Chopra @paraschopra]

Video and slides from my session on automating research are live 👇

[Lossfunk @lossfunk]

@rao2z @fchollet @drmichaellevin @MiTiBennett @eschwitz @jzl86 @Meaningness @MelMitchell1 @Plinz 11/ The broader claim: AI metaphysics is not useless.

But it becomes useful when it helps inspire empirical progress: new experiments, benchmarks, interventions, or better predictions.

Full paper: https://t.co/3nzTlaU0My

[Lossfunk @lossfunk]

🚨 Paper release

Accepted at ICML 2026 main conference!

AI is full of metaphysically loaded questions:
• Can AIs be conscious?
• Do LLMs understand?
• Do models have goals?
• What is AGI?

Our position: don’t judge such concepts by whether they capture the "true essence" of something.

Judge them by what they help us do.

[Lossfunk @lossfunk]

10/ Pinging researchers who inspired this paper.

@Meaningness - your essays on misuse of metaphysics kickstarted this.

@MelMitchell1 - we were hugely inspired by nuances you tease out in “Why AI Is Harder Than We Think”

@Plinz - we love how you operationalize AI consciousness as a specific testable hypothesis. Curious what you think about our paper.