Karush Suri

PhD at @UofT | Research Intern @lossfunk

[Lossfunk @lossfunk]

We graduated the 8th batch of @lossfunk residency yesterday! Amazing people, amazing presentations.

Along with this, we're announcing a pause on future batches to focus on our own research. So this was our last batch 😢

Over time residency evolved from builders to researchers, and our focus also evolved; we got feedback that 6 weeks isn't enough to get high-quality research done.

So we'll be focusing our own research agenda will full time researchers and research collaborations.

Also, we will continue build the AI x research community via open public talks that we host in our office.

[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.

[Vector Institute @VectorInst]

Ready to work on AI research that bridges academic theory and real-world application?

Applications for Vector's Research Internships are now open for Fall 2026 and Winter 2027 at https://t.co/nYCHRrXD7G

What awaits you: 🧵

[Lossfunk @lossfunk]

🚨 Submissions are now open for the Conference for AI Scientists (CAISc) 2026, co-organised by Lossfunk and @bitspilaniindia.

Submit to probe what happens when AI systems drive scientific discovery. Submissions are open until May 15!

Here is everything you need to know 🧵

[Lossfunk @lossfunk]

We are very glad to announce @AnthropicAI as the first sponsor for our Conference for AI Scientists (CAISc) 2026, co-organised by Lossfunk along with @bitspilaniindia.

Top three spotlight papers at the conference will be awarded $2,000 each worth of model credits.

This is very exciting for us because scientists across the world have been doing with amazing things with Claude and we believe these awards are a great chance for motivated teams to meaningfully multiply their research output.

Check out how senior scientists have been using Claude 🧵

[Karush Suri @karush_]

Excited to have joined the very fast-paced @lossfunk as a research intern over summer. I will be working with @paraschopra and team on tinkering reasoning models to understand their inner workings in scientific domains.

[Karush Suri @karush_]

a bit old but valuable result: Strong Functional Representation Lemma (SFRL) of (Li and El Gamal, 2017) shows that the conditional entropy decoded outputs given noise is close to the mutual information between inputs and outputs.

[Karush Suri @karush_]

“The first rule is to keep an untroubled spirit. The second is to look things in the face and know them for what they are.” - Marcus Aurelius

[Paras Chopra @paraschopra]

At @lossfunk, are hosting a fireside chat with the @AnthropicAI team this week at our office.

Focus will be uncovering their culture!

I’ll be in conversation with @katchu11, discussing what makes them tick and different from other labs!

Register below (online/offline) 👇

[Karush Suri @karush_]

"across 56 tasks, we show that the model consistently outperform sota." - 5th year phd, 2026

[Peyman Milanfar @docmilanfar]

people see the breakthrough result, but rarely see the daily grind that made it possible. you achieve so much more by being consistently productive and reliable than by being occasionally extraordinary.

[Jerry Tworek @MillionInt]

It’s the best moment in history for a small team to make a gigantic difference

[kache @yacineMTB]

you can outsource your thinking
but you cannot outsource your understanding

[Math Lady Hazel 🇦🇷 @mathladyhazel]

Contrapositives are for cowards.

[Peyman Milanfar @docmilanfar]

nothing gives you a bigger advantage than being underestimated

[Karush Suri @karush_]

finetuning open-source models teaches you tuning. but interpreting open-source models teaches you good science.

[Soumitra Shukla @soumitrashukla9]

3rd year grad students explaining their research progress

[Karush Suri @karush_]

pure sciences mould grit

[yuzu @yuzu_4ever]

one of the best things about doing a phd in physics is that you get an insane level of delusion that you can do anything in the world. there's nothing that i think is too difficult. literally put me anywhere and i'll figure it out i promise.

[Andrew M. Dai @AndrewDai]

Great research requires focus, not drama.

I've learned through the last decade that the best minds thrive when given stability and clear direction. Something I prioritized at Brain and DeepMind.

If you're looking to get back to making breakthroughs at the frontier, DM me.

[Karush Suri @karush_]

frontier models are research productivity drivers

[alz @alz_zyd_]

In an ideal world all funded PhD students should get $300/month extra funding specifically for access to frontier AI models