←
Apoorva Bhandari1,732 posts
Apoorva Bhandari avatar

Apoorva Bhandari

@apaxon

Forever sensemaking. Cognitive neuroscientist at Brown U

◷ Joined September 2009▧ 37 Photos and videos
Simone Azeglio avatar
⟳ Apoorva Bhandari retweeted
Simone Azeglio@simoneazeglio·Nov 17

Representational geometry describes what a neural population encodes. But it can be built in many ways, each with its own assumptions. In our new preprint, we derive one from a few simple requirements. Its distances turn out to be exactly tied to mutual information. 🧵 1/13

Post attachment
Apoorva Bhandari avatar
Apoorva Bhandari@apaxon·Oct 30

Excited to see this bet scale. Still lots to do to build even better student models that incorporate the full range of student beliefs and trajectories, but this is a really exciting development.

Introducing Ampersand Tutor (&Tutor) - we trained a model to learn like people, then trained &Tutor to teach it On a widely-used tutor model eval, &Tutor performed best, showing how better models of people can help build AI for humans Available through our API - try it!

Mohamady El-Gaby avatar
⟳ Apoorva Bhandari retweeted
Mohamady El-Gaby@GabyMohamady·Oct 26

World models are central to intelligence - but how do we actually use them to behave flexibly? We review emerging findings on how the PFC’s inner dynamics translate structured knowledge into action. osf.io/preprints/psyarxiv/j7… 🧵 1/15

Post attachment
Apoorva Bhandari avatar
Apoorva Bhandari@apaxon·Oct 8

There is precedent in the figure of Srinivasa Ramanujan whose math appeared just as alien to Hardy & Littlewood. In what reads like a prescient article, Stephen Wolfram argued that he was an 'experimental mathematician'. His words seem to apply to AI writings.stephenwolfram.com/…

Post attachment

Some early first-hand accounts of encountering a narrow superhuman intelligence as mathematicians grapple with the hundreds of big AI proofs released by OpenAI. Problems solved in inhuman ways that make us wonder what it means to actually know things... scottaaronson.blog/?p=10169

Post attachmentPost attachment
. avatar
⟳ Apoorva Bhandari retweeted
.@Radical_Amit·Sep 23

Do you know why the pain of Sahil Wakode resonates with Dalits across India? Because a part of the harassment and bullying that Sahil faced, which ultimately led to his suicide, is something many of us have experienced at some point in our lives. When I reached Allahabad for JEE preparation and went looking for a room to stay in, the first thing I was asked in almost every household I visited was my CASTE. The more households I visited, the more traumatic the search for a room became. Then my friend told me to lie about my caste. I did. I lied about my caste. It was the first lie I told after moving from my village to the city. And that experience alone was enough to make me see the reality of the society around me without any delusions about it.

Laura Driscoll avatar
⟳ Apoorva Bhandari retweeted
Laura Driscoll@lndriscoll·Aug 8

We describe compositionality as a continuum defined by 1) expressiveness of modules : what can the components do? and 2) complexity of syntax : what are the systematic rules that combine components?

Post attachment
Li Ye avatar
⟳ Apoorva Bhandari retweeted
Li Ye@liye_tsri·Jul 2

Our new preprint is out! Learning Shapes the Energy Cost of Neural Tasks biorxiv.org/content/10.64898… The human brain runs on ~20 watts. Computers and modern AI systems require vastly higher energy. How does the brain compute so efficiently? In our new study, we take a critical step toward directly measuring and understanding this. Using in vivo glucose sensors in behaving mice, we asked: what is the neuronal fuel cost of performing a learned task? Key finding: After learning, the same task required less fuel, suggesting learning may not only improve performance—it may lower the energy cost of computation. We propose an “energy minimization” view of learning. Neural circuits may optimize not only activity patterns or behavior but also energy cost. Understanding this could be critical for explaining the remarkable efficiency of biological intelligence.

Ido Aizenbud avatar
⟳ Apoorva Bhandari retweeted
Ido Aizenbud@IdoAizenbud·Jun 11

What can a neuron compute? Real biological neurons are complex, but how capable are they? Using a new method, we found that a single cortical neuron can classify cats vs dogs, recognize spoken words, and solve 10-bit parity, all tasks thought to require entire networks. (1/15)

Post attachment
Neuron avatar
⟳ Apoorva Bhandari retweeted
Neuron@NeuroCellPress·May 18

Online now: Precision fMRI reveals densely interdigitated network patches with conserved motifs in the lateral prefrontal cortex dlvr.it/TSbkcr

Post attachment