Lucas C Parra

@parralab.org

NeuroAI, applied neuroscience, machine learning, brain stimulation, neural signals, medical imaging - physicist by training. parralab.org

Should I still teach coding? I take pride in the biomedical signal processing class I have been teaching and refining since 2003. This is the first class where our engineering students learn to code with concrete problems to solve (60% of the grade). Should I still insist that they code themselves?

1. A bit of evolutionary biology. I'm really intrigued by a new perspective piece from Steve Frank that explores connections between how natural selection creates systems that generalize and recent work in machine learning about the surprising capabilities of massively overparameterized systems.

Generalization as the great leap in evolvability: insights from machine learning

Abstract. Natural selection encodes learned information in the genome. Learned solutions may be tuned specifically to past challenges, failing in altered e

academic.oup.com

(1/n) Thrilled to share my first paper at Meta FAIR! "EgoBabyVLM: Benchmarking Cross-Modal Learning from Naturalistic Egocentric Video Data" 👶 Human infants learn language from sparse, noisy multimodal input. Today's VLMs can't. We built a benchmark + challenge to close that gap. 🧵

Bild

Incredible! Scientists receive a rare standing ovation after announcing that a new treatment can double the survival time for pancreatic cancer, the deadliest cancer which was once considered “undruggable.” When we fund and support science, we can achieve anything! (Video: Dr. Wish Dhillon)

What are the real problems to be solved in continual learning? In my latest post, I tackle this question — reviewing where I think the field went astray in the past, how language models changed things, and where the real challenges remain. infinitefaculty.substack.com/p/what-are-t...

What are the real problems of continual learning?

Reflections on catastrophic interference, plasticity, and learning for the future in the era of large language models

infinitefaculty.substack.com

This paper changed my mind on how visual processing works in the brain. We knew it was dynamic given all the recurrent connections. But I thought it would converge to a fixpoint during a fixation. Not so, it seems ... The dynamic captures different aspects of the stimulus at different times. Nice.

Tim Kietzmann @timkietzmann.bsky.social · 3mo ago

A huge effort and a new take on what constitutes the "feedforward pass" across the visual system. Make sure to check out our new preprint:

Where, exactly, does learning happen in the brain? Out today in @nature.com, we identify a synaptic locus of birdsong learning and show that the circuit can be tuned to make birds learn faster - but at a cost. Read on👇 #neuroskyence 🧪 #prattle 💬 #bioacoustics Shareable link: rdcu.be/fiyrS

A synaptic locus of song learning - Nature

Combining a computational framework and optogenetic and chemogenetic manipulations within and downstream of the cortico-basal ganglia circuit identifies the specific cortico-basal ganglia synapse...

nature.com

"Quality is a relic now. Nobody wants it. Nobody pays for it. Nobody measures it. The metric is velocity. The metric is how fast you can ship a 3,167-line function that burns a quarter million API calls daily and call it 100% AI-written." 🔥 techtrenches.dev/p/the-snake-...

Claude Code's Source: 3,167-Line Function, Regex Sentiment

Anthropic claimed 100% of Claude Code is AI-written. A source leak exposed a 3,167-line function, regex sentiment analysis, and 250K wasted API calls daily

techtrenches.dev