Yes agreed dynamics are missing from that list. The functional and structural phenotypes he mentions are from snapshots of biology. I bet you need some measure of change in structure/morphology, activity, etc. maybe omics dynamics too. The tech might be there for some of this but not all.

Novel color via stimulation of individual photoreceptors at population scale Image display by cell-by-cell retina stimulation, enabling colors impossible to see under natural viewing.

This paper is the result of >20 yrs of hard work by talented students, postdocs and collaborators, but none more than my co-author Ren Ng and his team in EECS.
1/n
Novel color via stimulation of individual photoreceptors at population scale | Science Advances www.science.org/doi/10.1126/...

Per source, NIH has been instructed to hold off from issuing ALL awards to Harvard, Cornell, Columbia, Brown, and Northwestern (including med schools)

Agency staff have also been instructed to not speak about this funding freeze to grantees and applicants

I'm so sorry Bryan

🚨 Everyone needs to call their Dem Senator right now. They are starting to cave.

Tell them:
1. Vote NO on Cloture
AND
2. Vote NO on the Republican spending bill.

Don’t let them pivot to reconciliation. GOP doesn’t need Dem votes on that and they know it.

TODAY is the showdown.
🤳🏽: (202) 224-3121

Nice essay!

"LLMs, while they already have all of humanity's knowledge in memory, haven't generated any new knowledge by connecting previously unrelated facts...We're currently building very obedient students, not revolutionaries."

if you're at #ISCOP this week, check out Halely Balaban (and mine)'s work on the capacity limits of mental simulation.

(if you're not there, you can imagine it, but just don't imagine anything else at the same time)

Pre-print 🧠🧪
Is mechanism modeling dead in the AI era?

ML models trained to predict neural activity fail to generalize to unseen opto perturbations. But mechanism modeling can solve that.

We say "perturbation testing" is the right way to evaluate mechanisms in data-constrained models

1/8

GitHub - ViCCo-Group/thingsvision: Python package for extracting representations from state-of-the-art computer vision models Python package for extracting representations from state-of-the-art computer vision models - ViCCo-Group/thingsvision

There was a bug in our THINGSvision toolbox (github.com/ViCCo-Group/...) that affected a subset of TensorFlow/Keras models, including the harmonization models serre-lab.github.io/Harmonization/. We apologize for the issue! Pytorch models were not affected! 1/2

The 3D-PC: a benchmark for visual perspective taking in humans and machines Visual perspective taking (VPT) is the ability to perceive and reason about the perspectives of others. It is an essential feature of human intelligence, which develops over the first decade of life a...

Interesting context to your thoughts is that DNNs learn monocular depth cues from 2d images after self sup or object class training:

arxiv.org/abs/2411.17385
arxiv.org/abs/2406.04138
arxiv.org/abs/2404.08636

So perhaps it is embodied interactions per se that leads to human-like shape bias.

Very cool post. I also see "shape bias" in comp vision as being more like contour bias than what's described in developmental lit.

Different paths toward safe AI at different Marr's levels

Excited to release what we’ve been working on at Amaranth Foundation, our latest whitepaper, NeuroAI for AI safety! A detailed, ambitious roadmap for how neuroscience research can help build safer AI systems while accelerating both virtual neuroscience and neurotech. 1/N

thinking of calling this "The Illusion Illusion"

(more examples below)