Why do varied DNN designs yield equally good models of human vision? Our preprint with Mick Bonner shows that diverse DNNs represent images with a shared set of latent dimensions, and these shared dimensions turn out to also be the most brain-aligned. arxiv.org/abs/2408.12804

Last week The Atlantic featured an article on the rising popularity of race/IQ science on the right (theatlantic.com/technology/arc…). The obvious point that "intelligence is not like height" sparked an unusual amount of whinging. I wrote about how this is now more true than ever. 🧵

“somnium de reductione scientiae ad geometriam” The dream of reducing science to geometry? supposedly attributable to Descartes. Where is this from? Google’s failing me. Maybe it’s the Latin or an inaccurate translation.

The half-life of code is an interesting predictor of project quality. Linux, has one of the longest code half-life’s at 6.6 years. WordPress, less than 2. Every software change induces some risk. Repos with numerous "change bursts" have the highest incidence of defects.

Gary Marcus Thomas G. Dietterich Murthy Lab Ah yes, what do computer scientists need to know about cell types? In five words or less?

i have this theory that many things in life are easier in hard mode. There’s easy mode, then a medium mode where everything SUCKS and you are dying in grueling tedium, and the real way to go is to skip right over the medium part of the bell curve and go

Kahneman and Tverskys work is some of my favorite from experimental psych. Haven’t read Kahneman’s book but systematic power analysis apparently suggests only half the results cited in Thinking Fast and Slow are likely to replicate. 😢 See quoted posts’ thread for link.

Rarely categorical, always high-dimensional: how the neural code changes along the cortical hierarchy biorxiv.org/content/10.110…

Now there’s an interesting take 🤔

A remarkable post about declining global literacy rates. Article cites habit nurtured by social media to confirm rather than challenge your views as cause. Replies miss the point and use it as an excuse to confirm amerocentric political views, unwittingly illustrating the point.

This was a fun read that reminded me of the obviously innate nature of certain complex behaviors. They explain the genetic basis of fully functional circuits, and challenge simplistic cross applications of artificial neural network learning principles to brain development.

Human control of neural interfaces is more effective when the intrinsic neural manifold is accounted for. biorxiv.org/content/10.110… First demoed in macaque by Aaron Batista now translated to humans by Erica Busch. Gotta pay attn to those animal population code manifold studies!

“A 25 percent reduction in public R&D spending would reduce GDP by … [an effect] comparable to the decline in GDP during the Great Recession.” tinyurl.com/vweadewh Meanwhile the government is proposing a 44% cut to the NIH science.org/content/articl…