Post

Whaddya mean, seems we understand brains perfectly: 1. Neurons and synapses 2. Something networky emergenty sorta deal 3. See?

And of course we all know that children can't speak until everything ever written has been read to them.

okay lmao i am genuinely unsure what this means, and i'm actually curious what you are getting at

LLMs only become functional after training on an enormous corpus of text, whereas human language capabilities develop with far less of such information- the term IIRC is "poverty of the stimulus". So we know that the process by which LLMs produce output when prompted is *not* the process humans use.

Yep, it's an enormous oversimplification, actively detrimental to people's understanding of the tech, and contributes to the alarming growth of science denialism. Deep learning neural network models do, to a *very* limited extent, work like a human brain, but definitely NOT the other way around.

I wouldn’t say “work like a brain”, it’s more a model that simulates similar outputs for the same inputs (it’s not really based on actual biological neurons, just a more or less convenient representation of what they _might_ do).

Somehow I suspect this isn't the first time humans have thought the brain works just like some recently developed technology