Thanks to @[email protected] for the links!
Here’s a link to Caltech’s press release: https://www.caltech.edu/about/news/thinking-slowly-the-paradoxical-slowness-of-human-behavior
Here’s a link to the actual paper (paywall): https://www.cell.com/neuron/abstract/S0896-6273(24)00808-0
Here’s a link to a preprint: https://arxiv.org/abs/2408.10234
I could believe that we take 10 decisions based on pre-learned information per second, but we must be able to ingest new information at a much quicker rate.
I mean: look at an image for a second. Can you only remember 10 things about it?
It’s hard to speculate on such a short and undoubtedly watered down, press summary. You’d have to read the paper to get the full nuance.
The paper actually talks about the winners of memory championships (memorizing random strings of numbers or the precise order of a random arrangement of a 52-card deck). The winners tend to have to study the information for an amount of time roughly equivalent to 10 bits per second.
It even talks about the guy who was given a 45 minute helicopter ride over Rome and asked to draw the buildings from memory. He made certain mistakes, showing that he essentially memorized the positions and architectural styles of 1000 buildings chosen out of 1000 possibilities, for an effective bit rate of 4 bits/s.
That experience suggests that we may compress our knowledge by taking shortcuts, some of which are inaccurate. It’s much easier to memorize details in a picture where everything looks normal, than it is to memorize details about a random assortment of shapes and colors.
So even if I can name 10 things about a picture, it might be that those 10 things aren’t sufficiently independent from one another to represent 10 bits of entropy.
I didn’t see any point in linking to the paper since you have to pay to access it, but here you go if you want to do that:
https://www.cell.com/neuron/abstract/S0896-6273(24)00808-0
Send the DOI to a science nexus search bot on Telegram and you’ll get the paper.
10 bits means 2^(10) = 1024 different things can be encoded.
Yes thank you. I know how binary works!
I was responding to “Look at an image for a second. Can you only remember 10 things about it?” I didn’t think that was a fair characterization. I see you probably specifically meant 10 yes/no questions about an image, but I don’t think yes/no questions are a fair proxy for “things”.
In any case you can read the preprint here https://arxiv.org/abs/2408.10234v2 and they make it immediately clear that 10 bits/s is an order-of-magnitude estimate, and also specifically list (for example) object recognition at 30-50 bits/s.
Sure, that’s what I meant with the nuance missing from the press release.
It depends what/how is being encoded in those 10 bits.
A decision tree was just one example.
Thanks for the link.