The Paradoxical Slowness of Human Behavior | Hacker News Reader

The Paradoxical Slowness of Human Behavior

caltech.edu

11 pointsbelter10 comments

Related to: https://news.ycombinator.com/item?id=42449602 "The unbearable slowness of being: Why do we live at 10 bits/s?", https://www.cell.com/neuron/abstract/S0896-6273(24)00808-0?_...

I'm having trouble believing the 10 bits/s claim. That surely can't refer to the collective output of all motor neurons, can it? Athletes and musicians are surely outputting a rate of motor data at least 4 orders of magnitude greater than that. Even people who are sedentary at a desk surely output more than 10 bits/s of motor signals.

Or perhaps the author wants to narrow it to the average rate of verbal thought creation. But even then, I can think of counterexamples.

Any insight what this is actually referring to?

As seen commented in the other thread, I feel like this metric of 10 bits per second for “thinking” could be wildly incorrect. Firstly I can “think” with my imagination – conjure up whole evolving worlds in my minds eye – this would be megabits per second easily if on a display. Secondly, a lot of my “thinking” is done semi-below my conscious radar, in that I set a task, say writing, and my mind chews on it while I do other things – for example if I’m writing an essay or book chapter, I can be mulling it over in the background while cycling or walking and doing other things, and then later sit down and churn it out. I feel like a couple of philosophers need to bang their heads together with the researcher with regards to thinking. It’s way higher than 70 bytes per minute.

The subconscious reflex of avoiding an object thrown at your eye is certainly faster than 10 bits per second.

What if those worlds are encoded so efficiently that they can be communicated in only 10 bits/sec? Maybe there's tons of detail but not very much is actually probable, so one part of your mind can "fill in the gaps" without another part "saying" much to it. Like the married couple who have been together so long they don't have to say anything and each still knows what the other means.

Please excuse my technological illiteracy. Isn’t that just a form of compression that might reduce the (uhh) space that the image takes up without actually reducing the information processing?

I don't have a great way to think about the distinction between computation and communication. It seems to me that the "bits/s" measure they're using doesn't even have the right units? That said, information seems fundamental and almost physical, whereas something like "FLOPS" feels tied to a general class of implementations.

Maybe you can imagine that some part of your mind really doesn't need a very high bitrate description of the world. For example, we read novels and play text-based adventure games, and that's sufficient to conjure a rich inner world. And at what rate do we read? Probably not more than a few bits/s, I would guess? Let's see... Google's AI says

> The entropy rate of English is between 0.6 and 1.3 bits per character.

Let's just say 1 bit per character for round numbers.

It also says,

> A typical novel page usually contains around 3,000 characters

so, 3 kbits/page. Divide by 10 bits/s, and you get 300 seconds/page = 5 minutes/page.

This seems too long to me. I think we read faster than that.

And that's ignoring that "a picture is worth a thousand words".

So, if there is a single information bottleneck into the "conscious mind", my guess is that it has a higher bitrate than the quoted 10 bits/s.

Paper on Arxiv (no paywall): [1]

No major new results. It's just an observation that, for things for which human processing bandwidth is measurable, it comes about to 10 bits/sec.

This may simply reflect the writing rate of long-term memory.

[1] https://arxiv.org/pdf/2408.10234

> This may simply reflect the writing rate of long-term memory

I don't think write rate is the bottleneck; consider performance on reaction time tests, eg: https://humanbenchmark.com/tests/reactiontime (for what would we need a LT memory write for this task?)

(if I had to take a wild guess, I'd say our scavenging and gathering ancestors never had the need of decision making with cat-like reaction times?)

I don't see how you could do what a table tennis player does in 10 bits/sec. Even if we "distil" the torrent of visual information to (e.g. simplified) heading/speed/spin estimates of the ball and do something similar for desired motor movements.

Reply on news.ycombinator.com