Hacker News Reader

Comment by monocasa | Hacker News Reader

There's strong implications that we as humans have dedicated wetware hardware for streaming out speech from the data structures in our mind. That hardware seems to be bandwidth constrained to about 39 bits per second.

https://www.sciencemag.org/news/2019/09/human-speech-may-hav...

It makes sense that other animals may be able to cargo cult some of the underlying mechanical representations of speech, but would lack the neural hardware that propelled us to dominance on the incredible short time spans of human history and prehistory.

View on HN

Isn't it the absence of hands that has held them back? (from dominance)

Having hands plus less of or less good neural hardware means less dominance, as we see with our simian cousins, yet having the greatest brain on the planet wouldn't provide dominance without the ability to fabricate things.

Bears and raccoons have very dexterous hands.

Interesting, but they also don’t have the brain of a dolphin. We’re the ones who happened to get both.

I see no reason for that, it seems more like a pragmatic choice of reliability over capacity. (in other words to make speech comprehensible in noise)

There are other studies that show we can understand speech at a much higher rate. The critical bandwidth constraint seems to be on stream out, not stream in.

In good listening conditions though. Not in listening conditions in which speech needs to be regularly understood. Rap God is nearly 10 syllables per second at some point and a few more recent songs go even a big higher, so that doesn't seem to be the limit either, only that people don't normally need to learn to talk faster than they do.

I mean, Rap God is an example of the upper limit of stream out that took a lot of practice on his part and is trying to prove that he's one of top people in the world at that art, and I can understand it just fine.

It also doesn't seem to be achievable with on the fly statements, only memorized oration. Notice how my citation interestingly shows that the 39 bits per second isn't a count of syllables, but of semantic information. Some languages have faster, simpler syllables, but the overall information bandwidth flow ends up being the same. You probably can bypass that 39 bit with practice and memorization, essentially applying higher level tags to large bodies of what you're saying via practicing repeated phrases.

That seems to prove where the bandwidth constraint is, on the stream out side.

Reply on news.ycombinator.com