← Back to context

Comment by Noumenon72

12 hours ago

A non-groady interpretation might be that we know Manet was intelligent, and the portrait captured that well; you might not be able to judge food by its photograph, but a photo that makes it look delicious when it actually is delicious has managed to capture the deliciousness.

Personally, I don't find it groady anyway; pretty sure a neural network trained on IQs and photographs could find reliable signals of intelligence therein. I can imagine a species that evolved to conceal the appearance of intelligence, but in humans I think it's more something natural selection would try to broadcast.

> pretty sure a neural network trained on IQs and photographs could find reliable signals of intelligence therein.

exactly how?

Or I should say, all humans are intelligent baring extreme circumstances (for example signs of genetic defect, potential inbreeding, or FAS may possibly be a sign of lower IQ it's not a certainty of it). Are we trying to say there is some correlation between 'high' IQ and outward appearance, because that is quite the statement, and one that really has no scientific basis that I know about.

> pretty sure a neural network trained on IQs and photographs could find reliable signals of intelligence therein

It would start judging people by their clothes. Among other factors, none of which would be intelligence.

You will find correlation, but it would not be reliable at all.

If I trained a neural network and it learned that stethoscopes are strongly correlated with IQ, have I learned anything useful?

> pretty sure a neural network trained on IQs and photographs could find reliable signals of intelligence therein

Yeah maybe, however as a life-long photographer and former freelance DoP I would heavily caution on then using those images to infer a persons intelligence from that picture. Because in my experience the number of false-positives and false-negatives is high.

There are extremely intelligent people who always look like shit on camera, because they constantly move, so you always capture them with their mouth half open, mid blink. And then you have complete hollow-brains who look deep and dashing whenever a camera is around, but god forbid they open their mouth.

So if anybody decided to make machines decide who is intelligent based on pictures (sounds like modern eugenics), the amount of false classifications would be exhorbitant and have real consequences for real people.

And let's not forget that appearances can be altered, so once you use such a system those deemed to be most intelligent will be those who game it best. So judging intelligence directly is probably the more reliable way.