Worth commenting this because actually true without even caveats:
I’ve wanted to write about creativity and AI for a while, but let’s just slightly dip here.
It’s all quite correct in its simplification. The idea that the weather is not controlled by us, so if we verify an AI prediction being actually true, against the real world, as a form of science, this should give us a sort of proof that the AI works at a more fundamental level.
As with all AI problems, all the confusion and weirdness happens because we get the overall frame in reverse. It’s not simply that the way we predict the weather, and the way AI works, are all based on heuristics. It’s that human consciousness is all based on heuristics.
In this case this sort of evidence says that deep learning is not more fundamental than physics, but more fundamental than US.
Because even when looking at the weather, we are the filters of perception, the lens through which we observe. Of course the weather is “out there”, so it’s not a perception problem, as it can be verified. But our canons of prediction are wholly ours. We filter what’s relevant from what is not, and what we judge an accurate prediction. In the end it’s all filtered and biased by some human arbitrary canon.
It’s the same with the problem of creativity and copyright. We can all see that AI doesn’t invent anything: it simply regurgitates what it has been fed. It’s digital vomit, coming at a great cost. But the fundamental aspect that makes it look interesting and good to us, is the illusion of removing origins. As long you have the source, training material on one side, and output on the other, you see that AI has done nothing worthwhile. But the amount of data it gobbles up, and the complexity of algorithms mean that AI is GOOD at HIDING ITS TRACK. Obfuscating the trail between origin and end. To create the ILLUSION of being TRACKLESS.
What happens when you remove those tracks from perception? That the origins of things are gone, and so we attribute them ON THE SPOT, to the thing itself. If there’s no direct observable origin, then AI created it. AI “made” something new.
The trick to understand AI is not about figuring out its limits, to realize that there’s no “real” creativity there matching human beings. But to understand that we human beings don’t have the creativity we think we have. That we also are heuristic machines compressing data and re-elaborating something that is fed to us. What comes in comes out, just the same. The reason why we BELIEVE we have REAL CREATIVITY is because of how much we SUCK, through limited introspection, at tracking the origin of our own thoughts. We have no data on the data we use and we become, therefore we think OURSELVES origins. We think something special and magical happened inside us. That we became the ORIGIN of something. That we made it.
Same as AIs, we are heuristics machines that don’t have the creativity we think we have. And where that “feel” is itself born not out of an unique, special QUALITY, but it’s born of a FAULT: We are incompetent at tracking ourselves, therefore think ourselves SPECIAL. We celebrate incompetence. We make virtues out of sins.
AIs are simply teaching us we aren’t as smart and special as we think we are.
AIs still seem not advanced enough to match us, because they still need to become more stupid. Only then AIs will become truly human.
(well, then there’s the little problem of efficiency and energy consumption)