I'm not totally on board with saying life forms like cockroaches are intelligent in a way fundamentally different than AI. Species are a store of information for evolution as it performs optimizations based varying environments. Evolutionary algorithms in AI vary from being pretty simplistic efforts to approximate functions to being decent models of actual evolution. While I'm not a biologist, I'm pretty ok with comparing species to functions that are gradually modified over time to maximize their "fitness" or be more "optimal." So then lots of species could come close to failing your "parameters" test. Evolution has warped the hell out of all lifeforms through its parameters. Intelligence doesn't have to be free of parameters and molding. Of course, the cockroach example stuck out to me because there's more of an argument there than with humans. I'll point out lots of stuff in AI is just developing clever ways to learn certain functions (e.g., the function of detecting a human face) by gradually minimizing some error function over time. So it's hard for me to see how this broad framework puts us on a track towards some sort of "strong" conscious AI. No matter what, the consciousness aspect is blocked by computer architectures--at a low level, even good convolutional neural networks (really in vogue ones that rn are doing lots of stuff like facial recognition, voice recognition, learning how to paint like van Gogh) are just building functions that, when executed, are passing a sequence of low level instructions through the processor. Reading from memory, doing some computations, changing some registers, writing back to memory, etc., is clearly very different than the consciousness-creating processes on biological brains. But this doesn't lead me to automatically agree with your over all assertions that current programs and AI techniques aren't intelligent. At the very list, they're intelligent in a different sense, but I don't want that to sound trivial because it isn't and consciousness produced through evolutionary constraints shouldn't limit our idea of what counts as intelligent.
I think it is already possible to create AI that can program. I remember reading something about AI that can reprogram parts of itself, as well.
No it's not. By your own dick measuring standards of correctness on this topic, I am right and you are wrong.
Hygro implied we are being used by machines. Which implies machines have agency which of course they don't.
Maybe I just have a higher standard for intelligence. My cat is pretty dumb but my smart-phone is manyfold dumber.
No doubt but it's still not real AI in my view. 'Weak AI' has no intelligence in any conventional sense and the idea that the machines are controlling us as Hygro suggests is absurd. No doubt technology is changing our brains but we're willingly using it to (and if I were rich I'd use it even more so, buying all sorts of fancy neurofeedback equipment and monitoring all my biomarkers to optimize my health and brain function).
And AlphaGo beat the best human last year, it's pretty amazing but all this software is still designed by humans and if it had to think as slowly as humans it would stand no chance. These programs don't "think", they simply analyze staggering amounts of possiblites, they are calculators that's all.
When "AI" can create EDM as well as humans I'll be impressed.
I disagree.
That is a different question. There is a vague notion being thrown around that high intelligence would bring about internal being. That is the case with biological gene units, but we don't know exactly why, and we certainly don't know how to artificially recreate this phenomena. I think there is good reason to be cautious about the idea of some kind of spontaneous coming about of sentient programs.
