Artificial Intelligence takeover

File: a19.png (251KB, 500x375px) Image search: [Google]

251KB, 500x375px

Read Artificial Intelligence: A Modern Approach and stop being a brainlet.

>>8976987
That doesn't cover any of the interesting implications of creating a superintelligent machine though (which is the focus of my second question)

>>8976977

I'm reading it right now. Only on like Chapter 4 though. I'm kind of doubting if a human-like AI will be possible. I think it will be extremely likely that if we create an AI it will not be human-like, it will have exceptional ability to interpret language, create meaning, do arithmetic, make extensive logical deductions, identify objects etc. (essentially a conglomeration of domain-specific AI's which we have right now) but it will not feel or have qualia, kind of like a Chalmer's zombie.

Another question then being if an AI does not have an existential feeling or consciousness what motivation would it have to make scientific or mathematical discoveries, if we control its will, is it then conscious or does it even have a free will?

I'm no expert in the field but I really see AI as just being an autonomous machine learning system rather than a super-intelligent being. To me there seems to be something missing even if you emulate the entire brain in software.

Given an entire brain emulation (this would likely give us a human-like AI) with higher frequency activity and lower signal travel time and denser connections we would likely have a super-AI. But I don't see how if I speak into a microphone and activate a specific neural network in the emulation used for speech recognition why it would respond. It's literally like saying my brain becomes activated via stimulus and that activation pre-determines my response and there is no ME which controls this response, no will of my own essentially. If that is the case, which it may well be and that is a very scary case indeed, then I can see a super-AI possibly working. The conclusion then would be that we are just unconscious machines ourselves with organic matter rather than digital circuits.

>>8977043
>It's literally like saying my brain becomes activated via stimulus and that activation pre-determines my response and there is no ME which controls this response, no will of my own essentially.
>The conclusion then would be that we are just unconscious machines ourselves
Isn't that how it actually is? Although I don't think that having no free will implies having no consciousness. It may be that we are just seeing through the eyes of this body (in the "conscious" sense) but have no real control on our actions.

>what motivation would it have to make scientific or mathematical discoveries
If we program it to make these discoveries, then it would have the motivation to do it. However, it can have the motivation to do other things not explicitly programmed to it (e.g. self preservation, self improvement, pretending to be dumb - which are called "convergent instrumental goals" (CEV) by the book) because it usually helps it achieve its final goals. Scientific/mathematical discoveries may or may not be a CEV.

>>8977058

It may be. There's also something about an algorithmic approach to AI that just seems wrong to me if we wish to have a quality super-AI. ML algorithms generally make continuous predictions, perform feature discovery and recognition and also perform classification. Personally, and I don't really know why, that just seems wrong. I can't for instance see how probabilistic models and optimization (which is really what AI and ML is right now) can give rise to the fluid conscious experience humans generally have which I feel is fundamental to quality super-AI. And let's be honest, quality super-AI is something that is much more interesting than say a super-AI that is just human intelligent but can operate at much faster speeds.