Wednesday, October 20, 2010

Open Season

I don't mind off-topic posts. I sincerely can't see how they're harmful.

Indeed, I'm about to encourage them. This post has no topic. All comments are therefore offtopic. Post anyway!

Here's mine:

All comments are emailed to me, so I don't miss any.

RSS feeds are the reason I don't see any necessity to post regularly, though I will perhaps make the attempt at some point.

I can be a selfish bastard. I'd prefer not to post my email, but you can ask me to email you and I will. Spam filters are pretty good these days...but the best spam filter is just not to have a public email. On the other hand I can easily post and then delete my email, and then nobody has to have it out there publicly (for long).

I wonder if there's anyone who knows in detail the pros and cons of posting their email. It seems to just be the thing to do, without any thought either way.

7 comments:

adrian said...

I am resuming the "situated cognition" discussion here, since you made this post anyway and gave the green light to derail conversations (assuming they're somewhat interesting naturally).

Also don't forget that artificial intelligence is not the same as artificial consciousness.

Thanks for actually bringing this up. It is an interesting question as to whether intelligence and consciousness need each other or not, or whether our present language and rational tools even allow us to conceptualize the issue correctly yet.

In any case, another consideration is as to whether AI needs to even emulate NI at all. AI+ may be a different kind of superhuman intelligence, and AI++ possibly beyond any kind of meaningful speculation. I digress though, we have much more pressing problems to work out first.

Alrenous said...

The reason I bring up the distinction between AI and AC is because some AI researchers are trying to make better computers and/or robots, but other AI researchers are trying to make minds. These are different problems, and you should know explicitly which one you want to work on. A lot of confusion in AI research results from one side neglecting the differences of the other side - understandably, since they use the same words.


Perhaps as an AI designer one has more pressing problems, but as a philosopher, a working definition of intelligence is pretty important, because it's extraordinarily difficult to get anything done without precise definitions.

I realized intelligences do three things:
Learn
Reason
Create

Assimilate information, extrapolate/interpolate information, and generate information.
As far as I can tell that spans the entire space - everything the folk intuition about intelligence points to.

I don't yet know if there's a reason we normally conflate these three things.

It's interesting though that obviously very weak AIs are already here. Moreover, you're correct, there's no need for this definition when creating an AI. Just making an AI that successfully navigates a problem will invoke intelligence automatically and unavoidably. So I guess that's another strike again situated cognition.


I'm pretty sure AI shouldn't attempt to directly model NI, but not due to any definition of intelligence, which will be common to both. Rather, AI isn't implemented in neurons and shouldn't run around trying to live up to the example of neurons. The particulars of algorithmic efficiency are all different.


Unless AI+ can solve P=NP? then it's just going to be especially big or especially fast. I think the scale issues will create the qualitative phase changes AI is looking for.

After all, even humans have phase changes across the natural IQ range. I can't see any reason AIs couldn't have the same kind of thing, especially as they can in principle be built arbitrarily smart.

adrian said...

I think the scale issues will create the qualitative phase changes AI is looking for.

If I'm interpreting this correctly, this is a case of qualitative == quantitative; if {quantitative = sufficient enough}

The more I think about it, the less sure I am regarding just what the "artificial" really is. Has anyone ever defined it properly? Already we're starting to move away from silicone in components, and the lines between genetic engineering and robotics are starting to blur. Philosophy to the rescue for clarification I suppose.

By the way I got an iPad. Yes that's completely irrelevant and I'm a dick for saying it. Looking forward to your future updates. Also, I got a few ideas I'd care to send you and get a reaction on, but after a bit of polishing as their current state is rather underwhelming.

Alrenous said...

"Philosophy to the rescue for clarification I suppose"

Artificial just means made by human hands. Technically speaking if you constructed a bog-standard human from the ground up, it would be an artificial human.

Though normally there's no point in labelling things natural or artificial unless the artificial version is different from the natural version.

You're right that there's no fundamental distinction. Human hands are natural, and so jumbo jets are just as natural as mountains; jets just happen to go through a metacognitive machine as an intermediate step.

"By the way I got an iPad. Yes that's completely irrelevant and I'm a dick for saying it."

Hahahaha

"Also, I got a few ideas I'd care to send you and get a reaction on,"

And I'm looking forward to reacting to them.

Alrenous said...

"Very weak AIs are already here?" What am I smoking?

Well, there you go. Creativity is a technological problem. And, so I assert, are learning and reasoning.

Though naturally they don't know what philosophy is.

"With Emily Howell, Cope is, once again, challenging the assumptions of artists and philosophers"

Err...'challenging' is not the word I'd use.

adrian said...

AI is definitely capable of creating, and easily surpassing, modern "art" (music included). I have a hard time imagining software creating something like Mozart's Requiem Mass that, to my ears anyway, captures a timeless essence indescribable outside of the music itself. But I've also learned to be open to everything, so I'm not ruling out the possibility either. Perhaps that essence is merely an illusion and a sufficiently-evolved AI could produce a kind of harmony that generates the same sense of "perfection" in the listener.

Alrenous said...

Short tangent: Cope is wrong about creativity. Just rearranging existing music? Okay, but who came up with the bits of music that appeal to humans in the first place? A computer can never invent euphonious music. Even by the random process I allude to a few paragraphs down, eventually the music would need to undergo artificial selection by human ear.

Music is sound, which is physics. Physics is made of math. Math is never unique, it can always be repeated.

A computer could, in principle, stumble upon recreating the Mozart piece by chance. At that point you can reverse-engineer the process and start understanding how Mozart created music.

It looks to me like the issue you're having is one common to pre-Alrenous philosophy of mind. (In the article, Hofstadter wrestles with this, for example.)

The computer is not conscious. Mozart is. Even though the computer can say the same words (pieces) in the language of music as Mozart, doesn't mean that the computer is conscious like Mozart, nor that Mozart did not have a mind.

The difference is that when Mozart said those words, they actually mean that Mozart is experiencing the conscious state the words decode to. It's the difference between saying, "I'm in love," because I'm in love, and saying it because I'm slapping random phonemes together and that's what popped out this round.

But Mozart was part machine, as were we all. To encode his qualia as physical words, he needed to use that machine. All we're doing is starting to understand how that machine works.

Intriguingly it doesn't seem to even look dualist to this point, aside from my confidence that the computer isn't conscious, which is hardly unique to dualism.

So yeah, the 'essence' thing is just confused all around. There is an essence. It's not where you're looking for it.

AI can make all the music they want; they will never appreciate music. You need AC for that.

A question that fascinates me: will AC appreciate the same kind of music as us? It's probably fairly arbitrary...but what if it's not?