Can we build AI without losing control over it? | Sam Harris

[Edwardo Piet]

Well, you may have a greater understanding of something like intelligence but if the definition of intelligence is about comprehension and you're talking about things that have nothing to do with comprehension then we're not actually talking about comprehension.

That's not a useful definition in the field of artificial intelligence.

Let me demonstrate this by asking a few questions.

What does it mean to comprehend?

We know what it means to comprehend. We only need to consult the nearest dictionary for that.

An A.I. never may be able to actually understand or comprehend anything... but it is artifical intelligence. It's a simulation, it just has to behave as if it comprehends. You're only ever going to get pseudo-intelligence with an A.I.... unless it develops more human like intelligence... and even then it would be hard to distinguish 'genuine comprehension' from simulated. But that's only like the philosophical zombie problem where we have no proof that other humans are actually conscious or whether they just seem that way... and at the end... does it matter? What matters is whether the A.I. behaves as if it comprehends, or whether people behave as if they're conscious.... worrying about whether it 'genuinely comprehends' is a waste of time.... it's like worrying about if there's any 'real qualia' or 'real figment' or worrying about if there's any noumena or thing-in-itself.... at the end of the day empricism and phenonmenology is what matters (as well as logic and reason... Harris' conclusion follows from his premises, strawmanning a bunch of things he didn't say isn't really relevant... and that happens to him a lot).

[Edwardo Piet]

And only a very small portion of what the brain does is actually relevant to intelligence.

Depending on how you define intelligence.

But that doesn't help if you are trying to create artificial intelligence. This is why classical AI failed after trying to get some success after 40 years. This is why the fields of new AI / non-symbolic AI  and artificial Life came about.

Intelligence isn't just about reasoning about abstract things. It includes sensory processing, linguistics, motor co-ordination. This all requires learning and adaptation to an environment. Why wouldn't you call that intelligence?

Well it doesn't really matter whether A.I. has 'actual genuine intelligence' or not it just matters about how the field progresses.

I'm just saying you can be skillful at adapting to an enviroment without understanding anything or having any intelligence at all. A.I. is artificial intelligence. It doesn't have to be genuine intelligence. Example: The best chess players in the world by far are computers now... they are the 'smartest' players, the most 'intelligent' players that make the best moves and least mistakes... but they're not actually smart or intelligent at all. They don't understand a thing. It's all knowledge and memory rather than comprehension and understanding. Their intelligence is artificial. Does that matter? No. I was just saying that adapting in an environment isn't intelligence.

[abaris]

I'm just saying you can be skillful at adapting to an enviroment without understanding anything or having any intelligence at all. A.I. is artificial intelligence. It doesn't have to be genuine intelligence. Example: The best chess players in the world by far are computers now... they are the 'smartest' players, the most 'intelligent' players that make the best moves and least mistakes... but they're not actually smart or intelligent at all. They don't understand a thing. It's all knowledge and memory rather than comprehension and understanding. Their intelligence is artificial. Does that matter? No. I was just saying that adapting in an environment isn't intelligence.

You know why they're able to beat the best chess players? Not always, by the way, but often. Because they're able to calculate ahead with data provided by us. These computers are neither intelligent nor are they adapting. They're just using the routines programmed into them and use them in a very narrow field opf expertise.

The main problem seems to be that you neither know nor care what AI really means. You're just fascinated by the premise without looking at the basic foundations any premise needs to have any kind of validity. The reasons why Harris just gave a nice science fiction horror scenario are many. And all of them have been presented here. But you waved them away with a flick of your fingers.

I wonder if you would have just as fascinated if he had talked about alien invasion. Which would have been on similar lines.

[The Grand Nudger]

He actually did use the notion of an alien invasion, or at least contact.  It;s become obvious that Harris' critics, at least in this thread, don;t think they need to acually watch the mans clips to confidently proclaim how wrong he is about everything that;s in them, lol.

If i went back through this thread, all 20 pages of objections, and itemized them, it would be a clinic in shitty reason. I don't think we've missed a single fallacy, formal or informal. OFC Harris presented a sci fi horror scenario......it was his entire point that this particulkar sci fi horror scenario doesn't produce what he considers to be an appropriate response in people, himself included. When we think about future ai, we don't dread the potential consequence, it's too damned cool. That was his entire fucking point. H expressed the notion that we needed, in his words, a manhattan project for ai. I can't, for the life of me, figure out why any single thing he said, or he entire clip as a whole, is objectionable. All I've heard is that he's wrong about things he didn't say.

.........?

[Excited Penguin]

Rational Wiki used as a source? End the thread here. You might as well use Conservapedia. The page even has a warning about itself: "This page contains rather too many unsourced statements, and needs to be improved."

The statement Alasdair quoted is one of those unsourced statements.

It's still true. You have the means to find out for yourself.

[Edwardo Piet]

The main problem seems to be that you neither know nor care what AI really means.

I do. But as I said it's irrelevant. Whether superintelligence is 'real intelligence' or not is irrelevant to whether it's superintelligent. My point was that intelligence itself is about comprehension and adapting in an environment doesn't make something intelligent. You can adapt skillfully without adapting intelligently... just like dogs can smell well without it being anything to do with smarts.

You're just fascinated by the premise without looking at the basic foundations any premise needs to have any kind of validity.

Premises aren't about validity. Premises are about soundness, as is the argument is a whole, the argument in between is about validity.

You can have a valid argument based on both sound or unsound premises.

The reasons why Harris just gave a nice science fiction horror scenario are many. And all of them have been presented here. But you waved them away with a flick of your fingers.

It's irrelevant. It's a strawman of his actual argument.

If I lay out an explicit argument and then talk about implicit hypotehtical examples and say they're not necessary, to attack my unnecessary implicit hypothetical examples and pretend they're necessary instead even when I've explicitly said they're not instead of dealing with my explicit argument and to misrepresent your objections against my implicit hypothetical examples as objections against my explicit argument when you haven't even addressed my explicit argument is to strawman my explicit argument.

I wonder if you would have just as fascinated if he had talked about alien invasion. Which would have been on similar lines.

This is all strawmanning.