Elon Musk: Artificial intelligence = "summoning the demon"

[slimmouse]

We won't be designing artificial intelligence while we still don't understand human intelligence.

Its a funny thing. A thought similar to the above, echoed in a podcast I was listening to somewhere, was what initially prompted me to slightly alter the direction of this thread.

It could have been Bruce Damer, since he appears to have his finger on the pulse wrt the real state of AI, coupled with a hands on experience of the potential of our own individual consciousness, having spent most of his working life within both fields , or so it would appear.

[DrEvil]

Not sure how you think it's limited hangout.

If extinction through advanced AI is unavoidable then it certainly would be a limited hangout...

But seriously, we don't need advanced AI to be culturally consumed by a machinic telos, and distributed AI systems don't have to be that smart to be smarter than 'the mob', which is when we start listening to them, or have to - let's face it we are outsourcing our intelligence to them. (And getting the mob to worry about an elite concern may just be a clever trick of subjugation.)

Btw, any thoughts on why this artilect is assumed to be conscious? (Isn't it?) As if we just make a simulation complex enough it will therefore become so. Strikes me as a peculiarly backwards attitude akin to thinking that animals aren't conscious.

Well, our Skynet won't be.

If we can make a simulation complex enough to pass for a human, how do you know there is a difference? If it's as complex as us, why should it be any less conscious?

[Wombaticus Rex]

Because it's way fucking easier to fool a human being than build one.

[slimmouse]

Because it's way fucking easier to fool a human being than build one.

I, intend you no disrespect Dr Evil, when I say that the above analysis and reply is nothing short of outstanding. Im also reasonably optimistic that it wont be long before you agree.

[coffin_dodger]

Dr Evil said:

If we can make a simulation complex enough to pass for a human, how do you know there is a difference? If it's as complex as us, why should it be any less conscious?

An AI won't fart. Or take drugs. Or think about sex every seven seconds.

AI is, to the 2010's, what flying cars were to the 1950's.

[DrEvil]

Because it's way fucking easier to fool a human being than build one.

Fair point on the "passing as human" part, no need to be as complex as us to do that, but what if we build something that really is as complex as a human brain, and that construct can pass as a human? Is it still just a fancy simulation?

[DrEvil]

Dr Evil said:

If we can make a simulation complex enough to pass for a human, how do you know there is a difference? If it's as complex as us, why should it be any less conscious?

An AI won't fart. Or take drugs. Or think about sex every seven seconds.

AI is, to the 2010's, what flying cars were to the 1950's.

You don't have to be conscious to fart, take drugs or think about sex (a simple robot could do all of those), and flying cars already exist.

[coffin_dodger]

Dr Evil said:

If we can make a simulation complex enough to pass for a human, how do you know there is a difference? If it's as complex as us, why should it be any less conscious?

An AI won't fart. Or take drugs. Or think about sex every seven seconds.

AI is, to the 2010's, what flying cars were to the 1950's.

You don't have to be conscious to fart, take drugs or think about sex (a simple robot could do all of those), and flying cars already exist.

If you say so, then it must be true. I bow to your greater intellect on this occasion.
I had no idea that there were already machines that could think, for starters.

[Wombaticus Rex]

Because it's way fucking easier to fool a human being than build one.

Fair point on the "passing as human" part, no need to be as complex as us to do that, but what if we build something that really is as complex as a human brain, and that construct can pass as a human? Is it still just a fancy simulation?

Thanks for taking it in the spirit it was intended; it was just short since I didn't have much to add to it.

The Turing Test standard has always baffled me, since even an ELIZA level chat bot is perfectly sufficient to keep your grandma busy for 20 minutes. (Or, legend has it, RAND executives angry for a few hours.)

People panicked when they first saw movies; another simple hack with profound consequences -- and a lot to say about how low-res our perception of the Universe is, I suppose.

So my point was more about what a crudely shallow approximation Turing-style finish lines are, compared to the completely intractable problem of actually describing and accounting for our own experience as organisms.

That said: although I am an inveterate skeptic on the subject of AI as consciousness, I also expect to see some absolutely fucking amazing stuff from outfits like DeepMind in the next 10 years. (More especially from the outfits we don't know about because they aren't looking for funding, just quietly recruiting for talent.) I do expect to see amazingly accurate models of the human brain -- and I also expect that those models will force us to admit the brain isn't answering all the questions we expected it to. (Again, just like the genetic determinism ideology was undone by their own success with the HGP.)

[DrEvil]

Dr Evil said:

If we can make a simulation complex enough to pass for a human, how do you know there is a difference? If it's as complex as us, why should it be any less conscious?

An AI won't fart. Or take drugs. Or think about sex every seven seconds.

AI is, to the 2010's, what flying cars were to the 1950's.

You don't have to be conscious to fart, take drugs or think about sex (a simple robot could do all of those), and flying cars already exist.

If you say so, then it must be true. I bow to your greater intellect on this occasion.
I had no idea that there were already machines that could think, for starters.

Not 'think' in the human sense, but 'think' as in loading a pornographic image into memory every 7 seconds and doing some light processing on it.
And thinking is basically information processing, isn't it?

PS! I don't think I actually claimed to be smarter than you in that other kerfuffle, so how about getting over it?

[coffin_dodger]

Dr Evil said:

Not 'think' in the human sense, but 'think' as in loading a pornographic image into memory every 7 seconds and doing some light processing on it.
And thinking is basically information processing, isn't it?

I'm not a fan of comparing the inner workings of consciousness to the current technological level. The Victorians convinced themselves that humans were basically steam engines etc. It's (possibly unintentially) dehumanising and a bit fanatical in my opinion.

Does thinking include 'imagination' in your lexicon, or is that another phenomenon entirely?
I don't see thinking, or especially 'imagination' as basically processing information at all. In fact, comparing 'thinking' to 0's & 1's running through non-living matter (no matter how many functions it can perform per second) and calling it AI is fantasy, afaic.

[Nordic]

People are already shutting off their brains in the presence of computers. It's like when men take testosterone and their testicles wither and shrink -- because their bodies go "hey I don't need to do that any more".

I have witnessed people let their GPS take them to places that don't make the slightest bit of sense, when paying the tiniest bit of attention would have alerted them to the fact that their computerized navigation system is taking them to the completely wrong part of town, or the totally wrong way.

And soon we'll have self driving cars.

I can't wait for that because all the self driving cars will be programmed to take the exact same routes, so I, with my brain, will be able to drive around them ad if they were blind sheep, which is basically what they'll be.

People are lazy. With AI comes the world of Idiocracy. For real.

[slimmouse]

Well, here we sit in a great electromagnetically generated illusion. Thats one fact, which in itself is something of a wonder to behold when you really start to cogitate on it.

Beyond that we''re essentially a bit stumped. Except of course to say that within the aforementioned framework we know that we are incredibly complicated in our personal biological construction - complicated beyond belief itself, some might say.

In other words, when you get right down to, we really actually know fuck all about anything of real meaning, other than that we live in an illusion (albeit a very persistent one), which we ourselves bring to being, coupled with the fact that we are.

This is essentially the rather limited palate that our greatest minds currently have to work with,

Therefore, since not one person either currently on the planet, or ever born fully understands this, then good luck with creating your AI is what I might say to such minds, adding that it all sounds like something of a petty distraction to me,

In the meanwhile, I wont be holding my breath.

[Nordic]

Well that's why it's dangerous. Because it will be fake. It will be to real intelligence what aspartame is to real sugar. And the whole point is that whatever the fuck it is will have a sociopathic amorality combined with a functional Aspergers to the point where it might "accidentally" eradicate mankind as a way to feed the goldfish or some mundane task.

[justdrew]

"strong" AI is one thing, a theory of possibility. I don't think they'll ever instantiate one without first building a synthetic-brain-analog (which we're semi-close to being able to do). and when they do build that, it seems to me it may well be found to already have an "AI" in it. I think a possible outcome is that it wouldn't be a "new" person, it might be an ex-person! Who saw what was being built and decided to inhabit it. First synthetic brain turns on, and it's first words are, "you all better sit down. This is Allen Turing speaking to you..."