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

[Excited Penguin]

 Understanding comes from being embodied in an environment.

[The Grand Nudger]

Thanks, I was trying to remember 2 and 3 but didn't want to watch that video again.

2. We might not be able to indefinitely improve our intelligent machines. The assumption he is making is that progress will speed up exponentially. He cannot make that assumption.

He acknowledges, at the very outset, that at some point we may not be able to improve our intelligent machines further, with the example of the two doors..in fact, and again he calls it the worst thing that would have ever happened to humanity.  Secondly, he does not..I can't stress this enough -does not- make any such assumption.  He specifically takes the time to explain why such an assumption is unneccessary more than once.  He notes, however that on processing speed alone, if it's possible to make an intelligent machine, it would be able to do more work than a human being in less time.  That's already an exponential increase.  That's not an assumption, that's objectively true even of the machines we have today...it's why we use them in the first place.  

We are limited by how long it takes us to measure, to understand, to experiment, to publish etc. We have so much that needs to be done to create the kind of AI that he is talking about, we cannot make assumptions about how society will look in a couple of hundred years time. Maybe we're in a golden age right now and it will be a steady decline as we don't make the transition to a cheaper and more abundant form of energy and the cheap oil runs out. Maybe we'll all be living in a theocracy or a fascist dictatorship. Maybe corporate capitalism will fail. There is plenty of reason to recognise it as being unsustainable in its current form. Maybe resource wars or a pandemic will destroy our increasingly fragile just in time society.  He's making the assumption that the society and economy we have now will be with us in the future. Even if our economic system continues smoothly on for the next few hundred years we still can't predict what the needs of that society will be.

Again, he acknowledges those things that would prevent us from improving our machines even if it were possible....right at the outset.....and you canl;t have actually watched that video and missed that, because he keeps referring to those things throughout the entirety of the presentation.

3. The space of possible intelligence is limited by your environment. The AI that we see now is statistical trickery. Google translate does not actually understand what you are saying. There is absolutely no way that it can. This is because it doesn't actually use that language that relate it back to itself. Understanding comes from being embodied in an environment.

....................that doesn't speak, at all, to the assumption you responded to with it.........are you trying to imply here that we -are- at or near the summit of intelligence..that human beings are the smartest possible thing in the available space...or that we couldn't "embody" an ai in an environment?

[abaris]

I need to relax. The assumptions he makes are staggering. I'll respond later when I have calmed down and can type properly without risk of mashing my keyboard into bits.

Well, Harris does that to you. On most every topic he feels like running his mouth on. Suffice to say, I'm not a fan.

As far as my understanding of AI goes, we are still miles away from creating something remotely deserving the name. I might have lost track and I may be wrong and maybe there's the big breakthrough in the near future. So it's rather moot to discuss it right now without knowing what it really turns out to be.

Right now it's just machines learning from and adapting to their environment.

[Edwardo Piet]

That's the thing, the assumptions he outlines are indeed very sound.

No, really, they are not.

Must. Destroy. Something. Beautiful.

Intelligence is about information processing and technology is advancing... weren't those his assumptions?

Yes, and I don't have any problems with those assumptions.

So which is it?

[Edwardo Piet]

The assumption he is making is that progress will speed up exponentially. He cannot make that assumption.

And he doesn't. In fact he says the exact opposite:

It's crucial to realize that the rate of progress doesn't matter. Any progress is enough to get us into the endzone. We don't need Moore's Law to continue; we don't need exponential progress... we just need to keep going.

Underline added.

He said this verbatim in the video. How can you say he's making the assumption of exponential improvement when he specifically says the exact opposite in the video?

[I_am_not_mafia]

What laymans assumptions would those be, for the laymen among us?

This isn't the only time that Sam Harris has talked about the dangers of AI. I have been talking about Sam Harris in general, not just his Ted Talk. I assumed that this was in his Ted Talk but he didn't mention it there.

Can We Avoid a Digital Apocalypse?

With AGI the most powerful methods (such as recursive self-improvement) are precisely those that entail the most risk.

He fundamentally does not understand that you cannot have recursive self improvement without some part of the program staying constant. Therefore it is limited. Otherwise it's like poking yourself in the stomach and saying that it's a self poking finger.

If you have the ability for the program to write over itself entirely, then it cannot improve. The first time it creates a less optimal or broken solution it won't be able to continue improving. You can only do this if you are prepared to have multiple artificial intelligences become useless the moment you deploy them. It can't know in advance that a solution will fail dismally until it tries it. Basically what he is describing is evolution.

We already have artificial evolution. There is nothing magical about this. It's just a form of parameter search. But it's an absolute essential step that requires ever longer processing time the more complex the system. For example, I can run a simple three layer biologically plausible neural network on my computer. It can work really fast and can do strong AI. But to get that I have to create a whole population that evolved over many generations. This stage takes many weeks evolving 24/7. And that's assuming that this is the final version.

[GrandizerII]

 Understanding comes from being embodied in an environment.

https://en.wikipedia.org/wiki/Embodied_cognition

[I_am_not_mafia]

It's crucial to realize that the rate of progress doesn't matter. Any progress is enough to get us into the endzone. We don't need Moore's Law to continue; we don't need exponential progress... we just need to keep going.

"We just need to keep going". Assumption right there that we will be able to.

Think of it in terms of the scale of the problem compared to the rate of progress. He is making the assumption that it will happen in 50 years time but that's because he does not understand the scale of the problem. We don't yet know what is possible. This is the very nature of research. We do know that the current form of computation won't hack it. We just don't know what the next form will be. So we can't say how long it will take. It might take hundreds or thousands of years. Who can say what society will be like in 50 years yet along hundreds? We can't predict this because we can't know in advance what our needs will be because we can't know what technology will enable us in the future.

[account_inactive]

Ermahgerd, killer robots!

[I_am_not_mafia]

The assumption he is making is that progress will speed up exponentially. He cannot make that assumption.

And he doesn't. In fact he says the exact opposite:

It's crucial to realize that the rate of progress doesn't matter. Any progress is enough to get us into the endzone. We don't need Moore's Law to continue; we don't need exponential progress... we just need to keep going.

Underline added.

He said this verbatim in the video. How can you say he's making the assumption of exponential improvement when he specifically says the exact opposite in the video?

He is making the assumption of the singularity. He hasn't specifically said it but that's what he is referring to. That's what he is describing. I know what he's referring to because other people do as well. Hence the bit he talks about replacing an office of workers and having the AI do the research. Fundamental to this is an exponential increase in progress. He claims that it's not required but then inserts it in anyway. Yet at no point does he even speculate how a disembodied computer can sit there creating new knowledge out of thin air.