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

[The Grand Nudger]

I need to start a thread called "Can Sam Harris talk about AI without me losing control?"

I've ranted on the thinking atheist about Sam Harris. He's fear mongering and is talking out of his arse on this one I'm afraid Alisdair. At the same time destroying all credibility of the field when it's difficult enough to get any kind of work in it as it is compared to other scientific fields. He clearly has no practical experience of Artificial Intelligence and makes layman assumptions that demonstrate this.

What laymans assumptions would those be, for the laymen among us? He layed out three explicitly, are those the ones you take objection to? He seems to have a pretty positive view of the value of ai, even if he has concerns, he repeats a line more than once about it....and the theme of the talk is actually our -shared- inability to mount what he thinks is an appropriate response.

I won't repeat myself but in a nutshell what he's talking about is so far in the future that he could just as easily be standing up there warning of the dangers of space travel because one day we might invent some kind of inter-galactic drive, meet aliens and get destroyed.

Didn't he address this sort of comment, specifically, in the talk?  Does it matter how far in the future it might be, in your opinion, and if so, why?

It's mere speculation or fantasy, because we just don't know what kind of technology would allow us to achieve strong AI of the level he is talking about, so we can't say what form the AI would take. As I've said elsewhere, a quantum computer will give us completely different AI to a biological or DNA computer.

I don't think what type of technology might allow us to achieve the effect bears any relevance to his opinion as stated.  Will those different results, from different types of computers, be qualitatively different in the context of his concerns?  Not quantitatively different, because that's a non-issue.  He doesn;t seem concerned wih what it;s made out of, but what it will or might do, in addition to what -we- might do in response to it.  Is a quantum AI more or less potentially dangerous than a biological one, and why...what's the relevance?

Another more relevant example, it is as equally far fetched to stand up there and say that we shouldn't be researching neuroscience because we might edit ourselves to extinction. Yet he doesn't warn about this. After all, the knowledge that would be required would be the same for us to create the kind of AI that he's talking about. Neuroscience is a top down field which tells us what the brain does, computational neuroscience tells us how it does it, artificial intelligence tells us why it does it. All three are necessary scientific endeavours for understanding ourselves, you can't have one without the other, so as a neuroscientist he should really shut the fuck up or admit to being a total hypocrite.

You seem to be importing some other sort of objection here.  In the talk linked, that I just watched, he describes us halting our research and improvement as the worst thing that would have ever happened to humanity.  He's clearly not advocating for anything even remotely -like- what you've described above.....at least not in that video? He's expressing caution over what is very easily argued to be pandora's box.

Additionally, if you're going to talk about hypocrisy, does i make sense to confidently object by means of some proclamation of how far in the future something is, or his lack of knowledge? Am I missing something, do you know what it;s going to take to make the sort of ai he's describing (or even what it takes to produce intelligence in a human brian?), do you have a crystal ball or a newspaper clipping from the future?

Obviously, I'm not a fan, I don't know if this was just the last straw for you..or what he's said in the past...but the argument presented is pretty damn solid, insomuch as it follows and the assumptions he expresses can be ascribed at least some measure of soundness? It is a bit ludicrous to imagine that we could control something that could out-think us, and not just a little, a lot. Especially seeing as how it's intelligence that we use to control in the first place...and regardless of the timescale involved, it's equally ludicrous to only start to worry about a problem when you have it staring you in the face. That's what -I- got from the talk. Honestly, it comes off as a bit of a deepity, a "no-shit" kind of statement on the face of it.

History is full of impressive failures where people thought they had more time, or that a specific problem would likely never materialize, or be so distant as to be a non-issue, don't you think?