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

[Tiberius]

Here's one of his blog posts: https://www.samharris.org/blog/item/in-d...-profiling. 

I have reread that blog entry just for this occasion. I don't see how you interpreted that as a call for racial profiling, short of confusing Islam with a race. Would you care to provide me with your reasoning in this regard ?

It's this line in particular: "We should profile Muslims, or anyone who looks like he or she could conceivably be Muslim".

I agree, Islam isn't a race. So then, what do Muslims look like exactly? Sam Harris thinks we should profile people who looks like they could conceivably be Muslim. Now, one could argue that he meant people wearing traditional Muslims dress, however, in a follow up post, he states this:

"Within a few hours of publishing “In Defense of Profiling,” I had lunch with Ayaan Hirsi Ali, one of my favorite people on earth. (Of course, I told her that I thought she should be profiled at the airport, and we had a good laugh about my “racism.”)"

Ayaan Hirsi Ali is an atheist, an ex-Muslim. So she would not be found wearing Muslim clothing. Still, Harris thinks she should be profiled at the airport. Why, he doesn't say, but it seems pretty clear that he's using the term "Muslim" as a placeholder for anyone who looks African/Middle Eastern.

[Edwardo Piet]

I'm not sure what's going on here, and maybe Harris doesn't know his shit when it comes to AI, but the people saying that just keep -agreeing- with the very points he actually raised...........what gives?

QFT.

His conclusion follows from his premises. If AI keeps getting better and we keep going and AI becomes able to think for itself at a level far smarter then our own then it certainly could easily be very dangerous.

I don't see how it makes him a hypocrite. He gave his 3 assumptions and maybe he's completely wrong about them, but if those assumptions are true then his argument follows.

Like he said, the improvement doesn't need to be exponential.

@Tibby

I don't remember Harris ever talking of racial profiling but I do remember him talking about religious profiling. He gave himself as an example as being profileable after he was accused of racism.

But yeah I don't agree with him on that either actually, as well as utilitarian aggregation, I think he's rather paranoid about terrorism. But I'm not really interested in politics anyways.

I'm mostly interesting in his stance on free will, lies, objective morality and his arguments against religion. I also think he's very good in debates. I agree with him a lot of things but I'm not interested in his politics. He seems rather right wing last I checked and... I think this is because he wrongly aggregates utility. Like... he thinks torture is justifiable hypothetically (in a thought experiment) if it saves huge amounts of innocent people because the suffering of one person is outweighed by saving many people... but I don't agree with him that suffering or happiness can be aggregated because of the consciousness barrier.

[Edwardo Piet]

"Within a few hours of publishing “In Defense of Profiling,” I had lunch with Ayaan Hirsi Ali, one of my favorite people on earth. (Of course, I told her that I thought she should be profiled at the airport, and we had a good laugh about my “racism.”)"

Ayaan Hirsi Ali is an atheist, an ex-Muslim. So she would not be found wearing Muslim clothing. Still, Harris thinks she should be profiled at the airport. Why, he doesn't say, but it seems pretty clear that he's using the term "Muslim" as a placeholder for anyone who looks African/Middle Eastern.

Wasn't that a joke? Lol.

He's white and I remember him giving himself as an example for profiling.

[Excited Penguin]

No, Tiberius, he was clearly making a joke, even though I never read that quote you just digged up. He has talked about how we shouldn't profile celebrities as an example of bad use of resources. Ayaan is being hunted by religious maniacs and has to take great security measures to just stay alive. This is why the joke is so on point with her.

In that same blog post you initially linked to... He talked about how 3 year Olds and very old people are/were made to take off their shoes. You should really read it a second time, especially the addendum where he addresses precisely the sort of critique you're directing at him right now. 



Would I be correct in assuming you first heard of these talking points from a third party before familiarizing yourself with the author's writings ?

[Edwardo Piet]

When I speak of profiling “Muslims, or anyone who looks like he or she could conceivably be Muslim,” I am not narrowly focused on people with dark skin. In fact, I included myself in the description of the type of person I think should be profiled (twice). To say that ethnicity, gender, age, nationality, dress, traveling companions, behavior in the terminal, and other outward appearances offer no indication of a person’s beliefs or terrorist potential is either quite crazy or totally dishonest. It is the charm of political correctness that it blends these sins against reasonableness so seamlessly. We are paying a very high price for this obscurantism—and the price could grow much higher in an instant. We have limited resources, and every moment spent searching a woman like the one pictured above, or the children seen in the linked videos, is a moment in which someone or something else goes unobserved.

Source:

https://www.samharris.org/blog/item/in-d...-profiling

I think the quoted text here is true but I still think profiling is wrong. I'm definitely very anti profiling.

Basically I accept the premise here but not the conclusion.

My emphasis.

[abaris]

I agree, Islam isn't a race.

And it's the usual cowardly defence of people who don't want to be called racist because they hate group ´X, Y or Z. To say, group X, Y or Z aren't a race. The reason why I call them bigots and very rarely use the term racism in anything I say. A weasel's defence shouldn't be strengthened by using the wrong word.

[Edwardo Piet]

Yeah it isn't a race but it's still bigoted if you're anti-muslim rather than anti-islam.

[abaris]

I think the quoted text here is true but I still think profiling is wrong. I'm definitely very anti profiling.

Basically I accept the premise here but not the conclusion.

My emphasis.

I know what he constantly says. So why should he be included in his approach to profiling? Because he's got dark hair, because of his sharp nose? Because of his facial features?

He goes by looks alone, regardless if he generously includes his august person to give the impression of being open minded.

[I_am_not_mafia]

For all the pro Sam Harris supporters out there answer me this.

How will the AI be a threat?

[The Grand Nudger]

Perhaps in any of the ways he describes in the video............or that you yourself have expressed

I'd source it for you, again, but we saw how fruitful that was the last go round. To summarize, AI is a potential threat to us for all of the same reasons that we are an -actual- threat to ourselves and other creatures, based in many instances, purely on the products and effects of our greater relative intelligence and (assumed as a matter of definition when speaking of the sort of AI being referred to) and goal seeking...without a necessity of malice or intent of harm, or even noticing that we do it, even when we try to go out of our way -not- to harm. Maybe, just maybe, the more intelligent machine -correctly- weighs given human lives lower than some other value...and whilst we might say that would be "wrong" -despite doing it ourselves..it's hard to imagine why an ai of the type envisioned would have the sort of convenient but ultimately nebulous kneejerk aversion to harming people that our biological evolution hardwired into us. We might imagine that we could control it, but we have trouble controlling the relatively lesser intelligence of human beings in precisely the same regard..and these things, in his view, have us outclassed by a wide margin. Hence the prudence in worrying about it -before- they're too smart for our own good.

None of this even approaches malfunction, or mans use -of- ai, worries all their own....or the societal crisis that such machines would almost invariably create unless they were, as he puts it, immediately put to use for the benefit of all mankind rather than their erstwhile owners. This one, amusingly, is a risk no matter how strictly we limit the function of AI, if we can...because if we use it as nothing other than a disconnected researcher to crunch data that we hand feed it..without giving it access to the outisde world...whomever happens upon it first, even if it's only as smart as a human being, ends up thousands of years ahead in the research race in short order on processing speed alone, before the next person happens upon a similar technology. He lays this out explicitly...in the video. Hopefully, we're talking about almost unfathomably benevolent human inventors here, but I doubt that's going to be the case.

Like -any- paradigm shift in technology it has immense potential for both good and harm..just as understanding nuclear forces can help us build a powerplant or a bomb, understanding intelligence could help us to lift ourselves up out of the mire, or drop us firmly back into it. Tech is neutral like that, and human beings are poor assessors of risk. We might want to have a good headstart on that count.