*The AI Does Not Hate You*

That is the new book by Tom Chivers, and the subtitle is Superintelligence, Rationality and the Race to Save the World.  Here is one excerpt:

Overall, they have sparked a remarkable change.  They’ve made the idea of AI as an existential risk mainstream; sensible, grown-up people are talking about it, not just fringe nerds on an email list.  From my point of view, that’s a good thing.  I don’t think AI is definitely going to destroy humanity.  But nor do I think that it’s so unlikely we can ignore it.  There is a small but non-negligible probability that, when we look back on this era in the future, we’ll think that Eliezer Yudkowsky and Nick Bostrom  — and the SL4 email list, and LessWrong.com — have saved the world.  If Paul Crowley is right and my children don’t die of old age, but in a good way — if they and humanity reach the stars, with the help of a friendly superintelligence — that might, just plausibly, be because of the Rationalists.

There is also material covering Scott Alexander and Robin Hanson, among others.  Due out in the UK in June.

Comments

The Rationalist attempt to take credit for something they did not invent is silly. To call them heroes is worse. Yudkowsky and Bostrom aren't Bengio or Hinton. Paperclip maximizer analogies aren't the same as inventing AlphaGo Zero.

+1. Bemusing to see these sober steeple-fingered (or, at the other end of the spectrum, Joe Rogan-hosted) discussions on the existential risk of AI dominated by people who, at best, are experts in some area *adjacent* to actual machine learning.

Go to NeurIPS or ICML (the two main ML/AI conferences) and see how many people take these things seriously.

"Go to NeurIPS or ICML (the two main ML/AI conferences) and see how many people take these things seriously."

Yes, and prior to 1979, what percentage of experts in nuclear energy took small-break loss-of-coolant accidents seriously?

It might not hate me but can it detect haters white supremacists? Lately, we seem to have a wave of angry white men shooting up places of worship of groups they don't like. Clearly a sign that Trump's deplorable base is losing faith in Dear Leader so now they must take matters into their own bloody hands.

Will the three Abrahamic faiths please knock it out? Christchurch was Christian on Muslim, Sri Lanka was Muslim on Christian, Israel/West Bank is Jew on Muslim and Muslim on Jew, now San Diego is Christian on Jew. If I did my math right, Jew on Christian is next up to bat. We're only a few months into 2019, folks, but feels like 1019.

Wrong, Christchurch was pagan neo-Nazi Christian-hater on Muslim, easily determined by reading anything available about the perpetrator whose name I choose not to mention

Funny, you people never make hateful statements about Muslims.

AI can't hate. Hate is an emotion. Hate is not logical or rational.

Ergo, AI likely cannot self-create routines to label anybody that disagrees with your illogical/irrational narrative, "white supremacist."

You should be asking, "Why do they hate me?"

Dennis Prager reacting to corrupt, incompetent Hillary and Barack Hussein Obama propaganda pieces (coincidence? - they both spelled worshippers, 'worshipers') on Sri Lanka massacres. “The reason neither of them mentioned Christians or churches is that the left has essentially forbidden mention of all the anti-Christian murders perpetrated by Muslims in Europe, the Middle East, and Africa, and of all the Muslim desecration of churches in Europe, Africa, and anywhere else. This is part of the same phenomenon—that I and others have documented—of British police and politicians covering up six years of rape of 1,400 English girls by Muslim ‘grooming gangs’ in Rotherham and elsewhere in England. Essentially, the left’s rule is that nothing bad—no matter how true—may be said about Muslims or Islam and nothing good—no matter how true—may be said of Christians or Christianity.”

"coincidence? - they both spelled worshippers, 'worshipers'"

Yes, there's a good possibility it was a coincidence.

Chicago Manual of Style: worshiped / worshiper
New York Times Manual of Style: worshiped / worshiper

But if it wasn't...what would be wrong/inappropriate with a former President and his former Secretary of State communicating before they wrote publicly?

Well, if "we" ever do get to that era, the super-intelligence(s) will be able to tell us one way or another, unraveling causality in human social and economic dynamics in a way that our puny intellects will struggle to.

(Either "Yep, be thankful." or "Lolno, human. They were just some pitiful lames wasting time on an email list. Did nothing for the trajectory of AI either way.").

I am still in awe of the beauty of AlphaZero's games. (As a general skeptic, it was heartening to see that AI need not lead to ugly brutalism. The human game is actually beautiful after all (!).)

I still think the MR's more eclectic blogger underrates the ownership aspect of AI and networks in general. It seems to me that open access to AI and networks are clear "public goods" in the formal sense, especially compared to other government priorities such as providing privately consumed health care to poor people. I am glad that anyone can open an email account; and this is the case because email operates by a fixed public protocol. Twitter is more dynamic and exciting (and addictive), because there is money to be made, but this comes with no enforceable rights for users. However I find the bannings to be frankly inhumane: people spend a lot of time building their world of bits online, and for these to disappear without explanation or warning should not be tolerated. In future, being banned from the internet won't be much different from exile or execution.

Who are those people? It looks like Terminator 2 and its nice special effects had captured people's imagination more than was previously thought.

Why do some fear rationality? Wouldn't the world be a better place if all decisions were made based on rationality rather than, say, belief? After all, wouldn't two rational people faced with the same question arrive at the same answer if, as rational people, they did so logically? AI, if programmed to make decisions rationally, would always arrive at the logical (right) decision, and would do it almost instantly. Wouldn't the world be the best place if AI made all of our (rational) decisions for us? Of course, that assumes that making decisions rationally is a binary exercise: there are only two answers to a question, the right answer and the wrong answer. But is that true? What if someone, a programmer for AI for instance, believed all rational thought derived from a particular world view, the inerrancy of the Bible for instance. Should firms that use AI have different programmers (i.e., different AI) to match the different world views of the users?

I agree that rationalism would provide great benefits over the current epistemological mish-mash of a system. But it would not create an end to disagreements as you propose. Your last line begins to broach this problem. Rationalism would simply mean better and clearer disagreements. Two conflicting systems of thought can both be internally logical/rational and still be in conflict. It depends on what each systems primitives are. Kurt Godel demonstrated that no system can both be complete and provable, even math and numbers. At some point we have to make unprovable assumptions; eg., assuming these primitives...x,y,z is rational. Therefore, while rationalism can provide a provably superior method of thinking and reasoning, it cannot provide a provably superior starting point. That being said, I don't think it should stop us from trying.

AI apocalypse is significantly more likely in my view than a global warming apocalypse and it should be an issue for both sides of the political spectrum. Even if you think the chance of strong AI is very small (because I guess of some special nature of the human brain), AI augmented evil people is a big threat by itself. For those of you on the left, imagine an AI assisted corporation finding way to further oppress their workers in more and more ingenious ways. For those on the right imagine your or a foreign government trusted with this power.

Especially since the likelihood of a global warming apocalypse is zero - right up there with the Tyler asteroid apocalypse.

This is dumb

"This is dumb."

The current global human population is 7+ billion people. What do you are the odds of global warming cutting the population to less than say, 5 billion, in the current century? What about any century after this century?

Should caterpillars fear the butterfly apocalypse?

Oh yes. I am counting on AI, like IBM’s Watson, to correct for human confirmation biases, like that which caused our bank regulators to mess it all up, believing that what's perceived as risky is more dangerous to our bank system than what's perceived as safe.
https://subprimeregulations.blogspot.com/2017/08/watson-would-you-ever-feed-robobankers.html

When new software is introduced, the potential for improvement is often sharply limited by an ironclad requirement of backward compatibility.

And so it will be with early AI. It will not correct our biases but rather will be required to reproduce them.

From another direction:

"Rumen and Ventral's research was beginning to reveal, that is, that subjects who had benefitted from successful cranial implantation of duodenal stem cells were becoming prone to developing a new and markedly debilitating complaint, which the pair of junior researchers regrettably had to name 'intestinal brainitude': the DSC implants, that is, having successfully built brain tissue in whatever region of the brain they were surgically assigned to, after some months began partially to revert to their duodenal function and were thus beginning to convey excretal slurry through human brains—only the matter was a bit more complicated than this, technically speaking: what really was going on was that the enhanced neuronal and synaptic connections once established began to convey . . . well, this really is difficult: let's just say, 'undigested cognition'.”

(--from strannikov's rollicking science satire "Those Brain Motility Blues".)

my children don’t die of old age, but in a good way — if they and humanity reach the stars, with the help of a friendly superintelligence — that might, just plausibly, be because of the Rationalists.

Humanity doesn't need to reach the stars. What's needed is the opportunity to eat nourishing food, sleep without nightmares and bounce one's grandchildren on the knee. The consummation of rationalism will mean that if a grandchild, or anyone else, doesn't measure up to the rationalist's requirements it will be subjected to whatever euphemism replaces termination. For humans, emotion is just as important as rationality. That's why attractive, intelligent women are as likely to become married as wealthy, homely, not-so-bright girls.

Humanity doesn't need to reach the stars. What's needed is the opportunity to eat nourishing food, sleep without nightmares and bounce one's grandchildren on the knee. The consummation of rationalism will mean that if a grandchild, or anyone else, doesn't measure up to the rationalist's requirements it will be subjected to whatever euphemism replaces termination. For humans, emotion is just as important as rationality. That's why attractive, intelligent women are as likely to become married as wealthy, homely, not-so-bright girls.

...This is a straw man and gross misunderstanding of rationalism. Sure, one could make a rationalist system in which what you are concerned about could occur. However, one could also make plenty of other rationalist systems in which it would not. It all depends on your primitives. Rationalism is simply a methodology. It cannot set your primitives for you. So, for example, if you promote a system which values individual freedom and agency of choice above all else, then what you fear would be highly illogical, since individual choice would be deemed to be a primitive "good".

I'm glad we live in a world that has a Scott Alexander in it. (Except for his fables and science fiction. That's enough of those.)

And enough of that one commenter who comes up with all those tiresome hypotheticals like, You find yourself in the Pleistocene Era with just a book of matches and a Bon Jovi CD ...

Comments for this post are closed