78

You can have some fun with people whose anticipations get out of sync with what they believe they believe.

I was once at a dinner party, trying to explain to a man what I did for a living, when he said: "I don't believe Artificial Intelligence is possible because only God can make a soul."

At this point I must have been divinely inspired, because I instantly responded: "You mean if I can make an Artificial Intelligence, it proves your religion is false?"

He said, "What?"

I said, "Well, if your religion predicts that I can't possibly make an Artificial Intelligence, then, if I make an Artificial Intelligence, it means your religion is false. Either your religion allows that it might be possible for me to build an AI; or, if I build an AI, that disproves your religion."

There was a pause, as the one realized he had just made his hypothesis vulnerable to falsification, and then he said, "Well, I didn't mean that you couldn't make an intelligence, just that it couldn't be emotional in the same way we are."

I said, "So if I make an Artificial Intelligence that, without being deliberately preprogrammed with any sort of script, starts talking about an emotional life that sounds like ours, that means your religion is wrong."

He said, "Well, um, I guess we may have to agree to disagree on this."

I said: "No, we can't, actually. There's a theorem of rationality called Aumann's Agreement Theorem which shows that no two rationalists can agree to disagree. If two people disagree with each other, at least one of them must be doing something wrong."

We went back and forth on this briefly. Finally, he said, "Well, I guess I was really trying to say that I don't think you can make something eternal."

I said, "Well, I don't think so either! I'm glad we were able to reach agreement on this, as Aumann's Agreement Theorem requires."  I stretched out my hand, and he shook it, and then he wandered away.

A woman who had stood nearby, listening to the conversation, said to me gravely, "That was beautiful."

"Thank you very much," I said.

Part of the sequence Mysterious Answers to Mysterious Questions

Next post: "Professing and Cheering"

Previous post: "Belief in Belief"

78

107 comments, sorted by Click to highlight new comments since:
New Comment
Some comments are truncated due to high volume. Change truncation settings

Hmmm... and I thought you were going to suggest that, if you succeded in making an AI that you must be god. I would've loved to be there to offer him that option instead. LOL!

Of course, even if what you said is what we really mean, I'm not sure which one is more effective at getting people to think, but your story shows that it's usually good (and at least entertaining) to try being more direct, every once in a while. I just find it easier to break through the social convention of politeness with humor.

I am quite impressed at your capability of signaling your prodigious intelligence. Less pompously, moments like that make for fond memories.

Nice story.

Nice job, but the mention of Aumann's theorem looks a bit like a sleight of hand: did the poor fellow ever learn that the theorem requires the assumption of common priors?

Robin sort-of generalized it so that it doesn't. http://www.overcomingbias.com/2006/12/why_common_prio.html

My big question though is whether this exchange led to a lasting change in the fellow's opinion as to the possibility of AI. In practice it seems to me that most of the time when people decisively loose an argument they still return to their original position within a few days just by ignoring that it ever happened.

3PrometheanFaun9y
He probably didn't see it as an argument proper, but a long misunderstanding. Most people arn't mentally equipped to make high fidelity translations between qualia and words in either direction[superficially, they are Not Articulate. More key, they might be Not Articulable], when you dismantle their words, it doesn't mean much to them, cause you havn't touched their true thoughts or anything that represents them.

This story is related to the phenomena whereby the most intelligent and educated religious folks are very careful to define their beliefs so that there can be no conflict with observations, while ordinary people are more prone to allow their religion to have implications, which are then subject to challenges like Eliezer's. It is fun to pick holes in the less educated views, but to challenge religion overall it seems more honest to challenge the most educated views. But I usually have trouble figuring out just what it is that the most educated religious folks think exactly.

I've mentioned before that my attempt to salvage a belief in God ultimately resulted in something like H. P. Lovecraft's Azathoth, which might not be too surprising as it was that ardent materialist's parody of the God of the Old Testament.

A few questions and comments:

1) What kind of dinner party was this? It's great to expose non-rigorous beliefs, but was that the right place to show off your superiority? It seems you came off as having some inferiority complex, though obviously I wasn't there. I know that if I'm at a party (of most types), for example, my first goal ain't exactly to win philosophical arguments ...

2) Why did you have to involve Aumann's theorem? You caught him in a contradiction. The question of whether people can agree to disagree, at least it seems to me, is an unnecessary distraction. And for all he knows, you could just be making that up to intimidate him. And Aumann's Theorem certainly doesn't imply that, at any given moment, rectifying that particularly inconsistency is an optimal use of someone's time.

3) It seems what he was really trying to say was someting along the lines of "while you could make an intelligence, its emotions would not be real the way humans' are". ("Submarines aren't really swimming.") I probably would have at least attempted to verify if that's what he meant rather latching onto the most ridiculous meaning I could find.

4) I've had the same experience with people who fervently hold beliefs but don't consider tests that could falsify them. In my case, it's usually with people who insist that the true rate of inflation in the US is ~12%, all the time. I always ask, "so what basket of commodity futures can I buy that consistently makes 12% nominal?"

0omalleyt6y
To point 4 and inflation, the trick is to not invest in commodity futures (where the deflationary pressures of improved production technology cancel some of the inflationary pressures of currency devaluation) but rather assets. You can invest in the S&P 500 and achieve ~11% nominal returns. Now whether asset prices are relevant to "inflation" is dependent upon whether you are trying to answer the question of "how many apples could I buy for a dollar in 1960 versus today?" or the question "how many apples could I buy for a dollar today if they were produced with the same inputs and technological process as they were in 1960?"

That was cruel. Fun, but cruel.

A woman who had stood nearby, listening to the conversation, said to me gravely, "That was beautiful."

And people wonder why men argue so often.

Meanwhile, over at the next table, there was the following conversation:

"I believe science teaches us that human-caused global warming is an urgent crisis."

"You mean if it's either not a problem or can be fixed easily, it proves science is false?"

7rkyeun10y
Technically, it proves his belief about science is false. If he'd said "Science teaches us that human-caused global warming is an urgent crisis." then "You mean if it's either not a problem or can be fixed easily, it proves science is false?" applies. And yes, it in fact would. And then Science would (metaphorically) say, "My bad, thanks for that new evidence, I reject my prior theory and form a new one that accounts for your data and explains this new phenomenon that causes symptoms as if global warming were an urgent problem."
7Bound_up7y
"Technically, it proves his belief about science is false." True, though in the same way, Eliezer's success in producing an AI, even according to the dodgy specifications of his dinner companion, would only prove his belief about God wrong, not his belief IN God wrong. The AI data point would contradict Mr Dinner's model of God's nature only at a single point, His allegedly unique intelligence-producing quality.
-9rkyeun7y
2Capla7y
Sure. But religion is supposed divinely inspired and thus completely correct on every point. If one piece of the bundle is disproven, the whole bundle takes a hit.

Even if religion is divinely inspired, a person's understanding of one aspect of religion can be wrong without invalidating all of that person's other religious beliefs.

1Capla7y
Yep.

I only torture them to make them stronger.

I know that if I'm at a party (of most types), for example, my first goal ain't exactly to win philosophical arguments ...

Funny, I've always thought that debates are one of the most entertaining forms of social interaction available. Parties with a lot of strangers around are one of the best environments for them - not only don't you know in advance the opinions of the others, making the discussions more interesting, but you'll get to know them on a deeper level, and faster, than you could with idle small talk. You'll get to know how they think.

Or how they don't...

Of course, you may not be invited back, if you offend them badly enough....

[comment deleted]

Good catch, Pseudonymous. Robin, my guess is that they're crypto-skeptics, performing for their self-perceived comparative economic/social advantage. Eliezer, please don't make something that will kill us all.

"Funny, I've always thought that debates are one of the most entertaining forms of social interaction available."

We may not have rationality dojos, but in-person debating is as good an irrationality dojo as you're going to get. In debating, you're rewarded for 'winning', regardless of whether what you said was true; this encourages people to develop rhetorical techniques and arguments which are fully general across all possible situations, as this makes them easier to use. And while it may be hard to give public demonstrations of rationality, demonstrations of irrationality are easy: simply talk about impressive-sounding nonsense in a confident, commanding voice, and people will be impressed (look at how well Hitler did).

I think the idea of argument is to explore an issue, not "win" or "lose". If you enter an argument with the mentality that you must be right, you've rather missed the point. There wasn't an argument here, just a one-sided discussion. It was a bludgeoning by someone with training and practice in logical reasoning on someone without. It was both disgusting and pathetic, no different than a high-school yard bully pushing some kid's face in the dirt because he's got bigger biceps. Did the outcome of this "argument" stroke your ego?

All-in-all, I'm not sure this is a story you should want to share. To put this in uncomplicated terms, it makes you sound like a real a$$hole. (FYI, the MRA who posted is not Ames.) MRA, the difference between winning an argument with someone, versus pushing them into the dirt - well, there's a number of differences, really. The three most important are: First, I didn't force him to talk to me. Second, losing an argument makes you stronger. (Or rather, it gives you a precious chance to become stronger; whether he took advantage of it was up to him. Winning is a null-op, of course.) Third and above all, in factual arguments there is such a thing as truth and falsity. 1Belkar1511y Yes, but humans are emotional beings, and we must recognize this. Sure, it is his fault he is so ignorant, but the difference between calmly explaining to him what is wrong with his thinking, and making fun of him, is that you hurt him on an emotional level. One must always live on both planes, and must always recognize what kind of argument he is dealing with. Emotional arguments are things to stay away from, because they only strengthen a person in the sense that bullying someone strengthens them, as opposed to teaching them KaraTe. People are tested in the little things. The unimportant things. That is where your personality shows. (BTW, how can you like this, and hate Gemara?) 7TobyBartels11y I agree with your general points, but I don't think that they apply here. Why do you say that Eliezer hurt this guy emotionally? He did amuse a witness, and it's possible that later she came and laughed at the guy, or something, but there's no evidence for that. On the contrary, the guy got a little confusion, technically won an argument (after having to clarify his position), and just maybe got something to think about. The only really cheap shot is citing Aumann. 0jacit319y i second :) Aumann has some curious priors of his own. 4TimFreeman11y The link broke sometime between 2007 and 2011. Do you have another pointer or some summary of what it said? 0[anonymous]11y The Internet Archive has it [http://replay.waybackmachine.org/20060901104304/http://www.crossingwallstreet.com/archives/2005/10/does_todays_nob.html] . 7khafra11y Does Today's Nobel Prize Winner Believe in the Bible Code? [http://www.crossingwallstreet.com/archives/2005/10/does-today%e2%80%99s-nobel-prize-winner-believe-in-the-bible-code.html] Learn to use the time machine. Back in 2007, the page looked like this. "I believe science teaches us that human-caused global warming is an urgent crisis." "You mean if it's either not a problem or can be fixed easily, it proves science is false?" Science has been proved false many times. Those things proven to be false are no longer science. OTOH most religious beliefs are dogmatic. They can't be discarded from that religion without divine intervention/prophecy. 1TraderJoe10y [comment deleted] 0Danfly10y Data accumulated using the scientific method perhaps? Once you have the data, you can make inferences to the best explanation. If the theory held to be the best explanation is falsified, that becomes part of the data. It then ceases to be the be the best explanation. Where do people get the impression that we all have the right not to be challenged in our beliefs? Tolerance is not about letting every person's ideas go unchallenged; it's about refraining from other measures (enforced conformity, violence) when faced with intractable personal differences. As for politeness, it is an overrated virtue. We cannot have free and open discussions, if we are chained to the notion that we should not challenge those that cannot countenance dissent, or that we should be free from the dissent of others. Some people should be challe... (read more) We may not have rationality dojos, but in-person debating is as good an irrationality dojo as you're going to get. In debating, you're rewarded for 'winning', regardless of whether what you said was true Only if you choose to approach it that way. Johnny Logic: some good questions. “Tolerance is not about letting every person's ideas go unchallenged; it's about refraining from other measures (enforced conformity, violence) when faced with intractable personal differences.” That’s certainly the bare minimum. His beliefs have great personal value to him, and it costs us nothing to let him keep them (as long as he doesn’t initiate theological debates). Why not respect that? “How do you know what the putative AI "believes" about what is advantageous or logical?” By definition, wouldn’t our AI frie... (read more) 2rkyeun10y Correction: It costs us nothing to let him keep them provided he never at any point acts in a way where the outcome would be different depending on whether or not it is true in reality. A great many people have great personal value in the belief that faith healing works. And it costs us the suffering and deaths of children. Mark M., "His beliefs have great personal value to him, and it costs us nothing to let him keep them (as long as he doesn’t initiate theological debates). Why not respect that?" Values may be misplaced, and they have consequences. This particular issue doesn't have much riding on it (on the face of it, anyway), but many do. Moreover, how we think is in many ways as important as what we think. The fellows ad hoc moves are problematic. Ad hoc adjustments to our theories/beliefs to avoid disconfirmation are like confirmation bias and other fallacie... (read more) JL, I’ve programmed in several languages, but you have me correctly pegged as someone who is more familiar with databases. And since I’ve never designed anything on the scale we’re discussing I’m happy to defer to your experience. It sounds like an enormously fun exercise though. My original point remains unanswered however. We’re demanding a level of intellectual rigour from our monotheistic party goer. Fair enough. But nothing I’ve seen here leads me to believe that we’re as open minded as we’re asking him to be. Would you put aside your convictions and a... (read more) Mark D, "JL, I’ve programmed in several languages, but you have me correctly pegged as someone who is more familiar with databases. And since I’ve never designed anything on the scale we’re discussing I’m happy to defer to your experience. It sounds like an enormously fun exercise though." There are programs (good ol' chatter bots) that use methods like you supposed, but they are far from promising. No need to defer to me-- I am familiar with machine learning methods, some notable programs and the philosophical debate, but I am far from an expert ... (read more) A less personal response to the second bit I quoted from Mark D: Yes, changing our beliefs in the face of good evidence and argument is desirable, and to the extent that we are able to do this we can be called critical thinkers. Would you put aside your convictions and adopt religion if a skilful debater put forward an argument more compelling than yours? To the extent the answer is "No" my atheism would be meaningless. I hope the answer is "Yes", but I have not been so tested (and do not expect to be; strong arguments for false theses should not exist). First: The argument wasn't the author being an a$$hole. He was stating the nature of his business, which is a very normal thing to do at a social gathering. (We are, to a disturbing extend, defined by our income.) Godboy dismissed his profession as quixotic, leading the author to the notion that if he created a working AI, that it would disprove God, in the mind of his coparticipant in discussion. This was a logical inferrence, based on the statement that inspired it.

Second: The only winner in a conversation is the person who learns something. I believe, t... (read more)

Wouldn't it be easier to say, an AI is not a soul? In what sense do these two words have the same meaning? An AI is a non-existant entity which, due to the unflagging faith of some, is being explored. A soul is an eternal being granted life (human only?) by god (should that be capitalized?) Comparing them is what leads to the problem.

Before using Aumann one should ask, "What does this guy know that I don't?"

Douglas, (1) what makes you think that anyone was suggesting that "AI" and "soul" have the same meaning?, (2) in what way would "an AI is not a soul" be a useful substitute for anything else said in this discussion?, and (3) why should comparing the two notions lead to any problems, and in particular to whatever you're calling "the problem" here?

I don't think it's any more obvious that there are no AIs than that there are no souls. That is: perhaps, despite appearances, my computer is really intelligent, and thinks i... (read more)

g- the man said, "I don't believe AI is possible because only God can make a soul." "...If I can make an AI it proves your religion false?" Somebody in this exchange has equated the making of an AI with the making of a soul. That's why I would suggest that the words have been confused. An AI is not a soul would be useful in this discussion because it would clarify that the making of one would not invalidate the existence of the other or the statement that "only God can make a soul". Comparing the two notions would not be a pr... (read more)

Douglas: OK, I hadn't realised you were talking about him; my bad. And, sure, another approach Eliezer could have taken is to say "an AI and a soul aren't the same thing". But I don't see why that would be any improvement on what he actually did do.

Also: "soul" is used vaguely enough that I don't think Eliezer could justifiably claim that an AI wouldn't have to be a soul. If his interlocutor believed, e.g., that a soul is what it takes in order to have real beliefs, feelings, will, etc., then saying "oh no, I'm not talking about so... (read more)

g- you ask good questions. My point about AI and religion is that rather than pretending that one is related to the other, AI would benefit from clearing up this confusion. (So would the religious) Perhaps the way Elizer went about it was OK I would define "soul" as a non-corporeal being that exists separable from the body and that survives body death. (I want to say something about the soul being the true source of consciousness and ability-- OK, I said it)

cool :)

Seems more like Aikido. I sense a broken spirit more than a redirection of thought or processes of it and his belief system. Simply put, honey has always gotten me more flies than vinegar.

Rereading the post, I don't understand why the fellow didn't just say "I defy your ability to build an AI" in response to your first question. Maybe he was intimidated at the moment.

4thomblake11y
Because he wanted the ability to retain his religious belief in the face of successful AI - and he can anticipate, in advance, exactly which experimental results he'll need to excuse [http://lesswrong.com/lw/i4/belief_in_belief/]

I have to heartily disagree with those that seem to think it impolite to disagree with the religious. Remember this same person is going to go out and make life and death decisions for himself and others. Notice also that it was the theist who started the debate.

All you did was show that your argumentative skills were better. His intial belief mentioned souls, and i dont think you ever did. I'd like to see some sort of testability for souls :)

As to your reply of possibly proving his religion false, if he was better at arguing, he may have replied that at the least it might prove his understanding of religion false.

And of course its not as if you have created an AI.

3Hul-Gil11y
Your points are irrelevant. The man asserted that his religious beliefs meant Artificial Intelligence was impossible, and that's what the author of this post was debating about. No souls need to be tested, because the existence of souls was not contested. Nor did Eliezer say he had created an AI. I'm also surprised no one pointed out that Mark D's "reversal" scenario is totally wrong: if Eliezer was unable to create an AI, that does not at all imply that the man's own assertions were true. It might, at best, be very weak evidence; there could be many reasons, other than a lack of a soul, that Eliezer might fail.

I attended a lecture by noted theologian Alvin Plantinga, about whether miracles are incompatible with science. Most of it was "science doesn't say it's impossible, so there's still a chance, right?"-type arguments. However, later on, his main explanation for why it wasn't impossible that God could intervene from outside a closed system and still not violate our laws of physics was that maybe God works through wavefunction collapse. Maybe God creates miracles by causing the right wavefunction collapses, resulting in, say, Jesus walking on water, rising from the dead, unscrambling eggs, etc.

Recalling this article, I wrote down and asked this question when the time came:

"The Many-Worlds Interpretation is currently [I said "currently" because he was complaining earlier about other philosophers misrepresenting modern science] one of the leading interpretations of quantum mechanics. The universe splits off at quantum events, but is still deterministic, and only appears probabilistic from the perspective of any given branch. Every one of the other branches still exists, including ones where Jesus doesn't come back. If true, how does this affect your argument?&q

Now, what I don't get is why he let you force him to change his position. If he really believed that it was impossible for you to create AI, why wouldn't he have just said "yes," and then sit back, comfortable in his belief that you will never create an AI?

4lessdazed11y
He didn't believe his religion was definitely true, even though he belived that he believed that [http://lesswrong.com/lw/i4/belief_in_belief/]. There is nothing paradoxical about being wrong about things, even one's own beliefs. "My religion is true" is a statement with no consequences for his anticipations, so it is isolated from his belief network. He was deeply committed to "I believe in my religion", and that second belief required him to pretend that the first statement was a member of his belief network by pretending that he believed in consequences as if his religion were true occurred and just happened to be untestable. Once he realized that he had goofed and said he anticipated consequences if his religion were true different from if it weren't true in a testable area, he had to backtrack. By never anticipating different consequences if his religion is true than if it isn't, he protects his belief that he believes his religion is true from falsification. So he changed his position only insofar as he updated his belief framework, but he didn't change any core belief. Everything snaps into place if you realize the terms of his real belief the way the movement of planets makes sense if you realize the Earth is not the center of the Solar System, and the Sun is. His religious belief is meaningless, not just wrong, and his belief that he believes is wrong. People are capable of even more sophisticated levels [http://lesswrong.com/lw/s/belief_in_selfdeception/] levels of self deception than it seems this guy had. Welcome to Less Wrong [http://lesswrong.com/lw/2ku/welcome_to_less_wrong_2010/]!

This post's presence so early in the core sequences is the reason I nearly left LW after my first day or two. It gave me the impression that a major purpose of rationalism was to make fun of other people's irrationality rather than trying to change or improve either party. In short, to act like a jerk.

I'm glad I stuck around long enough to realize this post wasn't representative. Eliezer, at one point you said you wanted to know if there were characteristically male mistakes happening that would deter potential LWers. I can't speak for all women, but this post exemplifies a kind of male hubris that I find really off-putting. Obviously the woman in the penultimate paragraph appreciated it in someone else, but I don't know if it made her think, "This is a community I want to hang out with so I, too, can make fools of other people at parties."

6jsalvatier11y
Do you occasionally see other comments/posts that give you this same vibe?
2juliawise11y
I can't think of any. Maybe there have been comments, but they're not sanctioned in the same way a core sequences post is, so I'm more apt to dismiss them.
0Jotto99910y
Before I say anything I would like to mention that this is my first post on LW, and being only part way through the sequences I am hesitant to comment yet, but I am curious about your type of position. What I find peculiar about your position is the fact that Yudkowsky did not, as he presented here, engage the argument. The other person did, asserting "only God can make a soul", implying that Yudkowsky's profession is impossible or nonsensical. Vocalizing any type of assertion, in my opinion, should be viewed as a two-way street, letting potential criticism come. In this particular example the assertion was of a subject that the man knew would be of large interest to Yudkowsky, certainly disproportionately more than say whether or not the punch being served had mango juice in it. I'd like to know what you expect Yudkowsky should have done given the situation. Do you expect him not to give his own opinion, given the other person's challenge? Or was it instead something in particular about the way Yudkowsky did it? Isn't arguing inevitable and all we can do is try to build better dialogue quality? (That has been my conclusion for the last few years). Either way, I don't see the hubris you seem to. My usual complaints of discussions is that they are not well educated enough and people tend to say things that are too vague to be useful, or outright unsupported. However I rarely see a discussion and think "Well the root problem here is that they are too arrogant", so I'd like to know what your reasoning is. It may be relevant that in real life I am known by some as being "aggressive" and "argumentative". Though you probably could have inferred that based on my position but I'd like to keep everything about my position as transparent as possible. Thank you for your time.

If I were the host I would not like it if one of my guests tried to end a conversation with "We'll have to agree to disagree" and the other guest continued with "No, we can't, actually. There's a theorem of rationality called Aumann's Agreement Theorem which shows that no two rationalists can agree to disagree." In my book this is obnoxious behavior.

Having fun at someone else's expense is one thing, but holding it up in an early core sequences post as a good thing to do is another. Given that we direct new Less Wrong readers to the core sequence posts, I think they indicate what the spirit of the community is about. And I don't like seeing the community branded as being about how to show off or how to embarrass people who aren't as rational as you.

What gave me an icky feeling about this conversation is that Eliezer didn't seem to really be aiming to bring the man round to what he saw as a more accurate viewpoint. If you've read Eliezer being persuasive, you'll know that this was not it. He seemed more interested in proving that the man's statement was wrong. It's a good thing to learn to lose graciously when they're wrong, and learn from the experience. ... (read more)

1Jotto99910y
Interesting. Do we have any good information on the attributes of discussions or debates that are the most likely to educate the other person when they disagree? In hindsight this would be a large shortcoming of mine, having debated for years now but never invested much in trying to optimize my approach with people. Something I've noticed: when someone takes the "conquer the debate" adversarial approach, a typical-minded audience appears more likely to be interested and side with the "winner" than if the person takes a much more reserved and cooperative approach despite having just as supported arguments. Maybe the first works well for typical audiences and the second for above-typical ones? Or maybe it doesn't matter if we can foster the second in "typical" minds. Given my uncertainty it seems highly unlikely that my approach with people is optimal. Do you have any tips for someone interested in making a mental habit out of cooperative discussion as opposed to being adversarial? I find it very difficult, I'm an aggressive and vigorous person. Maybe if I could see a video of someone using the better approach so I can try to emulate them.
5thomblake10y
I hope you've noticed you changed the subject here. In the first paragraph you're trying to persuade the person with whom you are conversing; in the second paragraph you're trying to convince an audience. They might well require entirely different methods.
0Jotto99910y
You're right, I see now that the effect on audiences does not relate much to the one-on-one, so I should have kept a clear distinction. Thank you for pointing this out. I believe this obvious mistake shows that I shouldn't comment on the sequences as I work my way through them, but rather it is better if I only start commenting after I have become familiar with them all. I am not ready yet to make comments that are relevant and coherent, and the very last thing I want to do is pollute the comment section. I am so glad about the opportunity for growth this site has, thanks very much to all.
4thomblake10y
Meh. Comments on old sequence posts don't add much noise, as long as the comment threads don't explode.
1juliawise10y
An adversarial approach may impress spectators. In Eliezer's example, it impressed at least one. But I think it's more likely to alienate the person you're actually conversing with. I don't have objective research on this. I'm working from personal experience and social work training. In social work you assume people are pretty irrational and coax them round to seeing what you think are better approaches in a way that doesn't embarrass them. In social work we'd call it "collaborative empricism" or Socratic questioning. Here's video example [http://www.youtube.com/watch?v=kPzuSFUnYVc&feature=related] of a therapist not shouting "Of course you're not being punished by God!" It's more touchy-feely than an argument, but the elements (taking the outside view, encouraging him to lay out the evidence on the situation) are there.

Shortly after my stroke, my mom (who was in many ways more traumatized by it than I was) mentioned that she was trying to figure out what it was that she'd done wrong such that God had punished her by my having a stroke. As you might imagine, I contemplated a number of different competing responses to this, but what I finally said was (something along the lines of) "Look, I understand why you want to build a narrative out of this that involves some responsible agent making decisions that are influenced by your choices, and I recognize that we're all in a difficult emotional place right now and you do what you have to do, but let me offer you an alternative narrative: maybe I had a survivable stroke at 40 so I'd start controlling my blood pressure so I didn't have a fatal one at 45. Isn't that a better story to tell yourself?"

I was pretty proud of that interaction.

0juliawise10y
Nice work! That's the same idea as narrative therapy [http://en.wikipedia.org/wiki/Narrative_therapy]: drawing a new storyline with the same data points.
0Jotto99910y
Hmm! I found that actually quite helpful. The therapist didn't even voice any apparent disagreement, he coaxed the man into making his reasoning explicit. This would greatly reduce the percent of the argument spent in an adversarial state. I noticed that it also put the emphasis of the discussion on the epistemology of the subject which seems the best way for them to learn why they are wrong, as opposed to a more example-specific "You're wrong because X". Thank you for that link. Would it be useful for me to use other videos involving a therapist who disagrees with a delusional patient? It seems like the ideal type of behaviour to try and emulate. This is going to take me lots of practice but I'm eager to get it. Thank you for your help and advice!
0juliawise10y
I'm not sure. The kind of irrational beliefs you're likely to talk about with others are some kind of misconception or cached belief, rather than an out-and-out delusion like "people are following me everywhere", which probably stems from a chemical imbalance and can't really be talked away. You could try reading up on CBT, but the literature is about doing therapy, which is a pretty different animal from normal conversations. Active listening [http://en.wikipedia.org/wiki/Active_listening] might be a more useful skill to start with. People are less defensive if they feel you're really trying to understand their point of view.

If I were the host I would not like it if one of my guests tried to end a conversation with "We'll have to agree to disagree" and the other guest continued with "No, we can't, actually. There's a theorem of rationality called Aumann's Agreement Theorem which shows that no two rationalists can agree to disagree." In my book this is obnoxious behavior.

I'd find it especially obnoxious because Aumann's agreement theorem looks to me like one of those theorems that just doesn't do what people want it to do, and so ends up as a rhetorical cudgel rather than a relevant argument with practical import.

Agreed. If this was Judo, it wasn't a clean point. EY's opponent simply didn't know that the move used on him was against the sport's rules, and failed to cry foul.

Storytelling-wise, EY getting away with that felt like a surprising ending, like a minor villain not getting his comeuppance.

Interesting. When I am arguing with somebody, I usually get them to explicitly define every one of their terms and then use the definitions to logic them into realising that their argument was faulty. A more rational person could have escaped your comment simply by defining AI as human-like intelligence- ie, ability to create, dream, emote and believe without prior programming for those things. And yes, I am religious and my belief can be overturned by proof. If aliens are found with human-like intelligence, I will give up my faith entirely- but until then, just about anything else can be explained from within my ideology.

8wedrifid10y
Does that work with actual people?

[ Disclaimer: This is my first post so please don't go easy on me. ]

After reading a handful of comments I am surprised to see so many people think of what Eliezer did here as some sort of "bad" thing. Maybe I'm missing something but after reading all I saw was him convincing the man to continue the discourse even though he initially began to shy away from it.

Perhaps citing a theorem may have intimidated him a little, but in all fairness Eliezer did let him know at the outset that he worked in the field of Artificial Intelligence.

4Jiro8y
Old post, but... 1. He cited a theorem knowing that his opponent wouldn't be able to detect a bad cite. 2. It actually was a bad cite.

I've already seen plenty of comment here on just how awkward this post is to be so early in the Sequences, and how it would turn people off, so I won't comment on that.

However: Seeing this post, early in the sequences, led me to revise my general opinion of Eliezer down just enough that I managed to catch myself before I turned specific admiration into hero-worship (my early, personal term for the halo effect).

I seriously, seriously doubt that's the purpose of this article, mainly because if Eliezer wanted to deliberately prevent himself from being affecti... (read more)

1[anonymous]10y
Nah, didn't happen. The essay reports an adolescent fantasy featuring martial invincibility. I'm sure the author has grown up by now.
3wedrifid10y
So early in the sequences? It would seem to be worse later in what we now call the sequences. At the time this was written it was just a casual post on a blog Eliezer had only recently started posting on. Perhaps the main error is that somehow someone included it in an index when they were dividing the stream of blog posts into 'sequences' for reference.

There's a theorem of rationality called Aumann's Agreement Theorem which shows that no two rationalists can agree to disagree. If two people disagree with each other, at least one of them must be doing something wrong.

This seems like one of those things that can be detrimental if taught in isolation.

It may be a good idea to emphasize that only one person in a disagreement doing something wrong is far less likely than both sides in a disagreement doing something wrong.

I can easily imagine someone casually encountering that statement, and taking it to ins... (read more)

Eliezer, that's false reasoning. I'm not religious, so don't take this as the opening to a religious tirade, but it's a pet peeve of mine that intelligent people will assert that every belief within a religion is wrong if only one piece of it is wrong.

There are a billion and one reasons why a body of knowledge that is is mostly correct (not saying I think religions are) could have one flaw. This particular flaw doesn't prove God doesn't exist, it would only prove God souls aren't necessary for an intelligent life form to survive, or (perhaps, to a religi... (read more)

2[anonymous]9y
4wedrifid9y
Consider rewording that in such a manner that you can fit the 'hubris' label in while leaving the sentence coherent.
0TheOtherDave9y
Well, if person A's religion strictly implies the claim that only God can make a soul and this precludes AI, then the falsehood of that claim also implies the falsehood of A's religion. (A->B => -B -> -A) But sure, you're of course correct that if person A is mistaken about what person A's religion claims, then no amount of demonstrated falsehoods in person A's statements necessarily demonstrates falsehood in person A's religion. That said... if we don't expect person A saying "my religion claims X" given that person A's religion claims X, and we don't expect person A saying "my religion doesn't claim X" given that person A's religion doesn't claim X, then what experiences should we expect given the inclusion or exclusion of particular claims in person A's religion? Because if there aren't any such experiences, then It seems that this line of reasoning ultimately leads to the conclusion that not only the objects religions assert exist, but the religions themselves, are epiphenomenal.
1J Mann3y
I think the "strictly implies" may be stealing a base. Yes, being convinced of the existence of the AI would make the man rethink the aspects of his religion that he believes renders an AI impossible, but he could update that and keep the rest. From his perspective, he'd have the same religion, but updated to account for the belief in AIs.
0Lotska9y
It looks like false logic to me too, but I'm very aware that that is how many Christians "prove" their religion to be true. 'The Bible says this historical/Godly event happened and this archeological evidence supports the account in the Bible, therefore the Bible must be true about everything so God exists and I'm going to Heaven.' Which sounds very similar to 'This is a part of what you say about your religion and it may be proved false one day, so your religion might be too.' Is it okay to slip into the streams of thought that the other considers logic in order to beat them at it and potentially shake their beliefs?
1DSherron9y
Basically, the question here is whether you can use the Dark Arts with purely Light intentions. In the ideal case, I have to say "of course you can". Assuming that you know a method which you believe is more likely to cause your partner to gain true beliefs rather than false ones, you can use that method even if it involves techniques that are frowned upon in rationalist circles. However, in the real world, doing so is incredibly dangerous. First, you have to consider the knock-on effects of being seen to use such lines of reasoning; it could damage your reputation or that of rationalists in general for those that hear you, it could cause people to become more firm in a false epistemology which makes them more likely to just adopt another false belief, etc. You also have to consider that you run on hostile hardware; you could damage your own rationality if you aren't very careful about handling the cognitive dissonance. There are a lot of failure modes you open yourself up to when you engage in that sort of anti-reasoning, and while it's certainly possible to navigate through it unscathed (I suspect Eliezer has done so in his AI box experiments), I don't think it is a good idea to expose yourself to the risk without a good reason. An unrelated but also relevant point: everything is permissible, but not all things are good. Asking "is it okay to..." is the wrong question, and is likely to expose you to some of the failure modes of Traditional Rationality. You don't automatically fail by phrasing it like that, but once again it's an issue of unnecessarily risking mental contamination. The better question is "is it a good idea to..." or "what are the dangers of..." or something similar that voices what you really want answered, which should probably not be "will LWers look down at me for doing ..." (After all, if something is a good idea but we look down at it then we want to be told so so that we can stop doing silly things like that.)
1descent2y
The framing of the first sentence gives me a desperately unfair expectation for the discussion inside HPMOR- I'm excited.

Me: Writes on hand "Aumann's Agreement Theorem". Thank you Eliezer, you have no idea how much easier you just made my Theory of Knowledge class. Half of our discussions in class seem to devolve into statements about how belief is a way of knowing and how everyone has a right to their own belief. This (after I actually look up and confirm for myself that Aumann's Agreement Theorem works) should make my class a good deal less aggravating.

1descent2y
If it also states that the participants must be rationalists as Yudkowski specifies, you'll be sorely disappointed to find out how many people would identify as a rationalist

I would have loved to watch that.

That was beautiful!

I wrote a long post saying what several people had already said years ago, then shortened it. Still, because this post has made me mad for years:

1) Of COURSE people can agree to disagree! If not, EY is telling this guy that no two rationalists currently disagree about anything. If THAT were true, it's so fascinating that it should have derailed the whole conversation!

(Leaving aside, for the moment, the question of whether Aumann's theory "requires" a rationalist to agree with a random party goer. If it really did, then the party goer c... (read more)

While I understand the absolute primal urge to stomp on religious texts used to propagate compulsory heterosexuality, I do think this exchange ended up a bit of a poor game, when it seems like he'd be mostly interested in discussing how the emotions of programmed thought might differ from ours (and that's a fun well to splash around in, for a while)(though deposing of cult-friendly rhetoric is valuable too, even if you have to get nasty).

I'm mildly concerned about the Reign Of Terror Precept, but I also understand it. It's just disappointing to know that t... (read more)