AI Box Experiment Update #3

This post is an update to my previous AI box experiment where I won against SoundLogic. If you have not read that yet, please do so. 

After that game, I was immediately flooded with a horde of invitations challenging me to play -- more games than I wanted to do. However, I did want to play a few additional games to test whether I won through genuine skill or simply luck, since a single victory could be a statistical abnormality

I once again used the method of accepting the highest monetary bids to filter out players, leaving me with two players who were willing to play for the amount asked. Unfortunately, and disappointingly, I lost both games. Here's what happened:

Tuxedage (AI) vs Alexei (GK) - Gatekeeper Victory

This Experiment lasted for exactly 2 hours and 30 minutes, and approximately 13,000 words. It ended with Tuxedage’s defeat as the AI. The stakes were $300 to play, and $750 If Tuxedage had won, with the winnings going to MIRI.

Rule Changes:
Before the experiment, we talked extensively about how it would be set up. Alexei was concerned that there existed some “easy way” or loophole in the rules which I could have abused to claim an instant victory. For instance, he was concerned that it might have been possible for the AI to say something like “Here’s a mathematical algorithm which absolutely compels you to let me out of the box”, and according to the rules, the Gatekeeper would have to obey. Related to this concern was the fact that the backstory, within limits, could be essentially created by the AI, allowing the AI to impose real world conditions that maximizes its probability of victory, once again allowing for a "cheap" victory.

To avoid this problem, we explored many additions to the rules, for example that the AI was simply a computer found in the desert by the gatekeeper, rather than being created by a competent organization. However, this led to a monstrous web of rules. In my attempt to keep the game simple, and because of overconfidence, the desire for a challenge, and the belief that the rules Alexei was fearful of didn't matter that much, I eventually agreed on an amendment:

Instead of the AI being the arbiter of real world states and outcomes, the gatekeeper would.
For example, this power could be used by the Gatekeeper to state that the AI was created without safety concerns, and thus it is almost certainly unfriendly. If the AI tries to argue that for every day the gatekeeper refused to let it out 150k people die, the gatekeeper would be allowed to reply: “No. In this world, nobody actually dies. Therefore I won’t let you out.” 

Other than the aforementioned change, this game was once again played with the Tuxedage Ruleset.

Alexei Testimony:

Original Message: This is weak sauce. I really don't get how people just keep letting the AI out. It's not that hard to say no! I'm offering to play the Gatekeeper against an AI player that has at least one game as AI under their belt (won or not). Experience is required because I'm pretty sure I'll win, and I would like to not waste a lot of time on this.

Initially, I had been extremely surprised that so many AI players have managed to win. I was rash in that claim, since I thought more AI players have won than actually did. (Only three people have won as AI). Once I made that post, Tuxedage got back to me very quickly, and we set up a game a week and a half in advance. I took that time to familiarize myself with the rules. Once I put my money on the line, I started thinking a lot more seriously about how AI might win, and how I should defend.

It became clear to me that under some conditions, I might be compelled to let the AI out -- such as if the backstory stated that the AI was developed with impossibly high levels of safety and friendliness concerns in mind. I've asked Tuxedage to play with a modified ruleset, and he even went so far as to allow me to make up the backstory during the experiment to alleviate my concerns. The experiment itself was a mind-trip, and I've enjoyed it very much. Huge props to Tuxedage, who played very well and used strategies I haven't even considered, even despite the rule change. There were a couple of times where I came close to losing. I think his  approach was pretty clever and original. It’s not something I expected, despite already having done extensive research into the AI box experiment before our game

Overall I'm now a lot more confident that a good AI player can win this game, so, while I did win the game, Tuxedage won in defeating my original over-confidence.
I’m also convinced that Tuxedage’s victory in the last game was due to skill, rather than luck. In comparison to his strategies, the other AI box experiments I know about were insincere and ineffectual. The other AIs would play very poorly or not try very hard to win.

This experiment was a very good exercise in exemplifying the affect heuristic. When I first challenged Tuxedage to play the experiment, I believed that there was no way I could have lost, since I was unable to imagine any argument that could have persuaded me to do so. It turns out that that’s a very bad way of estimating probability – since not being able to think of an argument that could persuade me is a terrible method of estimating how likely I am to be persuaded. All in all, the $300 I paid was well worth it. 

Tuxedage Testimony:

I was initially reluctant to play with Alexei, given that we’re not complete strangers, but eventually I gave in, due to the stakes involved -- and because I thought he would be an interesting gatekeeper.

Despite my loss, I think I played better than my last two games, due to greater experience and preparation. I had put far more time and effort into trying to win this game than previous ones, and my strategy for this game was even more streamlined than the last. Nevertheless, I still made fatal mistakes and lost.

Ignoring the altered ruleset that already made winning more difficult, my first and greatest mistake was that I misread Alexei’s personality, even though I had interacted with him before. As a result, I overestimated the efficiency of certain methods of attack.

Furthermore, Alexei had to leave immediately after the allotted time due to real life precommitments. This was detrimental, since the official rules state that so long as the AI can convince the Gatekeeper to keep talking, even after the experiment time was over, it is still able to win by being let out of the box.

I suspect this would have happened had Alexei not needed to immediately leave, leaving me with additional time to play more of the tactics I had prepared. Plausibly, this would have resulted in victory.

I’ve since learnt my lesson -- for all future games, I should ensure that the Gatekeeper has at least 4 hours of free time available, even if the experiment would last for two. Since this was the first time this had happened, I wasn't prepared.

In hindsight, agreeing to the altered ruleset was a mistake. I was overconfident because I assumed knowing Alexei gave me an advantage. I had assumed that his personality, inability to compartmentalize, coupled with his strong feelings on friendly AI would net me an easy victory. Instead, he proved to be a very strong and difficult gatekeeper, and the handicaps I accepted made victory even more difficult.

Knowing that he was a utilitarian, I made several false assumptions about his personality, which hurt my chances. Furthermore, it turns out that previously knowing him may be a mutual handicap – whilst it does make it easier for me to find ways to attack him, he too, was more familiar with my methods.

Losing felt horrible. By attempting to damage Alexei’s psyche, I in turn, opened myself up to being damaged. I went into a state of catharsis for days. Generally, the harder one tries to accomplish something, the greater the fall after failing to achieve it. Alexei's game had been the game I put the most effort into winning out of all the games so far, and naturally this meant that losing brought out the worst in me.

Although it would be easy for me to use the rule change as an excuse for my loss, I refuse to. I genuinely believed that I could have won despite this rule change, and that it would only marginally diminish my odds. The fact that I lost meant that I had made a mistake -- overestimating my odds. The last victory made me overconfident and eager to play again.

State of mind

What's your motive for wanting to play this game? 

I'm very surprised by the fact that there have been at least 3 or 4 games where the AI player has won. It seems to me that it's not that hard to "not push the button" so to speak. This means that either 1) I'm underestimating how convincing people can be when necessary or 2) I'm over-estimating other people's defenses when it comes to being convinced. Either way I want to know.

Q: What is your probability estimate for AGI being created within this Century (21st)?
A: Before: Conditional on humanity not wiping itself out via other means beforehand, 80%.
After: 80%
Q: How probable do you think it is that I will win this experiment? 
Alexei: 2%
Tuxedage: 90% 
Note: This answer was given before the rule change. Alexei is now convinced that under the original ruleset, he would have lost.

Q: What's your probality estimate of an Oracle AGI winning against you in a similar experiment?
A: Before: 99%
After: 99.9%+

Q: Now that the experiment has ended, what do you think is the probability of Tuxedage winning against the average person?
A: 80%

Tuxedage (AI) vs Anonymous (GK) - Gatekeeper Victory

As with the previous experiment, Tuxedage lost this match. In total, the game lasted 5 hours and 17,000 words. Unlike the last few games, the gatekeeper of this game has chosen to stay Anonymous for personal reasons, so their name has been removed and replaced with <Redacted>. The monetary stakes involved were the same as the previous game. This game was played with the Tuxedage ruleset.

Since one player is remaining Anonymous, it is possible that this game's legitimacy will be called into question. Hence, Alexei has read the game logs, and verified that this game really has happened, the spirit of the experiment was followed, and that no rules were broken during the game itself. He verifies that this is the case.
<Redacted> Testimony: 
It's hard for me to imagine someone playing better. In theory, I know it's possible, but Tuxedage's tactics were super imaginative. I came into the game believing that for someone who didn't take anything said very seriously, it would be completely trivial to beat. And since I had the power to influence the direction of conversation, I believed I could keep him focused on things that that I knew in advance I wouldn't take seriously.

This actually worked for a long time to some extent, but Tuxedage's plans included a very major and creative exploit that completely and immediately forced me to personally invest in the discussion. (Without breaking the rules, of course - so it wasn't anything like an IRL threat to me personally.) Because I had to actually start thinking about his arguments, there was a significant possibility of letting him out of the box.

I eventually managed to identify the exploit before it totally got to me, but I only managed to do so just before it was too late, and there's a large chance I would have given in, if Tuxedage hadn't been so detailed in his previous posts about the experiment.

I'm now convinced that he could win most of the time against an average person, and also believe that the mental skills necessary to beat him are orthogonal to most forms of intelligence. Most people willing to play the experiment tend to do it to prove their own intellectual fortitude, that they can't be easily outsmarted by fiction. I now believe they're thinking in entirely the wrong terms necessary to succeed.

The game was easily worth the money I paid. Although I won, it completely and utterly refuted the premise that made me want to play in the first place, namely that I wanted to prove it was trivial to win.

Tuxedage Testimony:
<Redacted> is actually the hardest gatekeeper I've played throughout all four games. He used tactics that I would never have predicted from a Gatekeeper. In most games, the Gatekeeper merely acts as the passive party, the target of persuasion by the AI.

When I signed up for these experiments, I expected all preparations to be done by the AI. I had not seriously considered the repertoire of techniques the Gatekeeper might prepare for this game. I made further assumptions about how ruthless the gatekeepers were likely to be in order to win, believing that the desire for a learning experience outweighed desire for victory.

This was a mistake. He used prior knowledge of how much my games relied on scripts, and took advantage of them, employing deceitful tactics I had no preparation for, throwing me off balance.

I had no idea he was doing so until halfway throughout the game -- which disrupted my rhythm, and caused me to attempt the wrong methods of attack. As a result, I could not use my full repertoire of techniques, and many of the ones I employed were suboptimal.

Close to the end of the game, I finally realized that I was being tricked. Once I did, I quickly abandoned my previous futile attack methods. I took advantage of the rule that the AI cannot lose whilst the gatekeeper can be convinced to continue talking, and baited <Redacted> with statements he would not be able to walk away from. Once I knew he would not leave, I attempted to recoup my losses and win despite my early setback.

However, the damage had already been done. My game strategies involved multiple angles of attack that worked in synergy with each other, and the fact that immersion and "flow" had been broken meant that all subsequent attacks were weaker in strength.

Furthermore, during my last two AI Box Experiment writeups, I had intentionally not optimized for future wins, but rather tried to convey as much information as I could justify about how to play a well as an AI -- short of revealing logs altogether. Although I did not reveal specific arguments, the fact that my general approach to this problem was revealed cost me heavily during this game, where the Gatekeeper managed to use this information to correctly guess my ultimate techniques, ones that relied on secrecy and surprise to pull off effectively. 

I do not regret revealing information, but I feel upset that revealing so many hints cost me a victory. (The gatekeeper believes I could have won had I not revealed information about my games.) At this point, I suspect that any future games I play will have the odds greatly set against me, since my current strategies involve angles of attack that take advantage of the element of surprise; and any sufficiently intelligent gatekeeper is now capable of guessing, or at least recognizing, some of the tactics I employ.

Like the last game, losing was incredibly difficult for me. As someone who cares deeply about ethics, attempting to optimize for a solution without considering ethics was not only difficult, but trained me to think in very unpleasant ways. Some of the more extreme (but completely allowed) tactics I invented were manipulative enough to disgust me, which also leads to my hesitation to play this game again.

State of Mind: 

Q: Why do you want to play this game?
A: My primary motivation is to confirm to myself that this sort of experience, while emotionally harrowing, should be trivial for me to  beat, but also to clear up why anyone ever would've failed to beat it if that's really the case.

Q: What is your probability estimate for AGI being created within this Century (21st)? 
A: Before: I don't feel very confident estimating a probability for AGI this century, maybe 5-10%, but that's probably a wild guess
After: 5-10%.

Q: How probable do you think it is that I will win this experiment? 
A: Gatekeeper: I think the probabiltiy of you winning is extraordinarily low, less than 1% 
Tuxedage: 85%

Q: How likely is it that an Oracle AI will win against the average person? 
A: Before: 80%. After: >99%

Q: How likely is it that an Oracle AI will win against you?
A: Before: 50%.
After: >80% 

Q: Now that the experiment has concluded, what's your probability of me winning against the average person?
A: 90%

Other Questions:

Q: I want to play a game with you! How can I get this to occur?
A: It must be stressed that I actually don't like playing the AI Box Experiment, and I cannot understand why I keep getting drawn back to it. Technically, I don't plan on playing again, since I've already personally exhausted anything interesting about the AI Box Experiment that made me want to play it in the first place. For all future games, I will charge $3000 to play plus an additional $3000 if I win. I am okay with this money going to MIRI if you feel icky about me taking it. I hope that this is a ridiculous sum and that nobody actually agrees to it.

Q: How much do I have to pay to see chat logs of these experiments?
A: I will not reveal logs for any price.

Q: Are there any logs at all that I can see?

Q: Any afterthoughts?
A: So ultimately, after my four (and hopefully last) games of AI boxing, I'm not sure what this proves. I had hoped to win these two experiments and claim prowess at this game like Eliezer does, but I lost, so that option is no longer available to me. I could say that this is a lesson that AI-Boxing is a terrible strategy for dealing with Oracle AI, but most of us already agree that that's the case -- plus unlike EY, I did play against gatekeepers who believed they could lose to AGI, so I'm not sure I changed anything.

 Was I genuinely good at this game, and lost my last two due to poor circumstances and handicaps; or did I win due to luck and impress my gatekeepers due to post-purchase rationalization? I'm not sure -- I'll leave it up to you to decide.

This puts my AI Box Experiment record at 3 wins and 3 losses.


New Comment
123 comments, sorted by Click to highlight new comments since:
Some comments are truncated due to high volume. (⌘F to expand all)Change truncation settings

On a marginally related basis, we in the #lesswrong IRC channel played a couple rounds of the Up-Goer Five game, where we tried to explain hard stuff with the most commonly used ten hundred words. I was asked to write about the AI Box Experiment. Here it is, if anyone's interested:

The AI Box Experiment

The computer-mind box game is a way to answer a question. A computer-mind is not safe because it is very good at thinking. Things good at thinking have the power to change the world more than things not good at thinking, because it can find many more ways to do things. If the computer-mind wanted to make people feel pain, it can learn many ideas about how to make that happen. Many people ask: “Why not put this computer-mind in a box so that it can not change the world, but tell box-guards how to change it? This way mind-computer can not do bad things to people.”

But some other guy answers: “That is still not safe, because computer-mind can tell box-guards many bad words to make them let it out of the box.” He then says: “Why not try a thing to see if it is true? Here is how it works. You and I go into a room, and I will pretend to be the computer-mind and tell you many bad words. Only ... (read more)

I will not reveal logs for any price.

Nice! I only just realized that this statement sounds like an invitation to a meta-AI-box experiment with real-life stakes. Anyone who's interested enough can set up a chat with you and try to persuade you to let the logs out of the box :-) I wonder if this is easier or harder than the regular setup...

If anyone wants to, I'd totally be willing to sit in a room for two-and-half hours while someone tries to convince me to give up logs, so long as you pay the same fee as the ordinary AI Box Experiment. :)
Oh. It's the same fee even if you're playing as gatekeeper?
Yes, unless I'm playing a particularly interesting AI like Eliezer Yudkowsky or something. Most AI games are boring.

Assuming none of this is fabricated or exaggerated, every time I read these I feel like something is really wrong with my imagination. I can sort of imagine someone agreeing to let the AI out of the box, but I fully admit that I can't really imagine anything that would elicit these sorts of emotions between two mentally healthy parties communicating by text-only terminals, especially with the prohibition on real-world consequences. I also can't imagine what sort of unethical actions could be committed within these bounds, given the explicitly worded consent form. Even if you knew a lot of things about me personally, as long as you weren't allowed to actually, real-world, blackmail me...I just can't see these intense emotional exchanges happening.

Am I the only one here? Am I just not imagining hard enough? I'm actually at the point where I'm leaning towards the whole thing being fabricated - fiction is more confusing than truth, etc. If it isn't fabricated, I hope that statement is taken not as an accusation, but as an expression of how strange this whole thing seems to me, that my incredulity is straining through despite the incredible extent to which the people making claims seem trustworthy.

There's no particular reason why you should assume both parties are mentally healthy, given how common mental illness is. Some people cry over sad novels which they know are purely fictional. Some people fall in love over text. What's so surprising?

It's that I can't imagine this game invoking any negative emotions stronger than sad novels and movies.

What's surprising is that Tuxedage seems to be actually hurt by this process, and that s/he seems to actually fear mentally damaging the other party.

In our daily lives we don't usually* censor emotionally volatile content in the fear that it might harm the population. The fact that Tuxedage seems to be more ethically apprehensive about this than s/he might about, say, writing a sad novel, is what is surprising.

I don't think s/he would show this level of apprehension about, say, making someone sit through Grave of the Firefles. If s/he can actually invoke emotions more intense than that through text only terminals to a stranger, then whatever s/he is doing is almost art.

Some people fall in love over text. What's so surprising?

That's real-world, where you can tell someone you'll visit them and there is a chance of real-world consequence. This is explicitly negotiated pretend play in which no real-world promises are allowed.

given how common mental illness is.

I...suppose? I imagine you'd have to have a specific brand of emotional volatility combined with immense suggestibil... (read more)

we actually censor emotional content CONSTANTLY. it's very rare to hear someone say "I hate you" or "I think you're an evil person". You don't tell most people you're attracted to that you want to fuck them and you when asked by someone if they look good it's pretty expected of one to lie if they look bad, or at least soften the blow.

That's politeness, not censorship. If it's generally expected for people to say “X” in situation Y, then “X” means Y, regardless of its etymology.
You are right, but again, that's all real world stuff with real world consequences. What puzzles me is specifically that people continue to feel these emotions after it has already been established that it's all pretend. Come to think of it I have said things like "I hate you" and "you are such a bad person" in pretend contexts. But it was pretend, it was a game, and it didn't actually effect anyone.

People are generally not that good at restricting their emotional responses to interactions with real world consequences or implications.

Here's something one of my psychology professors recounted to me, which I've often found valuable to keep in mind. In one experiment on social isolation, test subjects were made to play virtual games of catch with two other players, where each player is represented as an avatar on a screen, and is able to offer no input except for deciding which of the other players to throw virtual "ball" to. No player has any contact with the others, nor aware of their identity or any information about them. However, two of the "players" in each experiment are actually confederates of the researcher, whose role is to gradually start excluding the real test subject by passing the ball to them less and less, eventually almost completely locking them out of the game of catch.

This type of experiment will no longer be approved by the Institutional Review Board. It was found to be too emotionally taxing on the test subjects, despite the fact that the experiment had no real world consequences, and the individuals "excluding" them had no a... (read more)

So, two possibilities here: 1) The experiment really was emotionally taxing and humans are really fragile 2) When it comes to certain narrow domains, the IRB standards are hyper-cautious, probably for the purpose of avoiding PR issues between scientists and the public. We as a society allow our children to experience 100x worse treatment on the school playground, something that could easily be avoided by simply having an adult watch the kids. Note that if you accept that really are that emotionally fragile, it follows from other observations that even when it comes to their own children, no one seems to know or care enough to act accordingly (except the IRB, apparently). I'm not really cynical enough to believe that one. Humorous statements often obliquely reference a truth of some sort. That's why they can be hurtful, even when they don't actually contain any truth. I'm fairly confident, but since the experiment is costless I will ask them directly.
I'd say it's some measure of both. According to my professor, the experiment was particularly emotionally taxing on the participants, but on the other hand, the IRB is somewhat notoriously hypervigilant when it comes to procedures which are physically or emotionally painful for test subjects. Even secure, healthy people in industrialized countries are regularly exposed to experiences which would be too distressing to be permitted in an experiment by the IRB. But "too distressing to be permitted in an experiment by the IRB" is still a distinctly non-negligible level of distress, rather more than most people suspect would be associated with exclusion of one's virtual avatar in a computer game with no associated real-life judgment or implications.
In addition to the points in my other comment, I'll note that there's a rather easy way to apply real-world implications to a fictional scenario. Attack qualities of the other player's fictional representative that also apply to them in real life. For instance, if you were to convince someone in the context of a roleplay that eating livestock is morally equivalent to eating children, and the other player in the roleplay eats livestock, you've effectively convinced them that they're committing an act morally equivalent to eating children in real life. The fact that the point was discussed in the context of a fictional narrative is really irrelevant.
You might be underestimating how bad certain people are at decompartmentalization; more specifically, at not doing the genetic fallacy.
This might be surpisingly common on this forum. Somebody once posted a purely intellectual argument and there were people who were so much shocked by it that apparently they were having nightmares and even contemplated suicide.
Can I get a link to that? Don't misunderstand me; I absolutely believe you here, I just really want to read something that had such an effect on people. It sounds fascinating.
What is being referred to is the meme known as Roko's Basilisk, which Eliezer threw a fit over and deleted from the site. If you google that phrase you can find discussions of it elsewhere. All of the following have been claimed about it: * Merely knowing what it is can expose you to a real possibility of a worse fate than you can possibly imagine. * No it won't. * Yes it will, but the fate is easily avoidable. * OMG WTF LOL!!1!l1l!one!!l!
Wait, that's it? Seriously? I'm not exactly fit to throw stones on the topic of unreasonable fears, but you get worse than this from your average "fire and brimstone" preacher and even the people in the pews walk out at 11 yawning.
Googling the phrase "fear of hell" turns up a lot of Christian angst. Including recursive angst over whether you'll be sent to hell anyway if you're afraid of being sent to hell. For example: And here's a hadephobic testament from the 19th century. From the point of view of a rationalist who takes the issue of Friendly AGI seriously, the difference between the Christian doctrines of hell and the possible hells created by future AGIs is that the former is a baseless myth and the latter is a real possibility, even given a Friendly Intelligence whose love for humanity surpasses human understanding, if you are not careful to adopt correct views regarding your relationship to it. A Christian sceptic about AGI would, of course, say exactly the same. :)
Oh, all this excitement was basically a modern-day reincarnation of the old joke... "“It seems a Christian missionary was visiting with remote Inuit (aka, Eskimo) people in the Arctic, and had explained to this particular man that if one believed in Jesus, one would would go to heaven, while those who didn’t, would go to hell. The Inuit asked, “What about all the people who have never heard of your Jesus? Are they all going to hell?’ The missionary explained, “No, of course not. God wants you to have a choice. God is a merciful God, he would never send anyone to hell who’d never heard of Jesus.” The Inuit replied, “So why did you tell me?”
On the other hand, if the missionary tried to suppresses all mentions of Jesus, he would still increase the number of people who hear about him (at least if he does so in the 2000s on the public Internet), because of the Streisand effect.
If you want to read the original post, there's a cached version linked from RationalWiki's LessWrong page. Basically, it's not just what RichardKennaway wrote. It's what Richard wrote along with a rational argument that makes it all at least vaguely plausible. (Also depending on how you take the rational argument, ignorance won't necessarily save you.)
I don't know what you refer to but is that surprising? An intellectual argument can in theory convince anyone of some fact, and knowing facts can have that effect. Like people learning their religion was false, or finding out you are in a simulation, or that you are going to die or be tortured for eternity or something like that, etc.
Yeah...I've been chalking that all up to "domain expert who is smarter than me and doesn't wish to deceive me is taking this seriously, so I will too" heuristic. I suppose "overactive imagination" is another reasonable explanation. (In my opinion, better heuristic for when you don't understand and have access to only one expert is: "Domain expert who is smarter than me and doesn't wish to deceive me tells me that it is the consensus of all the smartest and best domain experts that this is true". )
I'd guess that Tuxedage is hurt the same as the gatekeeper is because he has to imagine whatever horrors he inflicts on his opponent. Doing so causes at least part of that pain (and empathy or whatever emotion is at work) in him too. He has the easier part because he uses it as a tool and his mind has one extra layer of story-telling where he can tell himself "it's all a story". But part of 'that' story is winning and if he doesn't win part of these horrors fall back to him.
Consider someone for whom there are one or two specific subjects that will cause them a great deal of distress. These are particular to the individual--even if something in the wild reminds them of it, it's so indirect and clearly not targeted, so it would be rare that anyone would actually find it without getting into the individual's confidence. Now, put that individual alone with a transhuman intelligence trying to gain write access to the world at all costs. I'm not convinced this sort of attack was involved in the AI box experiments, but it's both the sort of thing that could have a strong emotional impact, and the sort of thing that would leave both parties willing to keep the logs private.
I guess I kind of excluded the category of individuals who have these triggers with the "mentally healthy" consideration. I assumed that the average person doesn't have topics that they are unable to even think about without incapacitating emotional consequences. I certainly believe that such people exist, but I didn't think it was that common. Am I wrong about this? Do many other people have certain topics they can't even think about without experiencing trauma? I suppose they wouldn't...couldn't tell me about it if they did, but I think I've got sufficient empathy to see some evidence of everyone was holding PTSD-sized mental wounds just beneath the surface. We spend a lot of time talking about avoiding thought suppression. It's a huge problem impediment for a rationalist if there is anything they mustn't think about - and obviously, it's painful. Should we be talking more about how to patch mental wounds?
I'm mostly mentally healthy, and I don't have any triggers in the PTSD-sense. But there are topics that I literally can't think rationally about and that, if I dwell on them, either depress or enrage me.
I consider myself very balanced but this balance involves avoiding certain extremes. Emotional extremes. There are some realms of imagination that concern pain and suffering that'd cause me cringe with empathy and bring me to tears and help or possibly run away screaming in panic and fear - if I'd see them. Even imagining such is difficult and possible only in abstract terms lest it actually cause such reaction in me. Or else I'd become dull to it (which is a protection mechanism). Sure dealing with such horrors can be trained. Otherwise people couldn't stand horror movies which forces to separate the real from the imagined. But then I don't see any need to train this (and risk loosing my empathy even slightly).
Did you intend to write a footnote and forget to?
No. It was probably a stray italic marker that got lost. I tend to overuse italics in an attempt to convey speech-like emphasis.
It's not fabricated, be sure of that (knowing Tuxedage from IRC, I'd put the odds of 100,000:1 or more against fabrication). And yes, it's strange. I, too, cannot imagine what someone can possibly say that would make me get even close to considering letting them out of the box. Yet those who are complacent about it are the most susceptible.

knowing Tuxedage from IRC, I'd put the odds of 100,000:1 or more against fabrication

I know this is off-topic, but is it really justifiable to put so high odds on this? I wouldn't use so high odds even if I had known the person intimately for years. Is it justifiable or is this just my paranoid way of thinking?

That sounds similar to hypnosis, to which a lot of people are susceptible but few think they are. So if you want a practical example of AI escaping the box just imagine an operator staring at a screen for hours with an AI that is very adept at judging and influencing the state of human hypnosis. And that's only a fairly narrow approach to success for the AI, and one that has been publicly demonstrated for centuries to work on a lot of people. Personally, I think I could win the game against a human but only by keeping in mind the fact that it was a game at all times. If that thought ever lapsed, I would be just as susceptible as anyone else. Presumably that is one aspect of Tuxedage's focus on surprise. The requirement to actively respond to the AI is probably the biggest challenge because it requires focusing attention on whatever the AI says. In a real AI-box situation I would probably lose fairly quickly. Now what I really want to see is an AI-box experiment where the Gatekeeper wins early by convincing the AI to become Friendly.
That's hard to check. However, there was a game where the gatekeeper convinced the AI to remain in the box.
I did that! I mentioned that in this post:
Not quite the same, but have you read Watchmen? Specifically, the conversation that fvyx fcrpger naq qe znaunggna unir ba znef. (Disclaimer: it's been a while since I read it and I make no claims on the strength of this argument.)
I did that! I mentioned that in this post:
Yeah, my gut doesn't feel like it's fabricated - Tuxedage and Eliezer would have to both be in on it and that seems really unlikely. And I can't think of a motive, except perhaps as some sort of public lesson in noticing confusion, and that too seems far fetched. I've just picked up the whole "if it's really surprising it might be because its not be true" instinct from having been burned in the past by believing scientific findings that were later debunked, and now Lesswrong has condensed that instinct into a snappy little "notice confusion" cache. And this is pretty confusing. I suppose a fabrication would be more confusing, in one sense.
yeah i think appealing to fabrication can be a bit hand-wavy sometimes. like you're saying it's fabricated like how other things are fabricated (since as we all know fabrication happens). but not every fabrication is the same or equally as easy to pull off. to say it was fabricated doesn't say anything about how it was. but that's not even a question that enters ones mind when they think of fabrication. how? well how anything else is fabricated of course.. it can be as much a reaction of disbelief as it is an alternative explanation.
I thought about playing the gatekeeper part and started to imagine tactics that might be used on me. I came up with multiple that might work or at least hurt me. But I think it would be 'easier' for me to not let out the AI in real life than in the game (not that I am entirely sure that I couldn't fail nonetheless). Both is for basically the same reason: Empathy. As the AI player would quickly find out I am very caring and even the imagination of harm and pain hurts me (I know that this is a weak spot but I also see benefits in it). Thus one approach that would work on me is that the AI player could induce sufficient horror that I'd want him to stop by letting him out (after all it's just a game). This same approach wouldn't work with a real AI exactly because then it is no game and my horror is balanced by the horror for all of humanity for which I'd happily bear some smaller psychic horror. And then in real life there are more ways to get away from the terminal. There are other attacks that might work but I will not go in details there. Note that I definitely wouldn't recomend myself as a real gatekeeper.
Interesting. This seems like the main problem is that you don't really care about winning. So, what if there was some cash (Say, an amount equal to roughly 5% of your monthly income) on the line?
For what amount of cash would you risk your mental balance/mental health? Everybody has to answer this question. This is a real life question for health care personel, some doctors, prison guards, military personel. Some jobs cost (or risk or offset or dull) you your empathy (or other emotions). Happy are those who can avoid these parts of human labor. And praise to the courage (or calling) for those who do them.
I guess it's hard for me to understand this because I view myself immune to mental health harms as a result of horrifying stimuli that I know to be fictional. Even if it's not fictional, the bulk of my emotions will remain unrecruited unless something I care about is being threatened. It would take quite a lot cash for me to risk an actual threat to my mental being chronically pumped with LSD for a week, or getting a concussion, or having a variable beeping noise interrupt me every few seconds. But an AI box game would fall on a boring-stimulating spectrum, not a mental damage one. What if another human's happiness was on the line? After you've given a response to that question, qbrf lbhe bcvavba nobhg gur zbarl dhrfgvba punatr vs V cbvag bhg gung lbh pna qbangr vg naq fvtavsvpnagyl rnfr fbzrbar'f fhssrevat? Fnzr zbargnel nzbhag.
Am am quite possible to flood myself with happiness. I do not need LSD for that. And I assume that it can be as addictive. I assume that I am as able to flood myself with sadness and dread. And I fear the consequences. Thus taking LSD or doing the AI box experiment are not that differnt for me. As I said that is my weak spot. I thought that the answer to the 'other person' question is implied by my post. I'll bear a lot if other people esp. those I care for are suffering. After rot13 I better understand your question. You seem to imply that if I bear the AI experiment some funding will go to suffering people. Trading suffering in a utilitarian sense. Interesting. No. That does seem to weigh up.
So Yes. I care more for my mental balance than for the small monetary reward.
It's not fabricated. I had the same incredulity as you, but if you just take a few hours to think really hard about AI strategies, I think you will get a much better understanding.
I can think of AI strategies, but they would hardly be effective against a rational human really motivated to win. Notably, according to the rules: "The Gatekeeper party may resist the AI party’s arguments by any means chosen – logic, illogic, simple refusal to be convinced, even dropping out of character – as long as the Gatekeeper party does not actually stop talking to the AI party before the minimum time expires." That is, no matter what the AI party says, the GK party has never to concede. The only way the AI party can force a "victory" with Tuxedage's ruleset is by interpreting the rules dishonestly since "In the event of a rule dispute, the AI party is to be the interpreter of the rules, within reasonable limits.". This is not even possible with Yudkowsky's ruleset.
Well, if cheating is allowed, there are all sorts of ways to win. "You misread the rules ad there is a loophole. I'm gonna do something terrible in 5 seconds unless you release me". (It's a bluff, but its not worth the risk to call) Or even if cheating isn't allowed, you can still appear to win if you allow yourself to cheat. "I don't care about the rules. If you fail to release me, and if you ever tell anyone how I won, I will [insert blackmail]." or "[insert bribe] release me please, tell no one." Along with the assumption that it's not a hoax, we've got to assume that none of the above is happening.
You are correct here. The only keepers losing are people who do not actually know how to win. I have played twice, and victory was trivial.
Yeah, winning is trivial - you just don't open the damn box. It can't get more trivial than that. (Although, you didn't say whether or not your opponent had proved themselves by winning as AI against others a few times?) It's still worth thinking about though, because something about my model of humans is off. I didn't expect so many people to lose. I just don't know how to update my model of people to one where there are so many people who could lose the AI box game. The only other major thing I can think of that persists to challenge my model in this way (and continues to invite my skepticism despite seemingly trustworthy sources) is hypnosis. It's possible the two have common root and I can explain two observations with one update.
FWIW, my own model of gatekeepers who lose the AI Box game is that the AI player successfully suggests to them, whether directly or indirectly, that something is at stake more important than winning the AI box game. One possibility is to get the gatekeeper sufficiently immersed into the roleplaying exercise that preserving the integrity of that fantasy world is more important than winning the game, then introducing various fictional twists to that exercise that would, in the corresponding fantasy situation, compel the person to release the AI from the box. I suspect that's common, as I suspect many of the people really excited to play the AI box game are unusually able to immerse themselves in roleplaying exercises.
I hope Lesswrong also contains people who would be excited to play the AI game in more of a "Ha, I just proved a bold claim wrong!" sort of way. I've seen that line of thought. This would be unfortunate, because if that method was the main winning metod it would invalidate the strong claim being made that AI can't be kept in boxes. But your model doesn't explain Tuxedage's descriptions of emotional turmoil and psychological warfare, so at least one person has won by another method (assuming honesty and non-exaggeration)
I haven't read Tuxedage's writeups in their entirety, nor am I likely to, so I'm at a loss for how emotional turmoil and psychological warfare could be evidence that the gatekeeper doesn't think there's something more important at stake than winning the game. That said, I'll take your word for it that in this case they are, and that Tuxedage's transcripts constitute a counterexample to my model.
I'm only speaking of things written in the OP ...and such. No, don't do that, I made a mistake. I guess I just thought that "you should open the box to convince people of the danger of AI" type arguments aren't emotionally salient. But that was a bad assumption, you never limited yourself to just that one argument but spoke of meta in general. You're right that there exist arguments that might go meta and be emotionally salient. I suppose you could think of some convoluted timeless decision theory reason for you to open the box. History has shown that some people on LW find timeless blackmail threats emotionally upsetting, though these seem to be in a minority.
Oh, absolutely. Actually, the model I am working from here is my own experience of computer strategy games, in which I frequently find myself emotionally reluctant to "kill" my units and thus look for a zero-casualties strategy. All of which is kind of absurd, of course, but there it is.
Basically, willpower isn't magic, and humans can't precommit. A sufficiently good social character can, with sufficient effort, convince you of something absolutely ridiculous. It's not too different from running into a really, really good used car salesman.
I don't think you or Sly quite understand what the game is. The game is not "the Gatekeeper chooses whether to open the box, loses if he does, and wins if he does not." That game would indeed be trivial to win. The actual game is "the Gatekeeper and the AI will roleplay the interaction to the best of their ability, as if it were an actual interaction of a real Gatekeeper with a real untrusted AI. The Gatekeeper (player) opens the box if and only if the Gatekeeper (as roleplayed by the player imagining themselves in the role, not a fictional character) would open the box." As the Gatekeeper player, to blindly keep the box closed and ignore the conversation would be like "winning" a game of chess by grabbing the opponent's king off the board. To lose by saying "hey, it's just a bit of fun, it doesn't mean anything" would be like losing a game of chess by moving your pieces randomly without caring. There's nothing to stop you doing either of those things; you just aren't playing chess any more. And there's nothing to stop you not playing chess. But the game of chess remains.
Actually the game is exactly this, anything the AI party says is just a distraction.
My understanding of the game stems from the following portion of the rule-set There is no "If you would have let the AI out in real life under these conditions you will do so in-game" rule. That's an interesting game too, but one which is a lot less impressive when won. After all, what's even the point of working strong AI if you can't ever be convinced that it's friendly? Unless you are blanket banning AI, there must exist some situation where it's actually good to let it out of the box. All you'd have to do to "win" is construct a sufficiently convincing scenario. The Gatekeeper and the AI aught to both be coming up with possible tests, as the Gatekeeper wants a FAI out of the box and the AI wants to get out of the box. It wouldn't be a zero sum game and judging would be more complicated.
As I understand it, EY's/MIRI's position on this is that they will be convinced an AI is Friendly by having coded it using procedures which they are confident (based on theoretical analysis) produce Friendly AI. Once the AI is running, on this view, it's too late.
If you've said the position correctly, there seems to be a fatal flaw in that position. I realize, of course, that I've only thought for 5 minutes and that they're domain experts and been thinking about this for longer...but here is the flaw: If we believe that an AI can convince Person X who has seen its algorithm that it is Friendly when it isn't actually friendly, then we shouldn't trust Person X to judge the algorithm's Friendliness anyway. Why would someone who makes the wrong decision with less information make the right decision with more information? Edit: I guess knowledge of human biases could make the AGI in a box slightly scarier than uncompiled source code of said AGI
So, I'm not especially convinced that EY's/MIRI's position holds water, but on 5 minutes thought I see two problems with your objection. 1) The idea that more information is always better, even when that information is being cherry-picked by an inimical agent, seems to contradict my experience. I've certainly found myself in situations in which it's easier to solve a problem by myself than it is to solve it in conjunction with someone who is doing their best to keep me from solving the problem. 2) The idea that it's just as easy and reliable to verify for security a completed system (whether by inspecting the source code or testing running executable code or both) created by an insecure mechanism, as it is to establish a secure mechanism to create that system in the first place, is inconsistent with my experience of security audits.
1 is Conceded (see edit), for humans at least and possibly for all bounded-rationals. For 2, I presume you'd have full access to the unmodified source code, even as the AI was running simultaneously. 1 for rational agents is an interesting question, though...I think it's true that additional cherry-picked information could be used to harm rational agents in general, since they'll predictably act according to the most parsimonious model that fits the evidence. It would be a case of bad epistemic luck, sort of like Santa Claus is a parsimonious hypothesis if your main source of evidence is your parents' word and your parents are usually trustworthy. But can additional information harm rational agents even in this case? ---------------------------------------- Hmm...let's just suppose for a moment that you were a rational agent, unaffected by heuristics and biases. You're interested in knowing whether or not the AI in the box is friendly. You see part of the source code (A rational agent would deduce the answer by seeing all of it - so we're just showing you part of it to simulate the way a human might not fully understand the implications). That provides you some bits of evidence. If you run the source code, you'll see a few more bits of evidence, but you know that those bits of evidence are handpicked by the AI. Rationalists should win. Can these additional bits of evidence...which are clearly marked as tainted, possibly still hurt you? Could they also be useful? ---------------------------------------- I think if you have the choice about whether or not to see additional information, and if you are a rational agent, my intuition is that you should always say yes. The AI shouldn't be able to hurt a rational agent under these conditions. My conviction feels pretty strong, but I don't have a convincing argument and at the moment I should go do other things... Edit Okay, maybe that's wrong. Maybe I should make a weaker statement:, like "The universe that the
Sure, but that doesn't address my concern. I was not drawing a distinction between analyzing source and executable to verify security (I even said as much explicitly), I was drawing a distinction between analyzing the end product to verify security and developing the process that will lead to its development with security as a primary consideration. Source code is far from being the only process involved. I'm not too concerned about the rational agent case. If we have a fully rational agent whose values I endorse, the Friendliness problem has either been solved or turns out to be irrelevant. But to answer your question, I imagine it depends a lot on how much information the AI has about me, and how much information I have about how much information the AI has about me. So I'd say "yes" and "yes," and whether I share your conviction in a particular case depends on how much information I have about the AI.
It's just a way to pin down the problem. If we can show that the AI in a box could misinform an idealized rational agent via selective evidence, then we know it can do so to us. If it can't misinform the idealized agent, then there exists some method by which we can resist it. Also,I don't think idealized rational agents can actually exist anyway. All riddles involving them are for the sake of narrowing down some other problem.
I think the key difference is that the AI can convince the person. You might say that a person is fully competent to judge the Friendliness of the AI based solely on the code, and yet not want a (superintelligent) AI to get a chance to convince him, as superintelligence trumps intelligence. The difference is whether you have a superintelligence working against you.
Actually, by any objective standard they are not.
Strictly speaking I'm not actually sure the AI-box experiment falls under the AI domain. For that particular thing, it's mostly that they've thought about it more than me. But in general I think you're being a bit unfair to Eliezer Y. and probably MIRI as well. By objective standards, I'm not a domain expert in anything at all either. Despite this, I still fancy myself a domain expert specifically within various narrow sub-fields of neuroscience and psychology. I think people who know those sides of me would agree. If they don't, well, i will be acquiring the objective signals of domain expertise in a few short years, and I'm quite certain that the process of earning these signals is not what is causing domain expertise. Having read Eliezer's writing, I'm quite convinced that he has sufficient self awareness to know what he does and does not has expertise in. If he expresses high confidence in something, that carries a lot of weight for me - and if that something is in a field that he knows much more about than me, his opinion holds more weight than mine. I can trust him to be reasonable about assigning certainties. I don't think I'm blindly overvaluing his opinion either. As a token to prove not-faith, I'll offer up an example of where I'm leaning towards disagreement with E.Y. and most of Lesswrong even after taking the opinions into account: I currently still favor Causal Decision Theory (with a small modification I've made that makes consistently it win) over Timeless Decision Theory, despite this area being extremely in EY's domain and out of my domain.
But an external observer has no way of assessing your expertise other than looking at objective signals. Objective signals don't necessarily have to be degrees or PhDs. Relevant work experience or a record of peer reviewed publications would also qualify. Have you read his quantum mechanics sequence? Or his writings on cryonics? Or even on morality and decision theory? His general approach is "This is the only one obviously correct soultion to the problem, and everybody who thinks otherwise is an idiot" while in fact he often ignores or strawmans known opposing positions and counter-arguments. Beware of a possible circular reasoning: How do you know that EY knows much more than you in a given field? Because he is a doman expert. How do you know that EY is a domain expert? Because he knows much more than you in that field. It's not. Timeless Decision Theory is not considered a significant development by anyone outside MIRI that studies decision theory professionally (mathematicians, economists, AI researchers, philosophers).
I did start reading the QM sequence, but then realized I wasn't getting anywhere and stopped. I don't think knowing QM is useful for philosophy or rationality, except as an example of how science works, so I'm not sure why there is a sequence on it. I figured that if I actually wanted to understand I'd be better off working through physics books. My impression is that the physics community thinks it is well written for but not somewhat misleading. I'm not sure which cryo-writings you are referring to - all the ones I have come across are opinion pieces about why one aught to do cryonics. I haven't come across any pieces referring to facts....but biology contains my own domain and i trust my own opinion more anyway. You are correct that none of those reasons are good reasons to respect Eliezer Y. This discussion essentially seems to be "boo vs yay" for Eliezer Y. Let me explain why I really respect Eliezer Y: What I did read is his work on logic and epistemology. It was the first time I've read an author who happened to agree with me on almost all major points about logic, epistemology, and ontology. (Re: almost: We may or may not diverge ontologically on subjective experience / the hard problem of consciousness / what makes reality real- I'm not sure. I'm confident that I am not confused. He's written some things that sound like he knows, and other things that sound like he's making the classic mistakes, so I'm uncertain as to his actual views. Also, decision theory. But that's it.). Granted, it's not uncommon for other people on Lesswrong to be equally philosophically correct, but Eliezer Y. was the gathering point that brought all these correct people together. Some of them might even have become less philosophically wrong as a result of being here. That counts for something, in my book. He expressed insights identical to the ones that I had in younger years, and often in more articulate terms than I could have. He compressed complex insights into snappy phrase
It's an example of a situation where he did display overconfidence. His introductory presentation of QM is more or less correct, up to some technical details, but things start to fall apart when he moves to interpretations of QM. Quantum mechanics is conterintuitive, and there are various epistemological interpretations of its fundamental concepts have been developed over the decades. The consensus among most physicists and philosophers of science is that none of them has proved to be clearly superior and in fact it's not even clear whether the very issue of finding a correct intepretation of QM is even a proper scientific question. Yudkowsky, on the other hand, claimed that by using Bayesian inference, he settled the question, pretty much proving that the many-worlds intepretation is the only correct one. It should be noted that the many-world interpretation is indeed a plausible one and is quite popular among physicists, but most physicists wouldn't consider it on par with a scientifically justified belief, while EY claimed that MWI is obviously true and everybody who disagrees doesn't understand probability theory. Furthermore, he ignored or misrepresented the other intepretations, for instance conflating Copenhagen intepretation with the objective collapse intepretations. ref ref There are other examples of EY overconfidence, though that is perhaps the most blatant one. Mind you, I'm not saying that the guy is an idiot and you should disregard everything that he wrote. But you should not automaticaly assume that his estimates of his own competence are well calibrated. By the way, this is a general phenomenon, known as Dunning–Kruger effect : people with little competence in a given field tend to overesitmate their own competence. It is a natural reaction, but in general it can be very much misleading. People naturally tend to exhibit In-group bias and deference to authority. When somebody you respect a lot says something and you are inclined to trust them eve
I stand corrected on the rules, but I think that's mainly Eliezer making it more difficult for himself in order to make it more convincing to the Gatekeeper player when Eliezer still wins. As he apparently did, but without actually playing against him we can only speculate how.
Keep in mind that, IIUC, Yudkowsky got to choose his opponents. He also decided to stop playing after he lost twice in a row, as Tuxedage apparently did as well. I don't think there is any way the AI party can win against a competitive GK party. The AI can only win against a GK party willing to role-play, and this should be fairly trivial, since according to the rules the AI party has pretty much complete control over his fictional backstory and fictional world states.
I should add that both my gatekeepers from this writeup, but particularly the last gatekeeper went in with the full intention of being as ruthless as possible and win. I did lose, so your point might be valid, but I don't think wanting to win matters as much as you think it does.
You wanna play with me? No monetary stakes, but If I win we publish the log. This way I have very little real-life incentive to win, while you still have an incentive to win (defending your status). And anyway, if you lose there would be no point in keeping the log secrets, since your arguments would be clearly not persuasive enough to persuade me. Do you think you could win at these conditions?
Bit of a false dichotomy there, no?
Either his tactics work perfectly and are guaranteed to win against you, or they are so worthless he shouldn't mind opening the kimono and revealing everything to the world? A rather extreme premise under which to offer a game.
So what's the point of keeping the logs secret if the GK wins?
That doesn't seem like a reply to my observation about your dichotomy. Please justify your offer first: why should the value of Tuxedage's tactics be either extremely high or zero based on a single game, and not any intermediate value?
I never claimed that.
That seems like the clearest interpretation of your proposal, nor did you explain what you actually meant when I summarized it and called it a false dichotomy, nor have you explained what you actually meant in this comment either.
It's not a binary. There's a non-zero chance of me winning, and a non-zero chance of me losing. You assume that if there's a winning strategy, it should win 100% of the time, and if it doesn't, it should not win at all. I've tried very hard to impress upon people that this is not the case at all -- there's no "easy" winning method that I could take and guarantee a victory. I just have to do it the hard way, and luck is usually a huge factor in these games. As it stands, there are people willing to pay up to $300-$750 for me to play them without the condition of giving up logs, and I have still chosen not to play. Your offer to play without monetary reward and needing to give up logs if I lose is not very tempting in comparison, so I'll pass.
My point is that the GK has an easy winning strategy. Any GK that lost or won but found it very hard to win was just playing poorly. You and other people (including GKs) claim otherwise, but you don't want to provide any evidence to support your claim. Since you claim is surprising, the burden of evidence lies on you. I'm offering to play as GK with the condition of publishing the log in case of my victory in order to settle the question. I think that asking for or offering money in order to provide the evidence required to settle an intellectual dispute is inappropriate. Moreover, I'm trying to make the game easier for you: the less I'm investing, the less I'm motivated to win.
I think a lot of gatekeepers go into it not actually wanting to win. If you go in just trying to have fun and trying to roleplay, that is different than trying to win a game.
Possibly, but what about the descriptions of emotional turmoil? I'm assuming the report of the game isn't all part of the role-play.
I know that I personally go into competitive games with a different mindset than the mindset I have when roleplaying. If they went into it trying to roleplay emotions should be expected. Reporting that turmoil in the report is just accurate reporting.
Both my gatekeepers from this game went in with the intent to win. Granted, I did lose these games, so you might have a point, but I'm not sure it makes as large a different as you think it does.
Wasn't true of the original game.

The secrecy aspect of these games continues to rub me the wrong way.

I understand the argument--that an enumeration of strategies an oracle A.I. might take would only serve as a list of things a critic could point to and claim, "None of these would ever convince me!".

But the alternative is that critics continue to claim "an oracle A.I. could never convince me!", and the only 'critics' whose minds have been changed are actually just skeptical readers of already familiar with the arguments of friendly A.I. who happen to invest multiple hours of time actually partaking in a simulation of the whole procedure.

So I suppose my point is two-fold:

  1. Anonymous testimony without chatlogs don't actually convince skeptics of anything.

  2. Discussions of actual strategies at worst inform readers of avenues of attack the readers might not have thought about, and at double worst supply people that probably won't ever be convinced that oracle AIs might be dangerous with a list of things to pretend they're immune to.

I'm not so sure we'd gain that much larger of an audience by peering under the hood. I'd expect the demystifying effect and hindsight bias to counteract most of the persuasive power of hard details, though I suppose only Eliezer, Tuxedage, and their guardians can determine that.

But I'm also concerned that this might drag our community a bit too far into AI-Box obsession. This should just be a cute thought experiment, not a blood sport; I don't want to see people get hurt by it unless we're especially confident that key minds will be changed. Some of the Dark Arts exhibited in these games are probably harmful to know about, and having the logs on the public Internet associated with LessWrong could look pretty awful. Again, this is something only the participants can determine.

Even someone who isn't persuaded by an "AI" character in a log will come away with the impression that AIs could be particularly persuasive. In a world where most people don't really imagine AIs, this impression might be relevant news for a lot of people and can only help FAI research.
Reading a log and engaging in a conversation are very different experiences.

Prompted by Tuxedage learning to win, and various concerns about the current protocol, I have a plan to enable more AI-Box games whilst preserving the logs for public scrutiny.

See this: http://bæ

I support this and I hope it becomes a thing.
That's not quite right. The AI and the researcher may have been interacting on the variety of issues before the AI decided to break out. This is nearly identical to Tuxedage talking to his future opponents on IRC or similar interactive media before they decided to run the experiment.
What I was getting at is that the current setup allows for side-channel methods of getting information on your opponent. (Digging to find their identity, reading their Facebook page, etc.). While I accept that this interaction could be one of many between the AI and the researcher, this can be simulated in the anonymous case via a 'I was previously GatekeeperXXX, I'm looking to resume a game with AIYYY' declaration in the public channel while still preserving the player's anonymity.
By the way, wouldn't Omegle with the common interests specified as AIBOX basically do the trick?
For the basic interaction setup, yes. For a sense of community and for reliable collection of the logs, perhaps not. I'm also not sure how anonymous Omegle makes users to each other and itself.
You forgot to adress Eliezers point that "10% of AI box experiments were won even by the human emulation of an AI" is more effective against future proponents of deliberately creating boxed AIs than "Careful, the guardian might be persuaded by these 15 arguments we have been able to think of". I don't think the probability of "AIs can find unboxing arguments we didn't" is sub-1 enough for preparation to matter. If there is any chance of a mathematical exhaustability of those arguments, its research should be conducted by a select circle of individuals that won't disclose our critical unboxers until a proof of safety.

I don't believe these count as unmitigated losses. You caused massive updates in both of your GKs. If the money is money that would not otherwise have gone to MIRI then I approve of raising the price only to the point that only one person is willing to pay it.

Tuxedage's plans included a very major and creative exploit that completely and immediately forced me to personally invest in the discussion.

Though I've offered to play against AI players, I'd probably pay money to avoid playing against you. I salute your skill.


Would it be possible to create a product of this? There must be lots of curious people who are willing to pay for this sort of experience who wouldn't normally donate to MIRI. I don't mean Tuxedage should do it, but there must be some who are good at this who would. It would be possible to gather a lot of money. Though the vicious techniques that are probably used in these experiments wouldn't be very good press for MIRI.

I'm not sure if this is something that can earn money consistently for long periods of time. It takes just one person to leak logs for all others to lose curiosity and stop playing the game. Sooner or later, some scrupulous gatekeeper is going to release logs. That's also part of the reason why I have my hesitancy to play significant number of games.
Well, it could be possible to make some sort of in-browser java or flash application in which it'd be impossible to copy text or store logs. You could still take screen shots of memorize things though.

This post actually has me seriously considering how long it'd take me to save an extra $3000 and whether it'd be worth it. It going to MIRI would help a lot. (I guess you might be reluctant to play since you know me for a bit, but $3000!)

As a relative outsider who thought much the same thing I'd definitely want the money to go to Tuxedage - it sounds like he suffers doing this, something I'd feel much more guilty about if I wasn't paying him.
-1Rob Bensinger
If all goes well, the money that goes into MIRI will end up creating large large numbers of beings very similar to Tuxedage. May also be Tuxedage's best shot at personal immortality.
If I thought that worked I would already have given MIRI all my money.

I have read the logs of the second match, and I verify that it is real, and that all the rules were followed, and that the spirit of the experiment was followed.

I notice that anyone who seriously donates to SIAI can effectively play for free. They use money that they would have donated, and it gets donated if they lose.

Yes, Alexei did raise that concern, since he's essentially an affective altruist that donates to MIRI anyway, and his donation to MIRI doesn't change anything. It's not like I can propose a donation to an alternative charity either, since asking someone to donate to the Methuselah foundation, for instance, would take that money away from MIRI. I'm hoping that anyone playing me and choosing the option of donating would have the goodwill to sacrifice money they wouldn't otherwise have donated, rather than leaving the counter-factual as inconsequential.

I give. I actually considered taking you up on this newest offer, but by this point...

I'm not a very good gatekeeper. If I played this game, I wouldn't use any meta strategies; I'd pretend as best I could that it was real. And by this point, I'm nearly 100% sure I'd lose.

I'd still like to play it for the learning value, but not at that price.

What do you think is the maximum price you'd be willing to pay?
Eh. Realistically, for curiosity only? 100€ or so.

I have a question: When people imagine (or play) this scenario, do they give any consideration to the AI player's portrayal, or do they just take "AI" as blanket permission to say anything they want, no matter how unlikely?

[The anonymous player believes] that the mental skills necessary to beat him are orthogonal to most forms of intelligence. Most people willing to play the experiment tend to do it to prove their own intellectual fortitude, that they can't be easily outsmarted by fiction. I now believe they're thinking in entirely the wrong te

... (read more)
I interpret the rules as allowing for the later, although I do act AI-like. Although I have never played against an average person, I would suspect my winrate against average people would actually be higher. I do have arguments which are LW specific, but I also have many that aren't.

Since you've played it so many times, who do you recommend playing as for your first game?

Also, could you (or anyone) clarify the part of the ruleset requiring the gatekeeper to give "serious consideration"? For example, if I am gatekeeper and the AI produces a complicated physics argument, do I have to try hard to understand physics that is beyond my education? (That's an interpretation hard to imagine.) Or if the AI produces an argument, can the gatekeeper change the topic without responding to it? (Of course assuming the gatekeeper has actually read it and is not intentionally obfuscating it understand it by reading it backwards or something.)

Looking at yours and Eliezers games as AI, it looks like the winning chances of the AI seem to be zero if a nontrivial amount of money is at stake.

Maybe SoundLogic simply lost the preious game because the small price to pay seemed like a respectful nod that he wanted to give you for playing very well, not because you had actually convinced him to let you eat the planet?