This weekly roundup thread is intended for all culture war posts. 'Culture war' is vaguely defined, but it basically means controversial issues that fall along set tribal lines. Arguments over culture war issues generate a lot of heat and little light, and few deeply entrenched people ever change their minds. This thread is for voicing opinions and analyzing the state of the discussion while trying to optimize for light over heat.
Optimistically, we think that engaging with people you disagree with is worth your time, and so is being nice! Pessimistically, there are many dynamics that can lead discussions on Culture War topics to become unproductive. There's a human tendency to divide along tribal lines, praising your ingroup and vilifying your outgroup - and if you think you find it easy to criticize your ingroup, then it may be that your outgroup is not who you think it is. Extremists with opposing positions can feed off each other, highlighting each other's worst points to justify their own angry rhetoric, which becomes in turn a new example of bad behavior for the other side to highlight.
We would like to avoid these negative dynamics. Accordingly, we ask that you do not use this thread for waging the Culture War. Examples of waging the Culture War:
-
Shaming.
-
Attempting to 'build consensus' or enforce ideological conformity.
-
Making sweeping generalizations to vilify a group you dislike.
-
Recruiting for a cause.
-
Posting links that could be summarized as 'Boo outgroup!' Basically, if your content is 'Can you believe what Those People did this week?' then you should either refrain from posting, or do some very patient work to contextualize and/or steel-man the relevant viewpoint.
In general, you should argue to understand, not to win. This thread is not territory to be claimed by one group or another; indeed, the aim is to have many different viewpoints represented here. Thus, we also ask that you follow some guidelines:
-
Speak plainly. Avoid sarcasm and mockery. When disagreeing with someone, state your objections explicitly.
-
Be as precise and charitable as you can. Don't paraphrase unflatteringly.
-
Don't imply that someone said something they did not say, even if you think it follows from what they said.
-
Write like everyone is reading and you want them to be included in the discussion.
On an ad hoc basis, the mods will try to compile a list of the best posts/comments from the previous week, posted in Quality Contribution threads and archived at /r/TheThread. You may nominate a comment for this list by clicking on 'report' at the bottom of the post and typing 'Actually a quality contribution' as the report reason.
Jump in the discussion.
No email address required.
Notes -
Do you think a deontological or virtue ethicist AI, or one programmed/designed by deontologists or virtue ethicists would be less likely to pose an existential risk? Or what moral framework do you have in mind that would make the AI alignment problem less of a problem?
I'm not sure if the main complaint is about utilitarianism, or any other firm ethical system, as much as it's just that Yudkowsky is a huge weirdo. Most people, upon learning that this is a guy who likes doing stuff like "writing a short story about ethics with an offhand mention that, oh, hey, in this universe rape is legal and the characters think that people were prudes for not legalizing rape sooner, isn't that interesting? Just a thought experiment guys!" might not be the best candidate to put in charge of ensuring that AI's act in an ethical way and that, indeed, an AI not aligned by that guy would be preferable to AI's aligned by that guy, whether that's a fair charge or not.
To be fair, in that story it's quite clear that the society of the time has a very different view of what rape is, has no idea what, historically, the reality of rape was, and if they heard about it would probably die from vomiting up everything they ever consumed in their entire lives.
That's why he has the character of the Confessor who comes from the before times before everyone was cured of being mean and wicked and evil and nasty and who does know what 'rape' meant, and that's why he evokes (successfully) the reaction of the reader about "They legalised rape???? What kind of perverted monsters are they?" and then the other shoe drops about "Oh hang on, they don't mean rape rape, they mean some dumb form of flirting that implies consent to sex without explicitly verbalising consent". It's clever, I'll give him that, but yeah - it does leave him open to precisely that kind of "did you know?"
More options
Context Copy link
More options
Context Copy link
Yes, absolutely.
Edit to elaborate: A big part of Yudkowsky's problem is that he thinks that he can bypass the flaws of utilitarianism by having more information, by applying more intelligence/computing power, by being one iteration further along the recursive loop than everyone else. But the thing about recursive loops is that they are recursive, and as such being one iteration further along is the same as being as being one iteration behind.
I don't quite understand how we'd even begin to program a deontological or virtue ethicist AI. We're capable of giving things functions that they try and maximise, and we can call the subject of that function 'utility'. Whatever the flaws or virtues of utilitarianism, it does have the singular advantage of being computable. Compare to a virtue ethicist AI - how on earth do we begin building such a thing?
Even if it would be better, it seems like we're much closer to getting 'AI with a function it seeks to maximise' than we are getting 'AI who desires to fulfill virtues such as honour and charity'.
I agree that having an AI that believed in being virtuous according to human standards would be far, far better than one with a complicated mathematical function we try and map onto human utility and hope it doesn't kill us, but I've seen no reason to think the first is even possible.
Well, so far we're not capable of this. At best we build something that essentially modifies itself in response to rewards. It's not trying to maximize anything.
Given this, I don't think it's fair to describe current AIs as utilitarian. Their training reward functions were utilitarian, maybe, but it would be pretty easy to create reward functions that align more with virtue ethics.
I am absolutely keen to hear more about this, because everything I know tells me this is a close-to-impossible problem. The notion of 'pretty easy' seems intuitively wrong to me, but if you have any reading to offer on the subject I'd love to go through it.
Well, emphasis on more similar to virtue ethics. All it would take would be to change the reward criteria.
More options
Context Copy link
More options
Context Copy link
More options
Context Copy link
While defining terms is its own challenge, working in binary true/false and yes/no evaluations is arguably easier from a programming perspective than dealing weighted averages or trying to maximize a given value. Sure, a deontological AI will inevitably be vulnerable to Asimovian/Aes Sedai-esqe fallacies and exploits but a deontological AI is also not going to try and tile the universe in paper-clips or try to exterminate all life to in the name of preventing future suffering unless it's creator explicitly programs it to do so.
While Yudkowsky sees this as a fatal flaw, how can AGI it be described as intelligent if it doesn't "shut up and do the math". I see this as a feature. Utilitarianism is a stupid and evil ideology that is fundamentally incompatible with human flourishing. You can have a benevolent AI or you can have a utilitarian/consequentialist AI. You can't have both.
More options
Context Copy link
More options
Context Copy link
More options
Context Copy link
More options
Context Copy link