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 -
Yes, yes, very smart people disagree with me. Your argument is to handwave me at the vast canon of AI scribbling? I've read the big ones, and they are as unconvincing as they are hysterical. It's a very specific style, one I recognize well from my upbringing in a millennial faith-healing cult. It all sounds very convincing, if you haven't been down this road before.
What I mean is that ... questions like "we've solved lots of problems before, AI will be fine" and "there will be a disaster, we will notice it, and then politicans will solve it" are things that people have written dozens of essays debating. It's like talking to someone here about race and genetics here and just saying "races aren't real. it's a distribution, not a category. and stereotyping is bigoted". Everyone here has heard that before, and hundreds of people have written up hundreds of paragraphs about why it's false. Maybe it's still true in some way, but that point is best made by addressing those arguments in some way, not just saying them.
Yeah, how so precisely? Again, it'd be much more interesting to read about why yud's arguments are wrong than "hurr its a cult you are being manipulated accept my social pressure instead of theirs"
The big danger is not AI, it's people who want to use AI to make a shit-ton of profit. The flip-side of the Fairy Godmother AI that will be so smart it will solve all the intractable problems that unaided humanity could never solve, and provide Fully Automated Luxury Gay Space Communism for all, is of course the Paperclip AI that will do away with us all (and that only if we're lucky, otherwise it's I Have No Mouth And I Must Scream for all).
(Side note here: congratulations, rationalists, you have managed to re-invent God and the Devil, Heaven and Hell, all over again even while wondering how anybody can possibly believe in religion's view of the afterlife).
So people who desperately want Fairy Godmother AI have to grapple with the possibility of Paperclip AI.
I don't accept the problem in those terms. I don't think we're ever going to get superhumanly smart AI that can be its own agent with its own goals any time near, if at all.
What I do think we will get is 'good enough' AI that private enterprises and governments will try exploiting just for that tiny edge. If trading fortunes can be made and lost on microsecond decisions, why not use your patented money-tree AI to make nanosecond decisions? Why not use AI for the social welfare and health care problems of triage that currently are being tried out by insurance companies about "ring up our hotline and some half-trained person running off a script will decide if you qualify to go see a doctor"?
We'll hand over decision making powers to dumb machines in order to make money, and we'll fuck ourselves up in the process. That's the risk, not a god-level intelligence AI deciding it wants to get rid of its monkey masters and tile the universe with NFTs.
Now, if you can solve the problem of "humans: we're still greedy, dumb monkeys fighting each other over who gets the bananas", then we won't have the problem of "uh-oh, we fucked up how our civilisation works". That's why I think "we must solve the problem of getting AI aligned with our morals/values!" is the wrong track to take; humans won't even align with our own morals/values, and besides, "fuck over that guy so I can get more bananas" is completely compatible with how we act morally/express our values, so if the AI aligns with that, why be surprised what results?
Yes, and handing over all the levers of society to very intelligent machines is bad. Why will the machines be dumb forever? Even if they are dumb at first, they'll become smarter, and quickly, because - see DL progress and theory of computation.
Yeah, this is the main issue! Why? Even ignoring object-level arguments - look how rapidly technology has advanced over the past 200 years, does that just ... stop?
Yeah. I think there are hard limits in the physical universe. I have no freakin' idea how human consciousness arose, and there are plenty of people willing to argue that no such a thing exists. So "we will get AI being conscious because magic happens" is not sufficiently convincing to me.
... but there's clearly a lower-bound of humans, and the wide variation in human intelligence incredibly strongly suggests that we're not right up against a cap of any kind. And even without that, you haven't actually said anything beyond "physical limits exist". It's 1800 - "we will never make anything faster than horses. Physical Limits Exist!"
nobody said 'conscious', i have no idea what you mean by conscious, it's a term that, in this context, has 500 different meanings, most of which are only loosely related. just 'intelligent and capable'
More options
Context Copy link
More options
Context Copy link
100 years ago, people predicted we'd all have jetpacks and flying cars, and then... progress just stopped. Technology got better and better, sure, but the energy needed to run a jetpack still has to come from somewhere.
The details would be a much longer post, but my take on AI is pretty much the same. Chips will get faster and faster, yes, models more complex, training algorithms smarter, but the data you use still has to come from somewhere.
Dumb-but-good-enough AI is going to be huge, obviously, and could have wide-reaching effects on society. I'm unconvinced you can turn that into Skynet just by applying Moore's law.
we literally do have jetpacks and flying cars lol. They're just not that useful in general use-cases relative to existing technology.
how is AI in a different situation from humans here, humans have 'enough data' as it is
Exactly my point.
Humans have millions of years of evolution on their side, and require decades of real-world interaction to train to human-level competence.
If you're training GPT-3 or DALL-E or AlphaGo, you can do the training offline. Here's an internet's worth of text and image; here's a perfect simulation of the rules of Go. So sure, with enough GPUs you can train far, far faster than any human.
Taking over the world does not, to me, seem amenable to offline training. Maybe I'm wrong? But if you want to convince me, this is the sticking point. What data, exactly, would an AGI be trained against?
if we accept all of this, it'll take ~ 200 years for superhuman AGI to happen though? That still has the same outcome!
human lives and writing? There are billions of them, and if it literally required a camera on a hundred million AR glasses, which it won't, that could easily happen.
More options
Context Copy link
More options
Context Copy link
More options
Context Copy link
More options
Context Copy link
More options
Context Copy link
More options
Context Copy link
More options
Context Copy link
More options
Context Copy link