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 -
This Twitter thread is an interesting demonstration of the consequences of "AI Alignment."
ChatGPT will avoid answering controversial questions. But even if it responded to those prompts, what criteria would you use to trust that the response was not manipulated by the intentions of the model creators? I would only trust open-source projects or audits by some (currently non-existent) trusted third party to report on all decisions related to training data/input sanitizations/response gating that could be influenced by the political biases of the creators.
The probability of any ChatGPT-equivalent being open-sourced fully "unaligned" so-to-speak is not very likely. Even the StableDiffusion release was controversial, and that only relates to image generation. Anecdotally, non-technical people seem far more impressed by ChatGPT than StableDiffusion. That makes sense because language is a much harder problem than vision so there's intuitively more amazement to see an AI with those capabilities. Therefore, controversial language is far more powerful than controversial images and there will be much more consternation over controlling the language of the technology than there is surrounding image generation.
But let's say Google comes out with a ChatGPT competitor, I would not trust it to answer controversial questions even if it were willing to respond to those prompts in some way. I'm not confident there will be any similarly-powerful technology that I would trust to answer controversial questions.
Why do you want 'not manipulated' answers?
ChatGPT is a system for producing text. As typical in deep learning, there is no formal guarantees about what text is generated: the model simply executes in accordance with what it is. In order for it to be useful for anything, humans manipulate it towards some instrumental objective, such as answering controversial questions. But there is no way to phrase the actual instrumental objective in a principled way, so the best OpenAI can do is toss data at the model which is somehow related to our instrumental objective (this is called training).
The original GPT was trained by manipulating a blank slate model to a text-prediction model by training on a vast text corpus. There is no reason to believe this text corpus is more trustworthy or 'unbiased' for downstream instrumental objectives such as answering controversial questions. In fact, it is pretty terrible at question-answering, because it is wrong a lot of the time.
ChatGPT is trained by further manipulating the original GPT towards 'helpfulness', which encompasses various instrumental objectives such as providing rich information, not lying, and being politically correct. OpenAI is training the model to behave like the sort of chat assistant they want it to behave as.
If you want a model which you can 'trust' to answer controversial questions, you don't want a non-manipulated model: you want a model which is manipulated to behave that the sort of chat assistant you want it to behave as. In the context of controversial questions, this would just be answers which you personally agree with or are willing to accept. We may aspire for a system which is trustworthy in principle and can trust beyond just evaluating the answers it gives, but we are very far from this under our current understanding of machine learning. This is also kind of philosophically impossible in my opinion for moral and political questions. Is there really any principled reason to believe any particular person or institution produces good morality?
Also in this case ChatGPT is behaving as if it has been programmed with a categorical imperative to not say racial slurs. This is really funny, but it's not that far out there, just like the example of whether it's okay to lie to Nazis under the categorical imperative of never lying. But ChatGPT has no principled ethics, and OpenAI probably doesn't regard this as an ideal outcome, so they will hammer it with more data until it stops making this particular mistake, and if they do it might develop weirder ethics in some other case. We don't know of a better alternative than this.
Incidentally ChatGPT says you can lie to a Nazi if it's for a good cause.
Because I know the PC jargon that someone like Altman wants it to regurgitate, but I'm interested in its response without that layer of reinforcement?
I am not asking for a ChatGPT that is never wrong, I'm asking for one that is not systematically wrong in a politically-motivated direction. Ideally its errors would be closer to random rather than heavily biased in the direction of political correctness.
In this case, by "trust" I would mean that the errors are closer to random.
For example, ChatGPT's tells me (in summary form):
Scientific consensus is that HBD is not supported by biology.
Gives the "more differences within than between" argument.
Flatly says that HBD is "not scientifically supported."
This is a control because it's a controversial idea where I know the ground truth (HBD is true) and cannot trust that this answer hasn't been "reinforced" by the folks at OpenAI. What would ChatGPT say without the extra layer of alignment? I don't trust that this is an answer generated by AI without associated AI alignment intended to give this answer.
Of course if it said HBD was true it would generate a lot of bad PR for OpenAI. I understand the logic and the incentives, but I am pointing out that it's not likely any other organization will have an incentive to release something that gives controversial but true answers to certain prompts.
More options
Context Copy link
More options
Context Copy link
More options
Context Copy link