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 -
Steve Hsu has a startup that claims to have solved the hallucination problem. AFAIK they’ve not yet described their technique, but Hsu is no confabulist.
This doesn’t seem like the sort of thing that can be “solved”. Human brains hallucinate all the time after all, so at a minimum, “solving” the hallucination problem would mean attaining superhuman performance in the domain of general truth-telling. But even determining what counts as a hallucination or not is tricky.
Our current understanding of physics is that traveling faster than light is impossible. Every reputable source will confirm that FTL travel is impossible. But suppose for the sake of argument that, in reality, FTL travel actually is possible; it’s just that no one knows the truth about it. So if an LLM in 2023 in this reality tells you that FTL travel is impossible, is that a “hallucination”? It’s a false statement, after all.
Or suppose that an LLM tells you that Bruce Jenner is, in fact, a woman. Is it hallucinating? Every reputable source will tell you that he’s a woman. But determining whether he actually is a woman or not depends on a complex set of underlying philosophical assumptions.
I’m not sure what their proposal for solving hallucinations could be besides “check everything you say with Google and reputable sources”. But who gets to define what counts as a reputable source, and what if the reputable sources are wrong anyway?
Sure, but that's a different problem and not really hallucination. Hallucinate is making things up of whole cloth, such as making fake citations, etc...
Hallucination will be easy to solve and will be done in the next 2-3 years high confidence. A LLM just needs an API to access a store of truth. This store of truth will be a simple database of factual information, not a neural net. For example, the database will know that Abraham Lincoln was born on February 12, 1809. When the LLM controller gets data that contradicts the store of truth it will interrogate the trustworthyness of that data much like a human does, and if appropriate replace the data.
Just like a human, this would fail to function in the case of a coordinated attack on the integrity of data. If you're looking for an AI to tell you that one political side or the other is "correct", then that may indeed be an impossible task. But it's not related to hallucination.
More options
Context Copy link
More options
Context Copy link
Hsu is free to claim whatever he likes, but if we assume that OpenAi has been largely honest about how GPT works in their patent applications and research papers, it seems to me that it is unlikely that there will be any solution to the so-called "hallucination problem" outside of a complete ground-up redesign.
How so? I agree that a LLM neural net will hallucinate. But you can make a system that uses an LLM as a controller that can query trusted data sources and the problem is basically solved. The future of AI is not a pure LLM solution. It is using LLMs as part of a system.
More options
Context Copy link
More options
Context Copy link
More options
Context Copy link