site banner

Culture War Roundup for the week of April 24, 2023

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.

11
Jump in the discussion.

No email address required.

It might be right 99% of the time, or even 99.99%, but like self driving cars this is exactly the kind of case where being good but not quite as good as a human is really dangerous.

Is getting things right 99.99%, or even 99% of the time not quite as good as a human doctor, rather than much better than a human doctor? I honestly don't know, and I'm not even sure how to quantify something like this to make like-for-like measurements. But surely at some % rate of success, the LLM would be getting things right at a rate better than could be expected of a typical (or even nth percentile) human doctor.

Which is scary in its own way. An LLM's "thought process" is currently completely unknowable, unlike that of a human doctor. So the types of mistakes it makes are likely to be more mysterious. Yet if it makes fewer mistakes and/or those mistakes are, on net, less harmful/unhealthy, then would it be our moral obligation to use those LLMs over doctors?

You know the one single artifact in the known universe that resembles an LLM the most?

The human brain. Another inscrutable black box for all practical purposes.

Sure, humans will give you a plausible-sounding answer when asked to explain their actions, but the current consensus as far as I'm aware is that those are almost entirely post-hoc responses, a nice-sounding package for the consumption of other humans than an accurate depiction of internal deliberations. Humans are constantly rationalizing, it takes a great deal of training to suppress those tendencies.

You know how we deal with this in practise? By simply asking the person to justify their thoughts. And it is entirely possible to ask the same thing of an LLM, the answers in both situations bear about as much relation to the truth of how they actually came to said conclusion!

then would it be our moral obligation to use those LLMs over doctors?

Only if you are a strict utilitarian which the vast majority of people are not.

Even then it might not in many cases, since there could be other benefits to having human doctors aside from the ability to provide diagnoses (e.g. human interaction would probably reduce anxiety about whatever illnesses and diagnoses and treatments are occurring).

I was thinking in % as accurate as a human, but didn't say that. I'll correct my original post.

I was also assuming it wouldn't be as good as a human, because of course a general model wouldn't accidentally be better than a specialist... Or would it?

I don't think it would be that hard to devise an experiment to get at least a rough idea of it's capabilities. Get some doctors, maybe professors, to devise questions of the sort OP was generating, present them to LLMs and real doctors (and maybe non-doctors with Google for an extra point of comparison), then have the professors grade the answers blind. I recall people giving LLMs math problems in this way, but I don't know if experiments have been performed with any rigor.

To your final point: what is the purpose of a doctor? Is it to heal at all costs, or is it to make people feel better? Many people go to doctors seeking specific prescription medication as a goal, when they could cure their ailment in a better but more laborious way. Some of them don't even have an ailment and only desire the effect of the meds. Many people also have procedures performed which physically make them less healthy, but fulfill their desires.

So, what should an LLM say to these people? "Don't take the pills, exercise" or "don't have surgery, improve your social life?"

I suppose it's not a new problem, but it does move the power and responsibility. Who is it moving to though?

If it interests you, GPT-4 successfully passed the USMLE, a pretty difficult exam that is intended to be a benchmark for the minimum level of competence a doctor needs to be able to practise. I'd say that counts for a lot.

I've thrown GPT 3.5 at a significant number of medical professionals, since I have several in my close and extended family. And even that antiquated model was sufficient to impress them, so I think the bar was already beaten before 4 showed up!

Just for the record, self made human is a real doctor, or a good enough liar to convincingly pass for one. His op is basically him starting your test and being blown away by the initial results.