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 -
Was a bit surprised to see this hadn't been posted yet, but yesterday Yudkowsky wrote an op-ed in TIME magazine where he describes the kind of regime that he believes would be necessary to throttle AI progress:
https://archive.is/A1u57
Some choice excerpts:
if its presence in the CW thread needs justifying, well, it's published in a major magazine and the kinds of policy proposals set forth would certainly ignite heated political debate were they ever to be seriously considered.
"Yudkowsky airstrike threshold" has already become a minor meme on rat and AI twitter.
I see we're back to trying to outlaw mathematics. I encourage everyone to read this article by Stephen Wolfram describing how LLMs work before panicking. I cannot understand the degree to which LLMs have apparently broken some people's brains.
Any concept which can be expressed in words is now, in principle, understandable by a computer. I am confused by your confusion.
Maybe you and I have different definitions of "understand" but this seems clearly false to me.
ETA:
Perhaps better to say, I don't see what this statement has to do with LLMs.
Let me break it down:
LLMs understand words and sentences (or at least the relationships between them).
If a concept can be expressed in words, then what that concept means is encoded by how the words that express that concept relate to the words that express any other concept.
Therefore, as an LLMs capability to relate words and sentences to each other approaches or exceeds human level, the LLMs understanding of concepts approaches or exceeds human level.
This is a very high-level explanation, but I think it captures the core intuition.
I must ask, what is the most powerful LLM you have personally interacted with? Default ChatGPT feels like talking to a 6-year-old after using Bing and GPT-4.
I guess (1) is where I disagree. I think there's a substantial difference between understanding the meaning of a word and understanding that word's statistical relation to other words. LLM's understand the latter, but not the former.
I've used Bing and it's still pretty easy to ask it questions that demonstrate it doesn't understand the semantic content of what it is being asked. Here's an interaction I had just now:
I asked it the same question three times and never once did it give me a story that actually had fifteen words in it. It clearly does not understand what "fifteen" means and how the presence of the word in the my sentence should constrain its answer.
Hmmm, you're right. GPT-4 does in fact have trouble writing 15-word stories.
On the other hand, I was able to get it to write a 68 word story on the first try.
Trying this experiment with multiple 2-digit numbers X, it looks like GPT-4 outputs a story with X ± 2 words. I have updated slightly against GPT-4's capabilities (I expected it to get the exact answer most of the time), but I still hold to my original thesis. This is a quantitative error, not a qualitative error. The presence of the number in the input sentence does in fact constrain the output. Asking for a story with 68 words gives an output with more words than asking for a story with 49 words, but less words than asking for a story with 96 words. The model does have some concept of what these numbers are.
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