Be advised: this thread is not for serious in-depth discussion of weighty topics (we have a link for that), this thread is not for anything Culture War related. This thread is for Fun. You got jokes? Share 'em. You got silly questions? Ask 'em.
- 71
- 3
What is this place?
This website is a place for people who want to move past shady thinking and test their ideas in a
court of people who don't all share the same biases. Our goal is to
optimize for light, not heat; this is a group effort, and all commentators are asked to do their part.
The weekly Culture War threads host the most
controversial topics and are the most visible aspect of The Motte. However, many other topics are
appropriate here. We encourage people to post anything related to science, politics, or philosophy;
if in doubt, post!
Check out The Vault for an archive of old quality posts.
You are encouraged to crosspost these elsewhere.
Why are you called The Motte?
A motte is a stone keep on a raised earthwork common in early medieval fortifications. More pertinently,
it's an element in a rhetorical move called a "Motte-and-Bailey",
originally identified by
philosopher Nicholas Shackel. It describes the tendency in discourse for people to move from a controversial
but high value claim to a defensible but less exciting one upon any resistance to the former. He likens
this to the medieval fortification, where a desirable land (the bailey) is abandoned when in danger for
the more easily defended motte. In Shackel's words, "The Motte represents the defensible but undesired
propositions to which one retreats when hard pressed."
On The Motte, always attempt to remain inside your defensible territory, even if you are not being pressed.
New post guidelines
If you're posting something that isn't related to the culture war, we encourage you to post a thread for it.
A submission statement is highly appreciated, but isn't necessary for text posts or links to largely-text posts
such as blogs or news articles; if we're unsure of the value of your post, we might remove it until you add a
submission statement. A submission statement is required for non-text sources (videos, podcasts, images).
Culture war posts go in the culture war thread; all links must either include a submission statement or
significant commentary. Bare links without those will be removed.
If in doubt, please post it!
Rules
- Courtesy
- Content
- Engagement
- When disagreeing with someone, state your objections explicitly.
- Proactively provide evidence in proportion to how partisan and inflammatory your claim might be.
- Accept temporary bans as a time-out, and don't attempt to rejoin the conversation until it's lifted.
- Don't attempt to build consensus or enforce ideological conformity.
- Write like everyone is reading and you want them to be included in the discussion.
- The Wildcard Rule
- The Metarule
Jump in the discussion.
No email address required.
Notes -
I would like a sanity check:
Is ChatGPT4 getting very noticably worse? My impression is that it has gotten much worse since it's release, in most visible areas such as engaging with questions without very elaborate prompting, answering them correctly (again without elaborate prompting), lying about what it knows and doesn't know and how to works, etc.
Questions I would get quick answers to before can barely be answered anymore (not CW stuff). Am I just imagining things or are they RLHFing it into uselessness?
Stanford says it has gotten worse.
https://fortune.com/2023/07/19/chatgpt-accuracy-stanford-study/
Thanks for the link!
More options
Context Copy link
More options
Context Copy link
I tend to find myself implementing my Wikipedia ruleset: trustworthy/useful on things that are not remotely political.
I currently use it for cooking recipes. Since most recipe websites are such garbage experiences that I'd rather trust the AI to slap together a recipe. I made a banana cake from an AI recipe.
I also still use chatgpt 3.5 haven't heard a compelling reason to use 4.
The chat also tends to have an annoying tendency to have a 3 paragraph answer minimum.
Does the AI get ingredient proportions right all the time?
Maybe?
Hasn't been horribly wrong yet. My banana cake did come out a bit closer to a banana bread in taste.
Verifying the instructions would require me to go use the horrible recipe sites I avoid.
More options
Context Copy link
More options
Context Copy link
More options
Context Copy link
FWIW, they claim it's the same model. Some people believe it may be difference in sampling strategy – speculative sampling, early exit decoding or some other tactic to cut costs.
More options
Context Copy link
How are you guys accessing ChatGPT4 anyways? I've used ChatGPT a couple of times and I was under the impression that GPT4 is only available via the commercial API with significant costs. Are you all paying for it, or is there some other site that lets you use it for free for some limited amount?
My employer has an account with unlimited gpt4 api access. I use the openai playground.
I recommend everyone use that short of making their own custom UI. It's way better than the ChatGPT web app interface.
More options
Context Copy link
You can sign up for ChatGPT Plus for $20/month. It comes with limited access to ChatGPT-4 (25 messages every three hours). That's a pretty fair price for having access to the most powerful AI in the world.
I'm not super against paying for it, I'm just a little reluctant to when I don't have any idea what to use it for that's actually useful. I've mostly used ChatGPT for things I might otherwise google for general explanations. It does mostly provide better results than whatever you'd find on Google.
More options
Context Copy link
Not if you can get api access
More options
Context Copy link
More options
Context Copy link
I've paid.
More options
Context Copy link
More options
Context Copy link
I don't know if ChatGPT 4 is getting worse, but I've been using the base GPT-4 API or thin wrappers over it from the beginning, and can't point to any deterioration myself.
More options
Context Copy link
https://www.businessinsider.com/openai-gpt4-ai-model-got-lazier-dumber-chatgpt-2023-7
There's been a lot of anecdotes about it lately. A thread in the orange site as well.
I see. Perhaps it'll get better in the future but the current version seems to have passed an inflection point in performance where it's not worth paying for or even using anymore.
More options
Context Copy link
More options
Context Copy link
Anecdotally I've seen it stop being able to transform markdown text into hyperlinks for some reason. Haven't used it much otherwise.
More options
Context Copy link
I've seen a people who use it saying so on twitter over the last months.
IIRC it was even addressed by OpenAI who said they improved performance because it was too costly.
More options
Context Copy link
I think our expectations have gotten so much better that its flaws are more visible. It’s hard to say how much RLHF (by customers, since release) is changing things.
Maybe so, but I've tried asking questions I've asked previously and it takes a lot of effort to make it answer, when it's at all possible, what it previously just answered.
More options
Context Copy link
More options
Context Copy link
More options
Context Copy link