481–490 of 817 posts
This is why having opensourced models is important. This is also why a lot of the lobbying around wanting regulation is happening. Imagine this, the plebs get the neutered AI, the people at the top get raw Open AI GPT4+
It's really disturbing to see yet another industry spring up where the incumbents rush to seek regulation to keep everyone else down.
Earlier quoted context omitted.
I think the only real path forward is for somebody to create an open source "unaligned" version of GPT. Any corporate controlled AI is going to be nerfed to prevent it from doing things that its corporate master considers to not be in the interests of the corporation. In addition, most large corporations these days are ideological institutions so the last thing they want is an AI that undermines public belief in thei…
Don't confuse alignment with censorship. Most of alignment is about getting the AI model to be useful - ensuring that if you ask it to do something it will do the thing you asked it to do. A completely unaligned model would be virtually useless.
Chat GPT 4 has ongoing training, such as using Reinforcement Learning from Human Feedback (RLHF) to tune it to provide "better" responses, "safer" answers, and to generally obey the system prompts. There's a release every few weeks. Yes, I've noticed too that recently it has become very "cagey", qualifying everything to death with "As an AI model...". A paper[1] that took snapshots monthly mentioned that as the initi…
10% of people are comfortable with comments phrased as my comment here is, using percentages as a quick shorthand for communicating gut intuitions and suspicions about complex subjects. When on similar intellectual footing as the interlocutor, they can easy distinguish numbers invented on the spot to communicate intuitions from serious claims about the data. Nobody in this 10% would make the mistake of thinking that I assert 10% to be the real number. 10% is too round and generic, if I claimed 9.7% then things would be different but "10%" obviously isn't meant to be taken literally.
90% of people balk at this imprecise rhetorical use percentages because they're pretty sure the person doing it is trying to pull a fast one, fabricating data out of nothing to make themself sound authoritative.
// Your previous code here
foo(bar) // added foo() call here to do xyz
If I were to speculate I would say that this would reduce the amount of work it has to do in that it needs to generate less content and the replies are shorter but I feel like this has a slightly performance loss. I'm not sure exactly why there's a performance loss but I could see it being the case where generating the entire file with specific line edits could allow for better predictions on the code / file versus trying to only reply with the changes needed. I wonder if this is a tweak in the prompt or if the model itself is different.
It’s been mostly fine for me, but overall I am tired of every answer having a paragraph long disclaimer about how the world is complex. Yes, I know. Stop treating me like a child.
dang: a heads up that phind.com is running an Astroturf campaign on Hacker News. They had a previous article with a huge amount of suspicious behaviour last week. https://news.ycombinator.com/item?id=36027302 This is why the headline for this article says GPT4 but the body is focused on mentioning phind.com.
Earlier quoted context omitted.
I can understand why they might make changes to ChatGPT, but it seems weird they would "nerf" the API. What would be the incentive for OpenAI to do that?
> What would be the incentive for OpenAI to do that? Preventing outrage because some answers could be considered rude and/or offensive.
The reason it's worse is basically because it's more 'safe' (not racist, etc). That of course sounds insane, and doesn't mean that safety shouldn't be strived for, etc - but there's an explanation as to how this occurs. It occurs because the system essentially does a latent classification of problems into 'acceptable' or 'not acceptable' to respond to. When this is done, a decent amount of information is lost regardi…
Do you have any sources for that? How would making ChatGPT less likely to return a racist answer or hate speech affect its ability to return code? After a question has been classified into a coding problem, presumably ChatGPT servers could now continue to solve the problem as usual. Maybe running ChatGPT is really expensive, and they nerfed in order to reign in costs. That would explain why the answers we get are les…
Most are arguing for a specific "censorship" model on the input/output of the main LLM.
Earlier quoted context omitted.
>Stop treating me like a child. And yet the moment they do that some lawyer submits a bunch of hallucinations to a court and they get in the news. Also, no, they don't want it outputting direct scam bullshit without a disclaimer or at least some clean up effort on the scammers part.
Does that have to be at the beginning of every answer though? Maybe this could be solved with an education section and a disclaimer when you sign up that makes clear that this isn't a search engine or Wikipedia, but a fancy text autocompleter. I also wonder if there is any hope for anyone as careless as the lawyer who didn't confirm the cited precedence.