291–300 of 323 posts
Perhaps it's expensive to self-censor the output, so they don't want to pay to self-censor every intrusive thought in the chain, so they just do it once at output.
They don't want to have to deal with the model's "thought crimes"
Earlier quoted context omitted.
The model will often recognise a request is part of whatever ${naughty_list} it was trained on and generate a refusal response. Banning seems more aimed at preventing working around this by throwing massive volume at it to see what eventually slips through, as requiring a new payment account integration puts a "significantly better than doing nothing" hamper on that type of exploiting. I.e. their goal isn't to have a…
I think once a small corpus of examples of CoT gets around, people will be able to reverse-engineer it.
The other part of the massive volume issue is it's not just "what clever prompts can skirt around detection sometimes" it's "detection, like the rest of it, doesn't seem to work for 100% of outputs so throwing the same 'please do it anyways' in enough times can get you by if you're dedicated enough" type problem.
Earlier quoted context omitted.
i don't write understand how that would lead to anything but a slightly different response. How can token prediction have this capability without explicitly enabling some heretofore unenabled mechanism? People have been asking this for years.
English is turing complete.
Earlier quoted context omitted.
There's no change in price. They charge the same amount per token from everyone. You pay more if you use more tokens. If some tokens are hidden, used internally to generate the final 'public' tokens is just a matter of technical implementation and business choice. If you're not happy, don't use the service.
Well imagine how it looks from the point of view of anti-discrimination and consumer protection laws: we charge this person an additional $100 because we have some imaginary units telling us they owe us $100... Just trust us. Not sure it will hold in court. If the both sides agree to a specific sum beforehand, no problem. But you can't just charge random amounts post factum without the person having any idea why they…
Is it actually different from paying a contractor to do some work for you on an hourly basis, and them then having to "think more" and thus spend more hours on problem A than probably B?
https://xcancel.com/SmokeAwayyy/status/1834641370486915417
The worst responses are links to something the generalized you can't be bothered to summarize. Providing a link is fine, but don't expect us to do the work to figure out what you are trying to say via your link.
The name "OpenAI" is a contraction since they don't seem "open" in any way. The only way I see "open" applying is "open for business."
They have several open models, including Whisper.
Earlier quoted context omitted.
Far, far more than 1% of people care. Sure, they are open in one sense: for business. But in the tech world, "open" specifically means showing us how you got your final product. It means releasing source code rather than just binaries (even free binaries!), or sharing protocols and standards rather than keeping them proprietary (looking at you Apple and HDMI). It doesn't matter if anyone can use ChatGPT, that has not…
Not enough people care to make considering a name change worthwhile. The net benefit of changing their name is negative. If I were Sam Altman, I would keep the name, changing it would hurt the company.
Earlier quoted context omitted.
Just no. AI being racist is still a popular meme. "Because the programmers are white males blah blah".
Why can't it be, if it were (I'm not saying that it is, mind) trained on racist material?
Earlier quoted context omitted.
Occam's razor: there is no secret sauce and they're afraid someone trains a model on the output like what happened soon after the release of GPT-4. They basically said as much in the official announcement, you hardly even have to read between the lines.
Stop using Occam's razor like some literal law. It's a stupid and lazy philosophical theory bandied about like some catch-all solution. Like when people say 'the definition of insanity is[some random BS] with a bullshit attribution[Albert Einstein said it!(He didn't)]