3 ms·

The more safeguards, regulations, copyrights they add, the less useful ChatGPT becomes. It's a thing with many products. And I don't think it's bad that it's ha

The more safeguards, regulations, copyrights they add, the less useful ChatGPT becomes. It's a thing with many products. And I don't think it's bad that it's happening - safeguards exist for a reason. And current generations of LLM is very prone to lying (I don't use term hallucination, as IMO it's an attempt to whitewash limits of AI). ChatGPT did a great job at showing people current state of AI. But they did even greater job at overhyping itself.

WanderPanda 3y ago
"Lying" has a deliberate notion to it which imo is not accurate in this case
- justapassenger 3y ago
  It's deliberate. It's not deliberate by the tech itself. But they pack it in the product and market it, with a general narrative that it's superhuman like experience that is going to replace humans. At this level, it's deliberate.
  - ben_w 3y ago
    The marketing is what you're complaining about? Log in: > This is a free research preview. > Our goal is to get external feedback in order to improve our systems and make them safer. > While we have safeguards in place, the system may occasionally generate incorrect or misleading information and produce offensive or biased content. It is not intended to give advice. *next page* > Limitations > May occasionally generate incorrect information > May occasionally produce harmful instructions or biased content > Limited knowledge of world and events after 2021 That comes up every time you log in, even if you've seen it before.
    - justapassenger 3y ago
      What you describe is not marketing - it's legal-ish language. For marketing - listen to Sam Altman says. Listen to what VCs say. Listen to what influencers say.
      - ben_w 3y ago
        “People are begging to be disappointed and they will be. The hype is just like... We don’t have an actual AGI and that’s sort of what’s expected of us.” - Sam Altman in an interview about GPT-4 with StrictlyVC "(GPT4) is a system that will look back and say it was a very early AI. It is slow, it is buggy, and it does not do a lot of things very well, but neither did the very earliest computers, and they still pointed to a path that is going to be really important in our lives even if it took a few decades," - Sam Altman in an interview about GPT-4 with Lex Fridman The "influencers" may be saying any old rubbish; but that was always so, and the subjects of their fantasies should not be blamed for nonsense being spouted about them.
jedberg 3y ago
> I don't use term hallucination, as IMO it's an attempt to whitewash limits of AI Lying implies intent. I don't think the GPT models are purposely telling you misinformation. That's why we call them hallucinations. Because they model "thinks" it is telling the truth.
- JohnFen 3y ago
  "Hallucination" is a term that is just as problematic as "lying", for the same reason. Both terms imply a cognition that just isn't happening. The model doesn't "think" it's telling the truth. It doesn't "think" at all, at least in that sense. Nor does it "believe" or "disbelieve" anything. A better term than "lying" or "hallucination" would be "erroneous".

WanderPanda 3y ago

"Lying" has a deliberate notion to it which imo is not accurate in this case

justapassenger 3y ago
It's deliberate. It's not deliberate by the tech itself. But they pack it in the product and market it, with a general narrative that it's superhuman like experience that is going to replace humans. At this level, it's deliberate.
- ben_w 3y ago
  The marketing is what you're complaining about? Log in: > This is a free research preview. > Our goal is to get external feedback in order to improve our systems and make them safer. > While we have safeguards in place, the system may occasionally generate incorrect or misleading information and produce offensive or biased content. It is not intended to give advice. *next page* > Limitations > May occasionally generate incorrect information > May occasionally produce harmful instructions or biased content > Limited knowledge of world and events after 2021 That comes up every time you log in, even if you've seen it before.
  - justapassenger 3y ago
    What you describe is not marketing - it's legal-ish language. For marketing - listen to Sam Altman says. Listen to what VCs say. Listen to what influencers say.
    - ben_w 3y ago
      “People are begging to be disappointed and they will be. The hype is just like... We don’t have an actual AGI and that’s sort of what’s expected of us.” - Sam Altman in an interview about GPT-4 with StrictlyVC "(GPT4) is a system that will look back and say it was a very early AI. It is slow, it is buggy, and it does not do a lot of things very well, but neither did the very earliest computers, and they still pointed to a path that is going to be really important in our lives even if it took a few decades," - Sam Altman in an interview about GPT-4 with Lex Fridman The "influencers" may be saying any old rubbish; but that was always so, and the subjects of their fantasies should not be blamed for nonsense being spouted about them.

jedberg 3y ago

> I don't use term hallucination, as IMO it's an attempt to whitewash limits of AI Lying implies intent. I don't think the GPT models are purposely telling you misinformation. That's why we call them hallucinations. Because they model "thinks" it is telling the truth.

JohnFen 3y ago
"Hallucination" is a term that is just as problematic as "lying", for the same reason. Both terms imply a cognition that just isn't happening. The model doesn't "think" it's telling the truth. It doesn't "think" at all, at least in that sense. Nor does it "believe" or "disbelieve" anything. A better term than "lying" or "hallucination" would be "erroneous".