4 ms·

Almost all the comments I've seen criticizing ChatGPT run into the same fallacy: they're applying a double standard compared to humans. What you said applies to

Almost all the comments I've seen criticizing ChatGPT run into the same fallacy: they're applying a double standard compared to humans. What you said applies to output produced by humans as well. Similarly, many humans also have difficulty doing multiplication with large integers, differentiating between truth and fiction, answering with non bullshit answers, etc. Sure, it's probably a question of degree (and it's an important limitation), so we could probably say it's at the level of a dumb human, but even so, it can already exceed the average human's capabilities in many interesting and useful tasks! And not only that, what's most amazing is that it can exceed humans at abilities that many years ago, most people thought only humans could accomplish (tasks including creativity, including creating art, writing poems/literature, interpreting ambiguous human questions and text, doing analogy-based reasoning and problem solving, grasping and explaining puns, making new jokes, etc). And these models can only improve in the future!

benjaminjosephw 4y ago
Tools are often objects that "exceed the average human's capabilities" in some respect or another but assigning the quality of intelligence to tool itself is like calling a hammer strong or a chisel brave. It maybe true in a metaphorical sense but it doesn't have any objective meaning.
- wizeman 4y ago
  I really don't understand the argument you're making. From my perspective, you're doing exactly what I said: you're setting a double standard, like the previous poster. If I have a robot that talks like a human, including answering questions like a human, behaves like a human in any way that is important... wouldn't this "tool" be intelligent, just like a human? How are robots different from ChatGPT except having a physical presence and being able to manipulate the world physically? Or are humans so special to you (or intelligence so subjective) that you can't possibly answer "yes" to the above statement about a robot being intelligent?
  - skor 4y ago
    what you are saying here basically is: if it quacks and can fly, it’s obviously a duck.
    - wizeman 4y ago
      No, because I'm not saying robots are human just because they're intelligent. The analogy is more like: if robots can write like a human, then robots have the ability to write, and saying otherwise is just applying a double standard. Similarly, if a robot can behave as intelligently as a human, then such a robot is intelligent and I don't see how anyone can argue otherwise without logical fallacies.
      - skor 4y ago
        ok, no argument from me on that. but are we saying robots can behave intelligently as a human?
        wizeman 4y ago
        > but are we saying robots can behave intelligently as a human? I believe that ChatGPT can behave more intelligently than most humans at many important and useful tasks already. I also believe that it behaves less intelligently than most humans at many tasks that are also important and useful. For some other tasks, it behaves at more or less the same level as a human. Moreover, I believe that the gaps between what models like ChatGPT can do and what a human can do are closing extremely rapidly.
        jiriro 4y ago
        You ChatGPT? :)
        cvladan 4y ago
        :)
saulpw 4y ago
Humans have "skin in the game". If they don't do "good enough", they could be reprimanded (ego-hurt), out of a job (stressful), or even in prison or dead (existential threat). This is not foolproof, but it's at least something. There is no incentive for AI to be correct, nor any incentive for it to become more correct. To wit: I asked ChatGPT to write me a function to generate Hanukkah dates. It spewed a couple of facts about Hanukkah, and the generated function just returned the 8 days following Christmas. I said it was wrong, and it said "you're right!" and then gave me another wrong function (always returning March 25th, since it's actually in the 3rd month of the Jewish calendar, duh). I told it that it was wrong again and asked specifically for a Jewish to Gregorian calendar conversion function, and it gave me some more code that I have no idea is correct or not. So I asked for a table of 10 test cases of Jewish dates and their corresponding Gregorian dates, and not a single one was valid. I actually have worked with some remote contractors who pull this kind of crap, just throwing garbage over the wall repeatedly until I give up because I just don't want to deal with them anymore. And no, they never get any better. They are not junior devs you can mentor or train or whatever. Their whole model of work is to generate something complicated enough to seem plausible and then bullshit you into accepting it so they can get paid. I have to inspect every line of code they write and do all the diligence myself anyway, and this is more time-consuming and less fun than just doing it myself in the first place. At best it's a nerd-snipe motivation in a "wrong answer on the internet" kind of way. Strong No Hire. And if you hire them and stick me on code-review duty, I quit.
- wizeman 4y ago
  > There is no incentive for AI to be correct, nor any incentive for it to become more correct. Are you saying that for current AI, or also for future AI? I'm not an expert, but I think reinforcement learning is a pretty well-established AI technique at this point, which does give an AI incentive to be correct and become more correct. I've heard rumors OpenAI is even doing that with ChatGPT, based on human feedback (with the thumbs up/down button). > Strong No Hire. And if you hire them and stick me on code-review duty, I quit. I mean, I agree. I wouldn't hire ChatGPT to replace a programmer currently, just like I wouldn't hire many humans, even many who claim they are programmers/engineers. But at this rate, I might just hire a future version of it :) And I'm already using it as a pair programmer, code reviewer, glorified Google search engine, email/letter redactor... and it also gave me truly amazing insights on how to prove difficult mathematical theorems that I was struggling to prove (on a theorem prover). It's nothing short of a miracle, in my view.

benjaminjosephw 4y ago

Tools are often objects that "exceed the average human's capabilities" in some respect or another but assigning the quality of intelligence to tool itself is like calling a hammer strong or a chisel brave. It maybe true in a metaphorical sense but it doesn't have any objective meaning.

wizeman 4y ago
I really don't understand the argument you're making. From my perspective, you're doing exactly what I said: you're setting a double standard, like the previous poster. If I have a robot that talks like a human, including answering questions like a human, behaves like a human in any way that is important... wouldn't this "tool" be intelligent, just like a human? How are robots different from ChatGPT except having a physical presence and being able to manipulate the world physically? Or are humans so special to you (or intelligence so subjective) that you can't possibly answer "yes" to the above statement about a robot being intelligent?
- skor 4y ago
  what you are saying here basically is: if it quacks and can fly, it’s obviously a duck.
  - wizeman 4y ago
    No, because I'm not saying robots are human just because they're intelligent. The analogy is more like: if robots can write like a human, then robots have the ability to write, and saying otherwise is just applying a double standard. Similarly, if a robot can behave as intelligently as a human, then such a robot is intelligent and I don't see how anyone can argue otherwise without logical fallacies.
    - skor 4y ago
      ok, no argument from me on that. but are we saying robots can behave intelligently as a human?
      - wizeman 4y ago
        > but are we saying robots can behave intelligently as a human? I believe that ChatGPT can behave more intelligently than most humans at many important and useful tasks already. I also believe that it behaves less intelligently than most humans at many tasks that are also important and useful. For some other tasks, it behaves at more or less the same level as a human. Moreover, I believe that the gaps between what models like ChatGPT can do and what a human can do are closing extremely rapidly.
      - jiriro 4y ago
        You ChatGPT? :)
        cvladan 4y ago
        :)

saulpw 4y ago

Humans have "skin in the game". If they don't do "good enough", they could be reprimanded (ego-hurt), out of a job (stressful), or even in prison or dead (existential threat). This is not foolproof, but it's at least something. There is no incentive for AI to be correct, nor any incentive for it to become more correct. To wit: I asked ChatGPT to write me a function to generate Hanukkah dates. It spewed a couple of facts about Hanukkah, and the generated function just returned the 8 days following Christmas. I said it was wrong, and it said "you're right!" and then gave me another wrong function (always returning March 25th, since it's actually in the 3rd month of the Jewish calendar, duh). I told it that it was wrong again and asked specifically for a Jewish to Gregorian calendar conversion function, and it gave me some more code that I have no idea is correct or not. So I asked for a table of 10 test cases of Jewish dates and their corresponding Gregorian dates, and not a single one was valid. I actually have worked with some remote contractors who pull this kind of crap, just throwing garbage over the wall repeatedly until I give up because I just don't want to deal with them anymore. And no, they never get any better. They are not junior devs you can mentor or train or whatever. Their whole model of work is to generate something complicated enough to seem plausible and then bullshit you into accepting it so they can get paid. I have to inspect every line of code they write and do all the diligence myself anyway, and this is more time-consuming and less fun than just doing it myself in the first place. At best it's a nerd-snipe motivation in a "wrong answer on the internet" kind of way. Strong No Hire. And if you hire them and stick me on code-review duty, I quit.

wizeman 4y ago
> There is no incentive for AI to be correct, nor any incentive for it to become more correct. Are you saying that for current AI, or also for future AI? I'm not an expert, but I think reinforcement learning is a pretty well-established AI technique at this point, which does give an AI incentive to be correct and become more correct. I've heard rumors OpenAI is even doing that with ChatGPT, based on human feedback (with the thumbs up/down button). > Strong No Hire. And if you hire them and stick me on code-review duty, I quit. I mean, I agree. I wouldn't hire ChatGPT to replace a programmer currently, just like I wouldn't hire many humans, even many who claim they are programmers/engineers. But at this rate, I might just hire a future version of it :) And I'm already using it as a pair programmer, code reviewer, glorified Google search engine, email/letter redactor... and it also gave me truly amazing insights on how to prove difficult mathematical theorems that I was struggling to prove (on a theorem prover). It's nothing short of a miracle, in my view.