3 ms·

As far as I know GPT has no motivation. It responds because it is programmed to do so, it doesn't do it because it wants to. The dopamine system is an internal

by corobo 4y ago

As far as I know GPT has no motivation. It responds because it is programmed to do so, it doesn't do it because it wants to. The dopamine system is an internal reward process. Does GPT give itself internal props ('consciously' or not) in any way for doing things it enjoyed doing or that proved to be useful? Legitimate question, I do not know. I'm not getting into the weeds on whether a while loop is sentient. That's an internal prompt sure but it's not motivated by anything, it's programmed.

fooker 4y ago
> internal props in any way for doing things it enjoyed doing or that proved to be useful. Yes. > whether a while loop is sentient. My point with this exercise was not to prove a while loop to be sentient. The idea is that any definition of sentience you can come up with can be artificially defeated. I’m not claiming that ChatGPT is sentient, merely that you can not prove the opposite with our current understanding.
- corobo 4y ago
  Could you expand on the yes? Or just chuck me some keywords I can look up. What is it's equivalent to dopamine called? I'm truly trying to know more. All of my opinions so far are direct evidence and feeling based rather than knowing the tech. If I'm wrong I'm wrong. [Edit: I just asked GPT itself. It reckons it does not have an equivalent to dopamine, closest it's got are algorithms designed to learn and improve based on feedback. I'm guessing that means the thumbs up and down and maybe some dev mode tools, external feedback mechanisms in any case] If we can't prove that something is sentient we'll just have to assume everything is sentient unless we can prove it isn't. In this case ChatGPT cannot do anything without being prompted (via function in a while loop or a chat message, lol). It is strictly a program and therefore is not sentient. Happy to reevaluate this basic hurdle when ChatGPT is able to send a useful/relevant/question/philosophical (any reasoning other than random) message first.

by corobo 4y ago

fooker 4y ago

> internal props in any way for doing things it enjoyed doing or that proved to be useful. Yes. > whether a while loop is sentient. My point with this exercise was not to prove a while loop to be sentient. The idea is that any definition of sentience you can come up with can be artificially defeated. I’m not claiming that ChatGPT is sentient, merely that you can not prove the opposite with our current understanding.

corobo 4y ago
Could you expand on the yes? Or just chuck me some keywords I can look up. What is it's equivalent to dopamine called? I'm truly trying to know more. All of my opinions so far are direct evidence and feeling based rather than knowing the tech. If I'm wrong I'm wrong. [Edit: I just asked GPT itself. It reckons it does not have an equivalent to dopamine, closest it's got are algorithms designed to learn and improve based on feedback. I'm guessing that means the thumbs up and down and maybe some dev mode tools, external feedback mechanisms in any case] If we can't prove that something is sentient we'll just have to assume everything is sentient unless we can prove it isn't. In this case ChatGPT cannot do anything without being prompted (via function in a while loop or a chat message, lol). It is strictly a program and therefore is not sentient. Happy to reevaluate this basic hurdle when ChatGPT is able to send a useful/relevant/question/philosophical (any reasoning other than random) message first.