Did ChatGPT just form a thought? • Hacker News Board

Re: Did ChatGPT just form a thought?

#11

Post by MarcScott » Tue, Jan 10, 2023, 9:21 PM UTC

sysrpl wrote: post #10

MarcScott wrote: post #9

I tried this prompt >I want you to play a game of 20 questions with me. You will ask me questions, one at a time, to try and determine which animal I am thinking of. You will use my answers to inform your next question. You can guess the animal I am thinking of at any time, but you must make a guess after your twentieth question. It appeared to break at question 7, but it might be experiencing high usage at the momen…

That is common at the moment. Their servers are overloaded due to the wild popularity of ChatGpt and often the session will just "break", erasing any context or learning it had acquired in your session. You must then start over.

I hit a question limit on the last attempt, which I've never seen before.

Re: Did ChatGPT just form a thought?

#12

Post by gibsonf1 » Tue, Jan 10, 2023, 9:32 PM UTC

Indeed, ChatGPT did not form a thought as it can't think or understand what a single word means. This technology will never produce intelligence for which base requirements include conceptual understanding and awareness by modeling space-time relationships.

Re: Did ChatGPT just form a thought?

#13

Post by dgeiser13 » Wed, Jan 11, 2023, 1:20 AM UTC

One more ChatGPT post.

Re: Did ChatGPT just form a thought?

#14

Post by pwinnski » Wed, Jan 11, 2023, 1:59 AM UTC

No, it did not.

Humans are pattern-matchers, indeed, pattern-overmatchers, so it's not unreasonable that you felt that way, but this does not reflect anything like a "thought" from the AI model.

Re: Did ChatGPT just form a thought?

#15

Post by bananapub » Wed, Jan 11, 2023, 11:58 AM UTC

no

Re: Did ChatGPT just form a thought?

#16

Post by _nalply » Wed, Jan 11, 2023, 3:08 PM UTC

pwinnski wrote: post #14

No, it did not. Humans are pattern-matchers, indeed, pattern-overmatchers, so it's not unreasonable that you felt that way, but this does not reflect anything like a "thought" from the AI model.

Yeah, it is hard to overcome our antropomorphic bias. We all know Disney animal toons who behave like humans and we don't have a problem identifying ourselves with the animal hero. So we will bias to maintain something having "thoughts".

However this is not enough to simply dismiss ChatGPT having "thoughts".

We need to understand first what it means to have "thoughts".

It is my opinion that ChatGPT does not have conciousness. For that it seems to lack self understanding which does not come from simply parroting some utterances found in the data used for the machine learning.

But could it have some inkling of a "thought" anyway? I am not sure but I don't think so. Yet.