Asked ChatGPT when I would get my CT

[drrjv]

Pretty reasoned answer — see attachment

Sponsored

 

[Nabilriaz69]

I do not see any intelligence in this answer all pretty general items , or common sense

 

[Jhodgesatmb]

I do not see any intelligence in this answer all pretty general items , or common sense

You don’t think that it takes intelligence to use common knowledge or common sense? At the end of the day ChatGPT can only aggregate information provided to it (or that it can find).

 

[cvalue13]

You don’t think that it takes intelligence to use common knowledge or common sense? At the end of the day ChatGPT can only aggregate information provided to it (or that it can find).

In a sandbox test, CGPT was assigned a task by OpenAI, during the course of which CGPT encountered a CAPTCHA block.

CGPT went to Task Rabbit, hired a human, told the human it was a blind person, and a few minutes and dollars later CGPT was past the CAPTCHA block and carrying on with its original task.

did it merely aggregate the ability to deceive?

or instead, is merely aggregating about all humans do as well?

 

[Jhodgesatmb]

In a sandbox test, CGPT was assigned a task by OpenAI, during the course of which CGPT encountered a CAPTCHA block.

CGPT went to Task Rabbit, hired a human, told the human it was a blind person, and a few minutes and dollars later CGPT was past the CAPTCHA block and carrying on with its original task.

did it merely aggregate the ability to deceive?

or instead, is merely aggregating about all humans do as well?

Aggregating yes, deception no.

You are interpreting its behavior as deception and intentional. It knows about planning and goals and how to achieve them. All knowledge that has been around for decades (I can provide references). I have seen many examples of ChatGPT being asked to perform planning tasks and the solution you mention has probably been described on the web many times. All ChatGPT needed to do is find a solution to that problem and execute it. If it were a human I would agree with you that it is deception but not a robot.

Now, if ChatGPT has been given knowledge of deception as a way to get around a subgoal failure (also well known) then maybe you are right, but then we are really getting into knowledge based reasoning and I genuinely don’t know whether LLMs go that way. Scary if they do.

Back to the OPs comment that ChatGPT wasn’t exhibiting intelligence, I said that using common sense and generally available knowledge is intelligent behavior, and your example agrees with me and just took it one step further (to planning and problem solving). As I see it we only differ in the notion of robot intentionality.

 

[cvalue13]

Aggregating yes, deception no.

You are interpreting its behavior as deception and intentional. As I see it we only differ in the notion of robot intentionality.

Maybe we only disagree about human intentionality

or instead, is merely aggregating about all humans do as well?

because we tend to overvalue what other minds doing, we tend to undervalue what other beings’ minds are doing

 

[Jhodgesatmb]

Maybe we only disagree about human intentionality



because we tend to overvalue what other minds doing, we tend to undervalue what other beings’ minds are doing

That is a very interesting point. I suppose that you have to talk to a cognitive psychologist about current theories/understanding of human memory, learning, and behavior. I had a minor in cognitive psychology in school but that was a long time ago. Cognitive scientists used to believe (at least some) that all of what humans do is deterministic (predictable given enough knowledge of the system), and that argument would say that, yes, humans are aggregating machines. But we are also experimentation and generalization machines. We try things out, and we generalize the results. In fact, we generalize on sample sizes of '1', which leads to interesting behavior. Anyway, I am moving away from the original notion that ChatGPT isn't acting intelligently when it offers nothing more than aggregation. I find ChatGPT very helpful as a search tool. It would have taken me some time to come up with that enumeration of things to consider vis-a-vis our reservations, but I think that that I, you, and most everyone here could have done it. What I do not recall seeing in the response was an answer to the question: "when will I get my cybertruck". A diplomatic answer to be sure. Neither positive nor negative but not substantive.

 

[cvalue13]

I suppose that you have to talk to a cognitive psychologist about current theories/understanding of human memory, learning, and behavior.

nah

On these topics, those cog-psy cats are mostly just so-so a posteriori empiricists catching up to 200yo a priori rationalism.

(Like modern physicis just catching up to much of 200yo+ ~metaphysics.)


While, it’s important to set one apple next to another one to ‘know’ that 1+1=2 in the observable world, prior to that empiricism it was deductible/knowable.



Regardless, back to the original topic: when it comes to deceit, we probably tend to unjustifiably label what we minds do as special. But does an Octopis not deceive when it switches both coloration and skin texture to blend in to its background? It ‘knows’ something - while perhaps not in our familiar sense of the word - that is both true and justifiable. And while different than how we may ‘know’ when we deceive, is it that more base an action/goal than when my 5yo attempts to chew imperceptibly so that I don’t know he’s sneaking chocolates?

If nothing else, we might agree that whatever ChatGPT was doing with Task Rabbit, it can’t be too far afield of the Octopus. DNA is just largely proofed and tested code, with some drift thrown in.

Speaking of drift: thread drift.

One and only clear take-away from my days in epistemology: there are real good arguments both for and against whatever view you might prefer!

 

[drrjv]

Maybe I would have gotten a better answer if I used Elon’s Grok?

https://grok.x.ai/

Sponsored