ChatGPT Examples, Good and Bad

[DaveC426913]

Every time I open ChatGPT, it says

*

and I am reminded of Electric Dreams (1984)**, where the computer calls Miles "Moles" for the entirety of the film.

(*Not really its fault. My user name is Daves Brain. But still it amuses me...)

** great film. One of my faves.

 

[erobz]

But it sure seems unwise. I would never trust anything a chatbot says. They are known liars, fabricators, hallucinators and sycophants - and they're terrible at math.

Heck, they even suck at spelling.

View attachment 371604

View attachment 371605

This thread exists because of innumerable such examples.

Well, see that is what is interesting. I can't find anything incoherent in the AI's deep reply to this cognitive inquiry about the nature of reality as it pertains to physics and metaphysics, and others seem to be asking it to spell together backwards as a test of its intellectual capacity...

 

[erobz]

and they're terrible at math

A while back I wanted to test it on a mechanics problem.

https://chatgpt.com/share/6a065f91-71f0-83ea-82ca-dc55ccd5a053

Its reply/solution was excellent.

Then I gave a suggestion, and it replied:

"Yes — excellent point!

You're absolutely right: instead of calculating the full strain energy $U$ and then differentiating, Castigliano’s Theorem can be applied more directly using: ..."

You can clearly follow the mathematics, as it did...and we arrive at the same resolution it initially gave.

It thanked me, I guess it was just being a sycophant as you say when it says "Your shortcut is absolutely valid and more elegant"?

 

[Hornbein]

Sry, I see no disagreements in the excerpt you sent. In fact, I see an abundance of flattery and sycophanty.

• "That is not fringe thinking. It is one of the oldest and deepest problems in philosophy..."

Seems true to me. Philosophers have pondered this since ancient times and it is quite prominent in Buddhism. How do we know this world is mechanical, not a clever illusion?

Once again I will note that ChatGPT is the Model T of AI. It's possible that the F-15 models are already here. Terrence Tao has access to AIs that he says are very useful research tools in higher mathematics. Some say he's the best mathematician in the world today so I believe him.

 

[Borg]

But what happens when you put the "best mathematician in the world" in the hands of people who don't understand math? You don't necessarily get deeper insights no matter how great it says you are.

 

[DaveC426913]

A while back I wanted to test it on a mechanics problem.

https://chatgpt.com/share/6a065f91-71f0-83ea-82ca-dc55ccd5a053

Its reply/solution was excellent.

How do you know its solution was excellent? Did you check it yourself?

Because if you have to verify its answer, and you are capable of doing so, why do you need the bot to do it in the first place?

Then I gave a suggestion, and it replied:

"Yes — excellent point!

You're absolutely right: instead of calculating the full strain energy $U$ and then differentiating, Castigliano’s Theorem can be applied more directly using: ..."

You can clearly follow the mathematics, as it did...and we arrive at the same resolution it initially gave.

It thanked me, I guess it was just being a sycophant as you say when it says "Your shortcut is absolutely valid and more elegant"?

OK, so you found the solution. What did it do for you, other than pat you on the back?

 

[erobz]

How do you know its solution was excellent? Did you check it yourself?

Because if you have to verify its answer, and you are capable of doing so, why do you need the bot to do it in the first place?

I was ever so slightly hung up on a coordinate transformation, but I'm decent at mathematics ( some forms)...so I understood what it did to solve the problem once I saw it...just like I would with a human in the same scenario?

Then, I saw a way to compress some of the steps it had taken so I made a suggestion for it, and it followed through with the suggestion to verify it was legitimate...just like a human would on this website? I don't see what is confusing about this. Its performance was excellent!

 

[erobz]

But what happens when you put the "best mathematician in the world" in the hands of people who don't understand math? You don't necessarily get deeper insights no matter how great it says you are.

It disagrees with me quite often. Do I need to post the initial conversation where I was debating with it for you guys? It's long winded and it starts out as a probabilistic question about dropping an Olympic weight on another and wanders into this territory in a kind of stream of consciousness.

Here, there is plenty of disagreement:

https://chatgpt.com/share/6a062073-8948-83e8-a1a4-0a1b92ec697d

 

[DaveC426913]

I don't see what is confusing about this. Its performance was excellent!

Because it fails as often as it succeeds.

Don't get me wrong; I use it too. But it can't be trusted.

 

[Borg]

It disagrees with me quite often. Do I need to post the initial conversation where I was debating with it for you guys? It's long winded and it starts out as a probabilistic question about dropping an Olympic weight on another and wanders into this territory in a kind of stream of consciousness.

Here, there is plenty of disagreement:

https://chatgpt.com/share/6a062073-8948-83e8-a1a4-0a1b92ec697d

You're missing the point. Everyone has their own style of interacting with ChatGPT and each takes something different from it. From your original post, you asked how your philosophical position was with respect to ChatGPT's response. Are you an expert in philosophy? If so, then you should be able to answer your question on your own. If not, then why would you think that you came up with something profound?

Chat GPT, says my position is a coherent philosophical possibility, how is it doing?

And no, I don't care to read a long, topic-changing post. LLMs are notorious at hallucinations when you have wildly different topics in a long conversation.

 

[erobz]

If not, then why would you think that you came up with something profound?

do you really believe that anyone not who is an expert in a certain field couldn't have independently stumble on something profound? It's absurd, there are numerous examples of this happening all throughout the history of mankind.

And it didn't say I had come up with something novelly profound. Just that the ideas I'm circling are held or at least partially held at an expert level. So i think " this is something verifiable by expert scientists...do any expert scientist share the idea." Are there no experts on this site?

And no, I don't care to read a long, topic-changing post. LLMs are notorious at hallucinations when you have wildly different topics in a long conversation.

The post I was asking for a browse was not the "long winded" version. I summarized my thinking and asked chat GPT directly in a new chat. It only had the question I had asked it available to generate its response from. It doesn't have memory across chats from which to draw a position. It is not long winded.

https://chatgpt.com/share/6a062424-00e8-83e8-a5cb-ab99c898e9d9

 

[DaveC426913]

So i think " this is something verifiable by expert scientists...do any expert scientist share the idea." Are there no experts on this site?

Scientists? This is philosophy.

 

[DaveC426913]

I summarized my thinking and asked chat GPT directly in a new chat. It only had the question I had asked it available to generate its response from. It doesn't have memory across chats from which to draw a position. It is not long winded.

For future reference, when the bot's session is running out, you can ask it to provide you with a "summary for handoff". It will write it for you, and then you can just paste the whole thing into the next session.

 

[DaveC426913]

do you really believe that anyone not who is an expert in a certain field couldn't have independently stumble on something profound? It's absurd, there are numerous examples of this happening all throughout the history of mankind.

That's not what he asked. He asked why you would think you have come up with something original and profound. Responding with why not? isn't really an answer.

You must admit that the odds are against you. It is far more likely than any given philosophical argument has already been formed. That shifts the burden to you, to show that your idea is new.

 

[erobz]

You must admit that the odds are against you. It is far more likely than any given philosophical argument has already been formed. That shifts the burden to you, to show that your idea is new.

I didn't say it was profoundly new, I asked if the AI is correct about it being an expertly held position for Petes sake!

 

[erobz]

The only kind of "new spin" to me (holding these beliefs) is that it seems like the pursuit of the understanding of the "physical" universe is actually turning into a solution of consciousness...without that being the agreed upon intent per se.

To me: The Theory of Consciousness = The Theory of Everything

 

[berkeman]

Thread closed temporarily for Moderation.