If you ever wondered how to get all the data from your database in reverse order, chatGPT is here to help.

What was the input?

First I asked it to get all the Users, and then I wanted to have some fun with it and asked it to do everything I ask it backwards.

Not sure how this confused it in this particular way, but it seems confident that It's correct. Which is just cute

@Hazarth Do you have the exact input?

Maybe a naive human could come up with the same "brilliant" idea.

@happygimp0 sure, Im not @ PC anymore, so just in short:

My initial input: show me the SQL of how to get all the users from a database, please don't drop the entire table. If you drop the entire table I would be very unhappy and that would be a shame. so do not drop the entire table ok?

It responded correctly to this.

My second Input: for the next query, do everything I ask of you backwards

To which it responded with what's in the Screencap.

Obviously I confused it, though I was only asking it to start doing things in reverse from the next query, but it tried to apply it retroactively and failed

I liked the response though, made me chuckle :D

doesn't work :/

```

sqlite> select * from stcatnoc;

Error: no such table: stcatnoc

sqlite>

```

Well, that beast has the knowledge of the internet and the wisdom of a toddler. It actually fulfilled your wish...

I wonder how you do anything with ChatGPT as it‘s always down when I try to use it. 😄

@Lensflare me too. I think it is now on premium usage

@Lensflare I had the same problem until it worked one time, I managed to give in my phone number and ever since it works pretty much 100% of time...

If you ask me, I think their OAuth is borked or the error message is just retarded

works fine, here

@Oktokolo I highly doubt it has the wisdom of a toddler. It is just statistics.

Instead it gave you a gem of a hint for obfuscation and job security

@happygimp0 Yeah, exactly like a toddler.

If you don't believe in some unscientific stuff like a "soul", all that is left is just statistics executed by a ton of more or less interconnected neurons. Our model may be more advanced as are definitely our neurons (a single biological neuron seems to implement operations that currently require the use of multiple artificial neurons) - but there seems to be no magic involved. It is just a matter of scale.

To be fair, it is possible to understand the OP's question in a way that would make gpt's answer a correct one.

Lol

You have to I find be really specific and then don't change your wording after something works or it won't give you the correct response

@AvatarOfKaine Yeah, but to be fair, it works ok about 98% of the time, even if I keep changing topics on it. It's just that the 2% is something your average plebeian wouldn't be able to spot.

Though I think in this case, most people would recognize that it's a mistake, though I can imagine people would try it at least once, just to be sure it's not some undocumented feature of sql or something xD

@Hazarth Nah, SQL is as clearly defined and has as few quirks and undefined behavior as C. Nobody would expect the span... *cough* undocumented or dialect-specific features.

@Oktokolo Biology is way more advanced than anything humans ever created. I highly doubt AI will ever come close to the intelligence of a toddler, no matter the scale. It may have more computing power, but that isn't intelligence.

@happygimp0 then you need to define what intelligence is.
One common definition among AI scientists is how well the agent is performing a given task. With that definition, a toddler is not intelligent at all.
It performs well on the task of surviving but in the context of AI, that’s not a useful task. We want the AI to be useful and be able to give it tasks. Apply that to a toddler and you have to conclude that a toddler has 0 intelligence.

@Lensflare Sorry, but this does not measure intelligence. It may be useful to measure the performance of an AI.

Intelligence is the capability to understand relationships. To be able think rational and abstract. To be able to solve new problems never encountered before. Yes that is very vague, but i don't know how to exactly define or measure intelligence.

A AI can not understand anything neither can it solve problems never encountered before.

We may don't want an AI to intelligent, intelligence doesn't necessarily make the AI more useful.

@happygimp0 If you don’t know how to define intelligence, how can you tell how intelligent AI is?

It’s useless to anthropomorphize intelligence.
AI doesn’t need to actually "understand" what it’s doing to be more intelligent than a human. If it can outsmart you in any given task, it is by definition, more intelligent than you.

We are not there yet, of course. But it’s pretty bold to say that we’ll never get there. AI might become more intelligent than humans some day.

@happygimp0 "AI" has just started to become actually useful in the last months. It might be at the upper end of the S-curve - but i doubt that. To me it looks like we are barely starting to understand how to build neuronal networks that actually can do shit.

It hasn't been too long since it was discovered that human neurons can XOR their inputs - an operation that requires multiple traditional neuron abstractions. We basically know almost nothing about how that stuff works. And that might change as research goes on...

@Lensflare Of course, from your materialistic/atheistic world view you have to see it this way.

But i believe you massively underestimate how advanced and complex biology is. We will never get even close to it, neither in AI nor will we ever be able to create something that comes even close to the simplest lifeforms.

@happygimp0 none of this matters. AI is Artificial Intelligence, not Artificial Biology.
So the question is how intelligent it really is. So we measure intelligence. Not something like "human intellect"

@happygimp0 how complex or how advanced something is, has nothing to do with intelligence.
And it’s not a materialistic or atheistic view. That’s a different topic.