LLMs, IIRC, are really bad at IQ-test type questions that require abstract reasoning, especially if they require multiple steps. So, something like
The box is yellow and red.
If the box is yellow, it is good.
If the box is blue, it is unhappy.
If the box is good and happy, the box is awesome.
If the box is red, it is happy.
Is the box awesome?
According to the provided statements, the box is yellow and red.
“If the box is yellow, it is good.” - So, since the box is yellow, it is good.
“If the box is red, it is happy.” - And since the box is red, it is happy.
Finally, “If the box is good and happy, the box is awesome.” - Therefore, because the box is both good (due to being yellow) and happy (due to being red), the box is indeed awesome.
Bard however, does not seem to get the answer right:
The answer to the trick question is indeterminate.
The box is both yellow and red, so it is both good and happy. However, the box is not blue, so it is not unhappy. Therefore, the box could be awesome, but it could also not be awesome.
The trick question relies on the ambiguity of the word “awesome.” In some contexts, “awesome” means “extremely good.” In other contexts, “awesome” means “inspiring awe.” The box could be either of these things, or it could be neither.
Without more information, it is impossible to say definitively whether or not the box is awesome.
Seems like it got kind of close, with the “The box is both yellow and red, so it is both good and happy”… but then falls apart afterwards.
Edit: I tried to debate with it:
Me:
To me, the answer is “Yes, the box is awesome” because the box is yellow thus it is “good” and it is also red, thus it is “happy”. Because the box is “good” and “happy”, the question defines the box as “awesome” - if that all makes sense!
Bard:
Yes, your interpretation makes sense. The box is yellow, so it is good. The box is red, so it is happy. Therefore, the box is both good and happy, which means that the box is awesome.
However, as I mentioned before, the word “awesome” is ambiguous. It could mean “extremely good” or “inspiring awe.” In the context of this trick question, it is not clear which definition of “awesome” is intended. Therefore, the answer to the question is still indeterminate.
Ultimately, the interpretation of the trick question is up to the individual. If you believe that the box is awesome because it is both good and happy, then that is a valid interpretation. However, if you believe that the word “awesome” has a more specific meaning, then you may interpret the question differently.
Which is interesting to say the least, its almost like its looking a bit too deeply into the question lol.
Not surprised. I got access to bard a while back and it does quite a lot more hallucinating than even GPT3.5.
Though it doubling down on the wrong answer even when corrected is something I’ve seen GPT4 do even in some cases. It seems like once it says something, it usually sticks to it.
The box is yellow and red.
If the box is yellow, it is good.
If the box is blue, it is unhappy.
If the box is good and happy, the box is awesome.
If the box is red, it is happy.
Is the box awesome?
To answer this question, we need to check if the box is both good and happy.
We know that the box is yellow and red.
We also know that if the box is yellow, it is good.
Therefore, we can conclude that the box is good.
We also know that if the box is red, it is happy.
Therefore, we can conclude that the box is happy.
Since the box is both good and happy, we can use the rule that if the box is good and happy, the box is awesome.
I was going to say you could give it a math problem that uses big numbers but tried one on GPT4 and it succeeded. GPT3 though will absolutely fail at nontrivial math every time.
Yeah. It still definitely does! The interesting thing is that it seems to be very good at estimating and the final answer it gives is usually pretty close to correct in my experience. Of course close doesn’t really count in math problems.
LLMs, IIRC, are really bad at IQ-test type questions that require abstract reasoning, especially if they require multiple steps. So, something like
The box is yellow and red.
If the box is yellow, it is good.
If the box is blue, it is unhappy.
If the box is good and happy, the box is awesome.
If the box is red, it is happy.
Is the box awesome?
is what I’d use.
Thats just an algebraic equation but with words.
This sounds like something chatGPT could easily convert. Although that is computationally high.
It would also have the side affects of blocking all math illiterate people from creating accounts.
A + B = C
A = X
D != Y
X + Y = Z
B = Y
Does C = Z?
exactly
From ChatGPT 4:
Yes, the box is awesome.
According to the provided statements, the box is yellow and red.
“If the box is yellow, it is good.” - So, since the box is yellow, it is good.
“If the box is red, it is happy.” - And since the box is red, it is happy.
Finally, “If the box is good and happy, the box is awesome.” - Therefore, because the box is both good (due to being yellow) and happy (due to being red), the box is indeed awesome.
Bard however, does not seem to get the answer right:
Seems like it got kind of close, with the “The box is both yellow and red, so it is both good and happy”… but then falls apart afterwards.
Edit: I tried to debate with it:
Me:
Bard:
Which is interesting to say the least, its almost like its looking a bit too deeply into the question lol.
Not surprised. I got access to bard a while back and it does quite a lot more hallucinating than even GPT3.5.
Though it doubling down on the wrong answer even when corrected is something I’ve seen GPT4 do even in some cases. It seems like once it says something, it usually sticks to it.
Bing had no trouble
Bing is GPT4 based, though I don’t think the same version as ChatGPT. But either way GPT4 can solve these types of problems all day.
That’s terrifyingly good wtf
I was going to say you could give it a math problem that uses big numbers but tried one on GPT4 and it succeeded. GPT3 though will absolutely fail at nontrivial math every time.
deleted by creator
Yeah. It still definitely does! The interesting thing is that it seems to be very good at estimating and the final answer it gives is usually pretty close to correct in my experience. Of course close doesn’t really count in math problems.
deleted by creator
Um wtf, I’m starting to doubt if I’m a human. 🤔