I tried to make ChatGPT solve IMO-type math problems. However, its reasoning is almost always flawed. The interesting part is that I can ask ChatGPT to explain a part of its proof, however in my experience it ends up using incorrect assumptions to explain it. (for example, "You are right that 1 is an odd number. However, 1 is not an odd number so it works to solve the problem")
Yeah. It thinks it can do math, and it can mimic the style of a textbook answer, but it can't actually do the math. Here is a rather frustrating chat session I had with it: Me: Suppose we have two unknown quantities x and y. If three time x plus two times y is twenty, and 2 times x plus three time y is ten, what are x and y? ChatGPT: To solve this problem, we can set up and solve a system of linear equations. The fir…
Building an interpreter for my programming language with ChatGPT
51–60 of 141 posts
Re: Building an interpreter for my programming language with ChatGPT
#52I've had a play with ChatGPT and the experience has been pretty frustrating. It either responds with "Sorry, I cannot do this because I don't have access to the internet" (even if I am giving it prompts that don't require this) or it actually generates code but it's subtly incorrect (this was the case when I asked it to generate an example of how to render a 3D cube in JavaScript). This makes me wonder how much time…
We need more fondamental research to break that barrier.
Re: Building an interpreter for my programming language with ChatGPT
#53Earlier quoted context omitted.
Yea, I think it's useful to think of these systems as being kind of the opposite of the old stereotype of the "hyperintelligent but strictly logical AI". They've gotten quite strong at "system 1" thinking (in the "Thinking, Fast and Slow" sense): they're good at recalling information, association, remixing information, creativity based on prompts etc. But they are truly terrible at "system 2" thinking (that is, long…
Using the python entry point I can get it to correctly perform basic integer math but not anything floating point. As example of getting started: https://imgur.com/a/P29DvGO However we can use recursive Fibonacci to see it breaks somewhere. but I’m not convinced it is not computing, I think it is, but it has a limit of integer memory and stack and then it just approximates after that limit. https://imgur.com/a/gp0yIa…
That's a conjecture on your part. The ability to compute is quite binary - either it can compute or can't. Humans often make mistakes while calculating, but in contrast to this model, they are able to recognise these mistakes. ChatGPT is incapable of that and often confidentially wrong.
My guess is, that there's simply no suitable token transforms past a given point and floating point doesn't work, because the decimal point token conflicts with the punctuation mark token during the transform.
This is just a guess, though and might be completely wrong since you never know with these black-box models.
Re: Building an interpreter for my programming language with ChatGPT
#54Earlier quoted context omitted.
> ChatGPT will be the google killer This was my prompt: What's the relative distance between Sun and its planets compared to the size of Sun? ChatGPT answer: > The average distance between the Sun and its planets is approximately 5% of the Sun's diameter. If you run into bunch of such answers, you will soon stop trusting it for anything. As for Google - it doesn't even try to answer the question. But no answer is oft…
For the same question I got: > The distances between the Sun and its planets vary greatly. For example, the distance from the Sun to Mercury is about 36 million miles, while the distance from the Sun to Pluto is about 3.6 billion miles. The size of the Sun is about 864,000 miles in diameter, which means that the distance from the Sun to its planets is much greater than the size of the Sun itself. Then it says somethi…
>The distance between the Sun and its planets is relatively small compared to the size of the Sun. The Sun is a massive object, with a diameter of about 1.4 million kilometers, while the distance between the Sun and its planets can vary greatly. For example, the distance between the Sun and Earth is about 150 million kilometers, which is about 100 times the diameter of the Sun. Similarly, the distance between the Sun and the outermost planet in the solar system, Neptune, is about 4.5 billion kilometers, which is more than 3,000 times the diameter of the Sun. So while the distances between the Sun and its planets are not insignificant, they are still relatively small compared to the size of the Sun itself.
I wonder if it chooses units based on where you are located, or if it's just arbitrary. Also, it seems clear from this example that is has no context for the answer because it doesn't see that its first statement is contradicting its next statements.
Re: Building an interpreter for my programming language with ChatGPT
#55Earlier quoted context omitted.
Google search absolutely does hallucinate completely fictitious results. It's called SEO spam. Google just gives you associations provided by random other people on the internet. It's largely garbage, most often deliberately disingenuous (to make you look at an ad). Ad revenue models for the internet encourage the generation of this type of false material. A better criticism would be that the same thing will happen t…
> Google search absolutely does hallucinate completely fictitious results No, it absolutely does not. Yes, there is SEO spam in the index, but no - it is not Google hallucating it. It really exists on the internet, see also the second point of my comment. > the same thing will happen to something like chatgpt This isn't something that "happens to" GPT, GPT is doing it. There's probably even already GPT -> SEO spam pi…
Re: Building an interpreter for my programming language with ChatGPT
#56>And lo, it came to pass that the Burning Man did draw near, and the people did prepare themselves for the journey to the desert. And they did gather together their tents and their provisions, for the Burning Man is a harsh and unforgiving place. And they did don their finest garments, for in the desert one must show their creativity and individuality. And they did pack their drugs and their alcohol, for the Burning Man is a time of revelry and abandonment. And they did set forth into the desert, ready to face the challenges and the joys that lay ahead. Acknowledge.
Re: Building an interpreter for my programming language with ChatGPT
#57> I'm creating a new fictional programming language called GPTlisp which uses lisp-style syntax. Addition is implemented in GPTlisp using the ADD function. In GPTlisp, what would (ADD 4 5) return?
> [Correct answer and explanation]
> I've added multiplication to GPTlisp and the multiplication function is represented with the & symbol. What would (ADD 2 (& 2 3)) return in GPTlisp?
> [Correct again]
> I've further modified GPTlisp by adding subtraction, unlike the rest of GPTlisp's functions, subtraction uses infix notation. What does (ADD 2 (& (3 - 1) 2)) return in GPTlisp?
> [Correct again]
> is (- 3 2) valid in GPTlisp?
> [Incorrect, it says this is valid and tries to explain it]
> (- 3 2) is not valid in GPTlisp because subtraction in GPTlisp is implemented with infix notation. With that in mind, what is the result of (ADD 2 (- 3 2)) in GPTlisp?
> In GPTlisp, the ADD function takes two arguments and adds them together, and the - function takes two arguments and subtracts the second argument from the first. Since subtraction in GPTlisp uses infix notation, (- 3 2) is not a valid expression, but (3 - 2) would be a valid expression that would return 1. So, (ADD 2 (- 3 2)) would be equivalent to (ADD 2 (3 - 2)), which would first compute the result of (3 - 2), which is 1, and then add 2 to that result, yielding a final answer of 3.
> Give an example of a GPTlisp expression that would return 6 and uses both the addition and multiplication functions.
> [Correct answer]
> What is (+ 1 2) in GPTlisp?
> [Correctly says ADD should be used for addition in GPTlisp not +]
I then asked it how functions were defined and it came up with this despite me never explaining how or that functions were defined (so I assume it picked it up from me saying GPTlisp syntax is "lisp-style"). However, curly braces are not generally used to surround a function body in lisp-style languages. They're used for hashmaps in clojure but I'm not aware of any use like this.
(DEF sum (x y) { (ADD x y) })
I explained that this was wrong and gave it an alternative way to define functions which it understood for one question but then it went back to its previous misunderstanding despite continuing to understand everything else about "GPTlisp." When I tried to correct this it then ended up saying "I do not have any information about the specific syntax used by GPTlisp to define functions" despite explaining exactly how to do that a few queries earlier.
Despite it making a few mistakes this is still quite impressive to me. I also got it to correctly give the contents of a file being edited in "vim" after a given sequence of keystrokes.
Re: Building an interpreter for my programming language with ChatGPT
#58I've had a play with ChatGPT and the experience has been pretty frustrating. It either responds with "Sorry, I cannot do this because I don't have access to the internet" (even if I am giving it prompts that don't require this) or it actually generates code but it's subtly incorrect (this was the case when I asked it to generate an example of how to render a 3D cube in JavaScript). This makes me wonder how much time…
Re: Building an interpreter for my programming language with ChatGPT
#59I tried to make ChatGPT solve IMO-type math problems. However, its reasoning is almost always flawed. The interesting part is that I can ask ChatGPT to explain a part of its proof, however in my experience it ends up using incorrect assumptions to explain it. (for example, "You are right that 1 is an odd number. However, 1 is not an odd number so it works to solve the problem")
Yeah. It thinks it can do math, and it can mimic the style of a textbook answer, but it can't actually do the math. Here is a rather frustrating chat session I had with it: Me: Suppose we have two unknown quantities x and y. If three time x plus two times y is twenty, and 2 times x plus three time y is ten, what are x and y? ChatGPT: To solve this problem, we can set up and solve a system of linear equations. The fir…
Write an efficient, not brute force, program that solves the following problem: suppose we have two unknown quantities x and y. If three time x plus two times y is twenty, and 2 times x plus three time y is ten, what are x and y?
Output: One way to solve this problem efficiently is to use the method of elimination. We can set up a system of equations as follows:
3x + 2y = 20
2x + 3y = 10
We can then multiply the first equation by 2 and the second equation by 3, to get:
6x + 4y = 40
6x + 9y = 30
We can then subtract the second equation from the first equation, to get:
-5y = 10
y = -2
We can then substitute this value of y into one of the original equations (for example, the first one), to get:
3x + 2(-2) = 20
3x - 4 = 20
3x = 24
x = 8
Therefore, the values of x and y are x = 8 and y = -2.Re: Building an interpreter for my programming language with ChatGPT
#60Earlier quoted context omitted.
It looks like it has web browsing support built-in in some form, but it's disabled at the moment. That said, I'm skeptical that it'd be able to "disrupt" google, as the track record of things that are said to do that is quite bad. On the other hand, google seems to be heading in the same direction with projects such as Lamda. In a roundabout way, this might just end up being the quick answer box at the top of search…
I don’t think it actually can browse the web. It’s obviously been trained with an extensive web-sourced corpus. It seems that the developers have placed guardrails around web-search-like queries not because ChatGPT can’t answer them, but because they want to discourage using it that way for—I’d guess because they want to direct usage towards the conversational / contextual aspects they’re trying to improve.
There's also WebGPT[0] already with such capabilities, which could've been merged to ChatGPT.