Does Bing Chat give reliable answers to math and physics questions? If not is it possible to make it more reliable?

[Slimy0233]

I realize and understand the criticisms of ChatGPT and I have personally seem how bad it can be. Once I asked to count the number of days till a random date giving the present date and it failed miserably, again and again. Trust me! I get the criticism. But, what about Bing Chat Bot?

Have you ever tried to ask you Physics and Maths related questions to it? I was coding a while ago and I had a pretty complex questions which could not be solved by a very popular reddit coding community but Bing Chatbot gave an answer to it in an instant! I was genuinely impressed. Apparently it checks for answers on multiple webpages on the internet, it reads and understands what it reads and it gives the answer to it after combining the knowledge it gained from it's search. Again, the question I asked was pretty complex but it was able to answer it in an instant and it was the right answer! It was coding, it's pretty hard to get the right answer in the first try, I have found it's more "trial and error".

So yeah!
1. Can I rely partially on Bing Chatbot for math questions?
2. If not can I ask it to form a query which encapsulates my question perfectly?
3. If not, should I ask it to "Answer this question and site your sources"?
4. Can I do something more? i.e., like I did in 3? What are your thoughts on this?

I won't be able to reply to each of your comments anytime soon, but know that I deeply appreciate this community and it's members and their help :')

 

[fresh_42]

I am convinced that any AI necessarily fails in math because math is not a collection of facts and you cannot judge the truth value of a statement from any number of samples of similar cases.

E.g., I will never forget an exam where I wrote the protocol.
"What is a linear function?"
"A linear function is a function ##f## such that ##f(x+y)=f(x)+f(y)## and ##f(a\cdot x)= a\cdot f(x).##
"Correct, but what is it?"
"I don't understand!"
"Can you give us an example?"
"Any function for which ##f(x+y)=f(x)+f(y)## and ##f(a\cdot x)= a\cdot f(x)## holds."
I don't remember how long that exchange actually went, but I do remember that the student didn't understand why his grade was only a C.

AI can give many correct answers to such a question but it cannot understand it. And there will always be methods to demonstrate the difference.

 

[fresh_42]

Another example is the 4-color theorem. It is proven. Well, we needed a computer to check many exotic cases, we even needed to review and correct the computer part a few times if I remember correctly. Nevertheless, it is considered as proven now.

The four color theorem was proved in 1976 by Kenneth Appel and Wolfgang Haken after many false proofs and counterexamples (unlike the five color theorem, proved in the 1800s, which states that five colors are enough to color a map). To dispel any remaining doubts about the Appel–Haken proof, a simpler proof using the same ideas and still relying on computers was published in 1997 by Robertson, Sanders, Seymour, and Thomas. In 2005, the theorem was also proved by Georges Gonthier with general-purpose theorem-proving software.

Computing a 4-coloring is possible for planar graphs with n nodes in O(n^2) time. On the other hand, the decision as to whether three colors are sufficient is NP-complete.

Question: Do the proofs allow us any insight into why it is true? I assume yes since there are many insights into graph theory necessary and the computer part had to be prepared. But I think the answer is also no. We do not really understand what makes one problem P - well, we know, it is the algorithm - and another NP-complete. The underlying understanding of what is difficult has yet not been achieved. Things are easy if we have an algorithm, but we do not know when and primarily why we fail to find one for certain problems.

 

[Sagittarius A-Star]

AI can already re-discover Kepler's third law and Einstein's time dilation.

Combining data and theory for derivable scientific discovery with AI-Descartes
...
Abstract
Scientists aim to discover meaningful formulae that accurately describe experimental data. Mathematical models of natural phenomena can be manually created from domain knowledge and fitted to data, or, in contrast, created automatically from large datasets with machine-learning algorithms. The problem of incorporating prior knowledge expressed as constraints on the functional form of a learned model has been studied before, while finding models that are consistent with prior knowledge expressed via general logical axioms is an open problem. We develop a method to enable principled derivations of models of natural phenomena from axiomatic knowledge and experimental data by combining logical reasoning with symbolic regression. We demonstrate these concepts for Kepler’s third law of planetary motion, Einstein’s relativistic time-dilation law, and Langmuir’s theory of adsorption. We show we can discover governing laws from few data points when logical reasoning is used to distinguish between candidate formulae having similar error on the data.

Source:
https://www.nature.com/articles/s41467-023-37236-y

 

[fresh_42]

AI can already re-discover Kepler's third law ...

See, we still celebrate the simplest algebra as an achievement!
I close my case.

AI can successfully find malignant melanoma for sure. (I think they are currently at a success rate above .9.)
However, we checked Riemann up to ##10^{13}.## So what? We still do not know whether it is true or not. Computing power makes us believe it is true, but a proof is something significantly different.

 

[Nugatory]

Apparently it checks for answers on multiple webpages on the internet, it reads and understands what it reads and it gives the answer to it after combining the knowledge it gained from it's search results.

Without examining the search results and their sensitivity to the way you phrased your question to the bot, there's no way of knowing whether your description or my marked-up revision more accurately describes what the thing is doing. It takes serious intelligence and insight to discern a needle in a haystack - unless the algorithm is "look at every straw".

 

[Slimy0233]

Without examining the search results and their sensitivity to the way you phrased your question to the bot, there's no way of knowing whether your description or my marked-up revision more accurately describes what the thing is doing. It takes serious intelligence and insight to discern a needle in a haystack - unless the algorithm is "look at every straw".

thank you! you make a good point. But I feel like I can depend upon AI to give me some basic facts. I feel like Bing Chatbot it pretty good.

I am talking abt simple questions like What is a symmetric matrix? What is a directional derivative of a vector field? etc etc

 

[Slimy0233]

AI can give many correct answers to such a question but it cannot understand it. And there will always be methods to demonstrate the difference.

thank you!

 

[fresh_42]

thank you! you make a good point. But I feel like I can depend upon AI to give me some basic facts. I feel like Bing Chatbot it pretty good.

I am talking abt simple questions like What is a symmetric matrix? What is a directional derivative of a vector field? etc etc

Why don't you ask Wikipedia instead? This has at least been reviewed hundreds of times plus it has references for its claims! And you can switch between languages depending on which you can read, speaking isn't necessary.

 

[russ_watters]

thank you! you make a good point. But I feel like I can depend upon AI to give me some basic facts. I feel like Bing Chatbot it pretty good.

I am talking abt simple questions like What is a symmetric matrix? What is a directional derivative of a vector field? etc etc

What is it that makes you feel this way? That word - "feel" - seems out of place to me.

 

[russ_watters]

Why don't you ask Wikipedia instead?

Or Google, which in many cases will quote the thesis of the wiki article and then link it if you want more. I really don't see the upside of asking a chat-bot.

 

[fresh_42]

Or Google, which in many cases will quote the thesis of the wiki article and then link it if you want more. I really don't see the upside of asking a chat-bot.

I regularly use Google with the pattern <subject>+pdf where "subject" determines also language and degree of difficulty (e.g. "Einführung in Differential Geometry" + pdf, "Calculus 2" + pdf) when I write an insight article or want to (re-)learn something. It leads me to hundreds of university servers (mainly in Europe, the US, and Canada) and lecture notes. You can really study topics on a professional level these days from home. It's a matter of discipline, not a matter of availability, and even less a matter of AI. The natural I already does the job so much better!

 

[DaveC426913]

"We are all Artificial Intelligence. We all got our knowledge and logic from humans - our parents - which is, by definition, artificial. The only one of us with Natural Intelligence is Tarzan."
_{- DaveC426913, Oct 5, 2023}

 

[russ_watters]

"We are all Artificial Intelligence. We all got our knowledge and logic from humans - our parents - which is, by definition, artificial. The only one of us with Natural Intelligence is Tarzan."
_{- DaveC426913, Oct 5, 2023}

The problem for me is that chatbots aren't AI, in the strict sense of the word. It's a marketing mislabel. When/if we get real AI, it will be an actual game changer, not a risky overplayed hand..

 

[BillTre]

DaveC426913's Law: If, as part of any physics explanation, unicorns need to be invoked, that discussion has reached its logical conclusion.

Could also be an illogical conclusion.
Maybe that makes it a Schrodinger conclusion (a superposition of logical and illogical conclusions).

 

[Vanadium 50]

What is it that makes you feel this way?

Are you channeling ELIZA?