4 ms·
Language models can generate a Python function that does the math perfectly. I bet you would get better results if you tweaked the prompt to say "Generate a Py
by yshrestha 4y ago
Language models can generate a Python function that does the math perfectly.
I bet you would get better results if you tweaked the prompt to say "Generate a Python program that solves X math problem" and then just ran the resulting Python script.
It does not need to be AGI to be useful.
- Kim_Bruning 4y agoThat is also a very valid and interesting thing to do. But it's also quite interesting to see how the model would do "by itself". All kinds of interesting lessons to be learned!
- yshrestha 4y agoYeah! It is interesting to try and figure out "what" the model is actually learning. It is a valid thread of scientific inquiry.
- lupire 4y agoYou mean "generate a Python function that calls a library that does math perfectly, right?
- thwayunion 4y agoUnlike 99.99% of human programmers, who can and often do implement everything in sympy/numpy from scratch ;-)
- yshrestha 4y agoExactly! Hey it gets the job done :) Software is just a tall wedding cake of abstractions built on top of abstractions.
- hgomersall 4y agoIn the limit, it's going to design an AI to write some python to call a library that does the math perfectly.
- lairv 4y agoThat could only generate constructivist [0] proofs, and there are many things done in modern maths which are not constructivist. Maybe a better approach would be to use Curry-Howard [1] correspondence to directly get proofs from generated programs [0] https://en.wikipedia.org/wiki/Constructivism_(philosophy_of_mathematics) https://en.wikipedia.org/wiki/Constructivism_(philosophy_of_... [1] https://en.wikipedia.org/wiki/Curry%E2%80%93Howard_correspondence https://en.wikipedia.org/wiki/Curry%E2%80%93Howard_correspon...
- swyx 4y agoyou can also tell the model that it doesnt know how to do math, and it respects that https://twitter.com/goodside/status/1568448128495534081 https://twitter.com/goodside/status/1568448128495534081
- nl 4y agoThis is pretty cool, although the "don't use outside the security sandbox" made me laugh: https://twitter.com/goodside/status/1568704302813700096/photo/1 https://twitter.com/goodside/status/1568704302813700096/phot...
- mlajtos 4y agoExactly, we need computer-equipped neural nets. Models need to use traditional UIs (including programming languages) and then we can talk about how to stop them. :)
- sergeyk 4y agoLive demo of this you can run: https://twitter.com/sergeykarayev/status/1569377881440276481 https://twitter.com/sergeykarayev/status/1569377881440276481