. @OpenAI's #Codex is to programming as Tesla's FSD 2021 is to driving.
Read the paper (esp Appendix B) carefully and you will realize there is a gap between the slick videos & reality: it is often correct on simple tasks, but frequently lost on more complex challenges.
1/3
OpenAI #Codex livestream worked because execs pre-tested & knew what Codex is good for (bite-sized tasks), steering clear of what it is weak at (understanding programming task as a whole).
Codex is amazing, but like current self-driving systems, it may never be reliable.
2/3
What do I think of @OpenAI's new 1st grade math video?
No published results, no peer review. We don't know what the training set is, what the accuracy is, nor how robust results would be to slight changes in wording.
PR ≠ Science
3/3
• • •
Missing some Tweet in this thread? You can try to
force a refresh
Let's start with a simple example drawn from my 2001 book The Algebraic Mind, that anyone can try at home: (2/9)
Train a basic multilayer perceptron on the identity function (ie mulltiplying the input times one) on a random subset of 10% of the the even numbers, from 2 to 1024, representing each number as a standard distributed representation of nodes encoding binary digits. (3/9)