[kaijia]

I dislike this line of research. It just demonstrates large models are capable of memorizing a large number of things, without any understanding.

So I tried the first problem in Appendix A on OpenAI playground.

When I use the prompt "# Sketch the graph of the function f(x) = x + |x|", with the model davinci-codex and a larger response length (other parameters as default), the result seems fine: https://pastebin.com/VT8tPbu6

When I change the prompt to "# Sketch the graph of the function f(x) = x + |x| + x*2", it becomes garbage. It changes the prompt to "# Sketch the graph of the function f(x) = x + |x| + x^2 + x^3 + x^4 + x^5 + x^6 + x^7 + x^8 + x^9 + x^10" and then writes new empty comment lines: https://pastebin.com/2bNEuqaH