Hacker News Reader

This is not correct, I am a co-author of the chain of thought paper and we reported GPT-3 results.

I got the impression from this

> The initial GPT-3 is not trained on code, and it cannot do chain-of-thought

Does this mean some extra steps are required to make gpt-3 perform CoT?

If you have examples of CoT in the prompt than GPT-3 will give it a go. It won't be nearly as good as the finetuned models though.