[Double_a_92]

More like: "ChatGPT can pretend to deobfuscate the code"

[yodon]

Are you disagreeing with the deobfuscation that ChatGPT produced?

[shever73]

According to Alex's own tweet, ChatGPT originally produced the obfuscated code too.

In my head, this is like asking someone to translate "Hello" into French and then asking them to translate "Bonjour" back to English. It proves nothing about capabilities or usefulness.

[twic]

This is the correct response.

Especially given that the "obfuscated code" is not syntactically valid. Even if you repair the syntax, it contains a number of errors, and eventually descends into gibberish (although some of what it does is not bad!). There is nothing to deobfuscate here.

[zekica]

I'm disagreeing that the obfuscated code even works. You shoudn't be able to deobfuscate code that doesn't work and get a resulting code that works.

[KomoD]

It's just guessing and happened to be right once

[o_m]

You can double check the answer and still save a ton of time.

[throwawayway99]

If you're double checking every line, how much time are you saving?