> Why are we flabbergasted that GPT can still infer that info with mild obfuscation? May be this was "obvious," but like a lot of research there's a bit of hin

> Why are we flabbergasted that GPT can still infer that info with mild obfuscation? May be this was "obvious," but like a lot of research there's a bit of hindsight bias in that sentiment. The issue now is that while you can easily do this yourself in terms of inference, the problem is GPT4 allows this inference to be automated.