I have a feeling that their test case is also a bit flawed. Trying to get index_value instead of index value is something I can imagine happening, and asking an LLM to ‘fix this but give no explanation’ is asking for a bad solution.
I think they are still correct in the assumption that output becomes worse, though
Yeah, reminds me a bit of bullshit titles like these:
No shit, your brain is also you
It seems that you can't invoke your rights pre-Miranda and before arrest, as seen in https://en.wikipedia.org/wiki/Salinas_v._Texas
I don't know how that works specifically, but I expect nobody will allow you ~~more~~ your rights whenever they can
It can be turned off after boiling, but it also has much better thermal insulation, so the water will stay hot for much longer than in an ordinary kettle
That's the real answer
You can transpile from C to Malbolge and then run it (this will probably take forever for most of C programs). I thought it can be used for obfuscation, and sure enough Wiki already states that:
Hisashi Iizawa et al. also proposed a guide for programming in Malbolge for the purpose of obfuscation for software protection
No thinking is not the same as no actions, we had bots in games for decades and that bots look like they act reasonably but there never was any thinking.
I feel like ‘a lot of agency’ is wrong as there is no agency, but it doesn't mean that an LLM in a looped setup can't arrive to these actions and perform them. It doesn't require neither agency, nor thinking
Maybe ‘feh-teh-peh’, but yeah, sometimes I see ‘an’ or ‘a’ where I don't expect them and am left wondering about mysterious ways of the English language
But if they go mom, they be a woman
It's kinda good but also reads a bit like
sorry we got caught we never intended to
Oh, this perspective didn't occur to me, it makes everything so much worse 😅
Yeah, if only QA vere not the first ‘replaced’ by AI 😠