Hacker News Reader

Comment by ashetr | Hacker News Reader

What does that have to do with the Turing Test? The TT has clear rules: There are judges that have a dialogue with anonymized AI/humans. The humans cannot cheat and impersonate a machine, they have to act normally. The AI obviously should try to sound human.

No AI would pass this test with experienced judges.

View on HN

You think current frontier models couldn't pass for a human on an online chat? You and I have very different perceptions of reality.

Sounds like something we could settle right here and now.

Ha ha! Fool! You've been talking to an LLM all this time! Your wife is actually Haiku 4.5.

Huh, should've known it was odd for my wife to always say I'm right (as the boomers would say)

I think that if you have a long enough chat, yeah, I think you can figure out who's meat. The original rules for the TT specified a short interaction, but I can probably accelerate it by pasting in large code snippets to force early compactions.

Humans are famously good at reading large code snippets.

They are famously good at ignoring large code snippets.

The test doesn't say that the judge has to be experienced. But I also don't care if some random gullible person can't tell the difference. Nothing passes the Turing Test for me yet.

Edit: Also doesn't say anything about who the human test subject is

Of course, and I'm sure OP wouldn't disagree with you, it was clearly a joke for emphasis. Some people round here need to clean and calibrate their humour detectors more often.

I wasn't responding to OP. I get the emphasis, the M6 Mac is very powerful.

How can you be certain you haven’t failed a Turing test?

I've never done a test. That means 1. you know it's a test 2. you get like 5 back-and-forths or 5 minutes 3. you have a human subject to compare to. But pretty sure I'd pass anyway, as the judge or subject.

Colloquially the Turing test is just a stand in for "can a human mistake a computer for a person." No need to overcomplicate it.

You’ve moved the goalposts.

You can always say “oh well these judges don’t have the experience to catch this type of AI.

The fact that you have to insert this qualifier, to ensure you always have a way to discredit the test, pretty much shows to me that we’re beyond it.

Reply on news.ycombinator.com