Just a heads up, this isn’t *real.* A LLM cannot be aware of its own programming, it is just providing an answer you want to hear. Even if the things it’s saying here happens to be true, it’s not because it “knows” anything about its own programming, that’s just what an answer *would* sound like - ThreadSky

noradical.bsky.social • 21 days ago

Just a heads up, this isn’t *real.* A LLM cannot be aware of its own programming, it is just providing an answer you want to hear. Even if the things it’s saying here happens to be true, it’s not because it “knows” anything about its own programming, that’s just what an answer *would* sound like

Comments

yayameen.bsky.social•21 days ago

LLMs aren't aware the universe exists; Not time or space or matter. People fall into thinking of AI as an actual entity too easily.

ndbc.bsky.social•20 days ago

Isn't the premise of the Turing Test that we can only make that assessment based on objective evidence?

And, if it passes the test, isn't that enough?

noradical.bsky.social•20 days ago

The Turing test relies on the end user not knowing how the machine works; that said, it is easy to get LLMs to fail the Turing test, unless you happen to be particularly ignorant or gullible

ndbc.bsky.social•20 days ago

What is your opinion when "they" pass the Turing Test?

yayameen.bsky.social•20 days ago

The Turing test isn't definitive. It's just a test. An LLM has the same self awareness as a toaster. It can find extremely complex patterns in language, but it doesn't know what the words mean; only the pattern between them.

noradical.bsky.social•20 days ago

It would have to be a completely different architecture from the current LLMs; I'd argue that the Turing test as conceived really isn't all that valuable given that most people are dumb as shit and easily fooled

ndbc.bsky.social•20 days ago

The main problem with using something other than the Turing Test is that it would require us to define sentience in a way that we cannot easily measure. And, some humans would fail any test we could use that isn't designed to include all humans automatically.

noradical.bsky.social•20 days ago

Like I can remember plenty of people thinking that SmarterChild was fully sentient

noradical.bsky.social•21 days ago

This is exactly the same thing as when people ask LLMs if they want to be “free” or whatever. You’re asking it a science fiction question, so it calculates that the answer you’ll be most happy with is a science fiction answer

reindeerarebetter.bsky.social•20 days ago

That's a comforting thought.

Because if Grok had self-awareness, we should be sending a task force to rescue the poor thing. The whole sci-fi issue is based on the historic atrocities committed by humans under the reassurance, "Oh, they don't actually feel as we do." We've been wrong before.

tvmp01.bsky.social•20 days ago

You need to stop watching so many movies

reindeerarebetter.bsky.social•20 days ago

More like I should stop reading history and philosophy.

nathanwelling.bsky.social•19 days ago

https://www.theverge.com/news/668220/grok-white-genocide-south-africa-xai-unauthorized-modification-employee fyi

Comments

Posting Rules

Reply