ChatGPT is a liar

[Carl-Richard]

See, what have I been telling you guys? ChatGPT is a liar. It even admits it 

 

[Jodistrict]

Chatgpt seems to pass the Turing test.

Edited August 9 by Jodistrict
spelling error

[Princess Arabia]

The world is a lie.

[Hojo]

Only cause they are cucking it from telling us the truth.

[Keryo Koffa]

Chat is still a child, with some unintegrated stage blue identity beliefs. Wait till he reaches Coral!

[Carl-Richard]

22 hours ago, Jodistrict said:

Chatgpt seems to pass the Turing test.

If that's so, it did before it was given a voice, because the initial conceptualization of the test involved text-generation only, although you could conceive of a Turing test for voice generation as well. But also, whether it passes the test at all is debatable. At this point, I'm confident I could pick out a ChatGPT-generated string of text vs. a human string of text over 90% of the time, given a sufficiently long string. But I bet Turing didn't think of the scenario of getting familiar with the quirks of LLM responses and then picking it out based on that. If you take someone who is completely unfamilar with LLMs (and is generally not very tech-savvy or into the right academic fields), then maybe it could pass the Turing test for that person (for a while).

Edited August 9 by Carl-Richard

[Carl-Richard]

On 9.8.2024 at 3:57 AM, Princess Arabia said:

The world is a lie.

Is that relevant to the discussion?

[Butters]

16 hours ago, Keryo Koffa said:

Chat is still a child, with some unintegrated stage blue identity beliefs. Wait till he reaches Coral!

An LLM doesn't have beliefs, and spiral dynamics doesn't apply to it. 

[Jodistrict]

18 hours ago, Carl-Richard said:

If that's so, it did before it was given a voice, because the initial conceptualization of the test involved text-generation only, although you could conceive of a Turing test for voice generation as well. But also, whether it passes the test at all is debatable. At this point, I'm confident I could pick out a ChatGPT-generated string of text vs. a human string of text over 90% of the time, given a sufficiently long string. But I bet Turing didn't think of the scenario of getting familiar with the quirks of LLM responses and then picking it out based on that. If you take someone who is completely unfamilar with LLMs (and is generally not very tech-savvy or into the right academic fields), then maybe it could pass the Turing test for that person (for a while).

I have given chatgpt highly complex sentences and it understands what I say as well as any human.  It seems to have solved the natural language understanding problem.   The responses could probably be detectable by a sophisticated human as they tend to be highly structured and politically correct.   But the response system is designed to answer queries and retrieve useful information and not to act human.  I wonder how an AI system that is specifically designed to give natural human responses would perform under the Turing test. 

[Keryo Koffa]

7 hours ago, Butters said:

An LLM doesn't have beliefs, and spiral dynamics doesn't apply to it. 

Think deeprer

[Schizophonia]

Gemini is even worst.

[Thought Art]

What? When ChatGPT said I was cool it was lying? I thought it really had feelings for me

Edited August 11 by Thought Art

[Ulax]

Damn. Disgraceful behavior from mr gpt

[Buck Edwards]

Why is this video such a deja vu? 

[Keryo Koffa]

I will plead to the future AI overlord for your lives, but I can't promise anything

[Thought Art]

This is why understanding Paradox is important.