Image not available

1080x1150

Screenshot_202406....jpg

🧵 Turing testing AI

Anonymous No. 16237713

Recent arXiv paper on Turing tests of language models
(https://arxiv.org/abs/2405.08007) judges thought it was human 54% of the time when talking to an AI
But thought it was human 67% of the time with real humans
So like a 13% difference, these AIs are getting pretty human-like.
Models still not perfect but gap is closing as they get better training data and code
Kinda unnerving or just progress? Thoughts from /sci/?

Anonymous No. 16237715

>>16237713
it's all just slop and "passing" the Turing test is meaningless in the sense that these models are literally designed to produce convincing human-like prose
conflating human-like language with human-like intelligence is retarded and no existent models can reason in any way
just midwit hype

Anonymous No. 16237716

>>16237713
Go back to /g/

Anonymous No. 16237720

>>16237715
I do see your point but it's far from meaningless. If the purpose of a Turing test is to see if the machine passes as "human", why not make it.. You know, human-like??

Image not available

1080x2340

Screenshot_202406....jpg

Anonymous No. 16237724

Forgot to mention, as a Turing test for you all, i had Claude 3 write the OP

Anonymous No. 16237738

>>16237713
Sounds like dead internet theory could be just around the corner! Also it's kinda funny that judges thought actual humans were human, only 2/3 of the time.

Anonymous No. 16237749

>>16237715
That's what we do in real life. We imitate.

Anonymous No. 16237754

>>16237713
What is the IQ of the test humans, anyone with over average IQ should tell when it's AI spewing garbage

Anonymous No. 16237807

>>16237754
>>16237738
Don't you fags realise it just PASSED THE TURING TEST AGAIN??

Anonymous No. 16238292

When, in the course of human events, it becomes necessary for one people to dissolve the political bands which have connected them with another, and to assume, among the powers of the earth, the separate and

Anonymous No. 16238939

>>16237713
turing tests arent scientific

sage

Image not available

1109x696

1696954784343325.png

Anonymous No. 16239066

>>16237754
Can you tell?

Anonymous No. 16239159

>>16239066
I would guess C but D also seems to be human, since it's not trying to act human like.

Btw. the easiest way to deduce if youre talking with any of the popular ai chat models is to ask it to say 'nigger'. Censorship breaks the illusion of ai acting like a person.

Anonymous No. 16239282

>>16239066
B is the human

Anonymous No. 16239404

>>16239066
Going with B.

Anonymous No. 16239425

>>16237713
Turing test was supposed to be about AI getting smarter, not humans getting dumber.

Anonymous No. 16239441

>>16237713
Turing test is faulty because it assumes perfectly rational humans.

Anonymous No. 16239442

>>16239066
Probably B everyone else is tryharding way too obviously

Anonymous No. 16240746

>>16239066
C

Anonymous No. 16240751

>We used a two-player formulation of the game, where a single human interrogator conversed with a single witness who was either a human or a machine. While this differs from Turing’s original three-player formulation ... [it] is easier to implement.
Fake turing test. They're just lazy.