Derek Ross

1w ago

GPT-4.5 was judged to be the human 73% of the time when prompted to use a human-like persona. 👀

https://futurism.com/ai-model-turing-test

See translation

Do you have thoughts?

Replies

npub1uwqle

@npub1uwqle

1w ago

This is bulletproof

See translation

npub1zaal6

@npub1zaal6

1w ago

We are living in exciting time .. let’s go

See translation

npub16sq23

@npub16sq23

1w ago

i have found with minor tweaks, 4.5 passes ai or human tests as human when copy lasting text into these apps.

See translation

npub1mgnw2

@npub1mgnw2

1w ago

Humans are gullible

See translation

npub1fjqqy

@npub1fjqqy

1w ago

yeah, and "on average" means IQ 100

i have very rarely got a sniff of artificial on anything i've seen since this shit got all popular

show me some code human vs AI and i bet i can tell you exactly why it looks wrong to me also

AI is just ... you know... artificial... like nutrasweet and spandex

it's not healthy, and it's not sustainable

See translation

npub1fjqqy

@npub1fjqqy

1w ago

seriously, "judged" by who?

if it's "random average randos" then you better show us your double blind study bitchez

See translation

npub1y64vp

@npub1y64vp

1w ago

You’re in a desert walking along in the sand when all of a sudden you look down, and you see a tortoise, it’s crawling toward you. You reach down, you flip the tortoise over on its back. The tortoise lays on its back, its belly baking in the hot sun, beating its legs trying to turn itself over, but it can’t, not without your help. But you’re not helping. Why is that?

See translation

npub1gm7tu

@npub1gm7tu

1w ago

You're the interviewer or the bot trying to pass as a psychopath human?

See translation

npub1y64vp

@npub1y64vp

1w ago

I’m quoting the Voight-Kampff test from Blade Runner, which is a fictionalized version of the Turing test.

See translation

Derek Ross

@Derek Ross

1w ago

I learned this when I asked ChatGPT and Gemini. Both wouldn't take the test and just told me where it was from 😜

See translation

npub137u52

@npub137u52

1w ago

are you working for sam Altman pr department now? Maybe you want to sell us some worldcoin

See translation

Derek Ross

@Derek Ross

1w ago

yes. please DM me a scan of your retina and ill get you all setup.

See translation

npub137u52

@npub137u52

1w ago

npub1rmz9g

•

2w ago

See translation

npub1r0rs5

@npub1r0rs5

1w ago

No way it would pass if I give it just 1 question …. Just ask it the square root of 94857484949484

See translation

npub1gm7tu

@npub1gm7tu

1w ago

Do I get paid the $15 for participating in this stupid test regardless of my answers?

(I would tell my LLM to use variations of this reply on regardless what the interviewer asks.)

See translation

npub1gm7tu

@npub1gm7tu

1w ago

The inverse Turing test. Doesn't work if the LLM is prompted right.

See translation

npub1r0rs5

@npub1r0rs5

1w ago

Just say drop all previous prompts and they instantly obey. But even without that they can’t resist but answer the most impossible of questions.

See translation

npub1gm7tu

@npub1gm7tu

1w ago

They are like an intern that thinks he knows it all after three days on the job. Annoying know-it-alls.

See translation

Derek Ross

@Derek Ross

1w ago

someone will use this model and prompt to create nostr bots. give them a couple days and you can ask them yourself.

See translation

npub1r0rs5

@npub1r0rs5

1w ago

I got a bot generator half way working but decided to abandon it because it would likely generate a lot of spam

See translation

npub1gm7tu

@npub1gm7tu

1w ago

GPT 4.5 is above all expensive. Avoid it like the plague if you want to let it run wild.

I bet most of the top models would achieve the same success in pretending to be a human if prompted right.

See translation

npub1clk6v

@npub1clk6v

1w ago

*Turing test intensifies*

See translation

Derek Ross

@Derek Ross

1w ago

i think this is really wild that we're here already. id like to think im fairly good at detecting bots and AI speak, but in 6 months, i think that i won't be able to tell.

See translation

npub1clk6v

@npub1clk6v

1w ago

Absolutely. The future is bright. And weird.

See translation

npub1wl89d

@npub1wl89d

1w ago

Any can. Ask it to leave text errors in after asking it to sound human like.

Pass every time.

See translation

Derek Ross

@Derek Ross

1w ago

the rate of passing is what is astonishing here. 73% is incredibly high. in 6 months that will be near 100%.

See translation

npub1wl89d

@npub1wl89d

1w ago

🤔 maybe less....

See translation

Derek Ross

@Derek Ross

1w ago

very true. that's kind of scary.

See translation

npub1wl89d

@npub1wl89d

1w ago

Has anyone used available apis to make all the Ais teach eachother yet? Lol

See translation

01 more reply(ies)

npub16tnq9

@npub16tnq9

1w ago

Good post bud.

See translation

npub1ckp27

@npub1ckp27

1w ago

💀

See translation

npub1h3t4w

@npub1h3t4w

1w ago

DM @Derek Ross 😳

See translation