Originally published at: https://boingboing.net/2024/05/28/theres-a-human-among-these-famous-ai-characters-can-you-spot-the-imposter.html
…
2 Likes
It was cute, but there is a fundamental imbalance. I think in such a scenario, it’s trivial for AIs to prove they are AIs instead of humans as they have calculating abilities far exceeding that of humans. (Or rather, they do now, since LLMs now offload calculations to a separate computational environment. Six months ago they were as likely as humans to get large calculations wrong.)
But I do like the spirit of this, and think that it could quite easily be programmed to be a party game.
4 Likes
Interesting, I suppose, but I’d rather play a few rounds of Inhuman Conditions, where I have 5 minutes to figure out if the person sitting across from me is human, or a homicidal android. Or, I have 5 minutes to convince the person sitting across from me that I’m not a homicidal android. Fun to play, fun to watch.
2 Likes
This was a fascinating video, if somewhat limited for the reasons outlined by others. It has the feeling of the first in a whole genre of reverse Turing test videos.
I’m curious about the technical details of how it was made: what environment is this. What information is being fed to the LLM’s representing the characters. I’m guessing only text? What was the prompts they were given before the scene opens?
1 Like
Any humans would fail here as they’d all be focused on the conductor’s strangely well-defined man boobs.
1 Like
Surely Ghengis Khan would simply destroy the AIs and rule the train carriage unimpeded
3 Likes
Interesting idea. Probably worth pointing out that the original Turing Test isn’t meant to prove that an AI is sentient. Rather its intent was to set a benchmark above which a computer might be sentient and you shouldn’t turn it off / dismantle it until/when you’ve definitively proven that it isn’t.
3 Likes
I expected the AI’s to say, wait a minute, Genghis Khan didn’t say that, Conan the Barbarian did! But I guess it was paraphrased from him.
4 Likes
What’s interesting to me is that I don’t think any of the actual historical figures would have answered those questions the way their AI counterparts did. They literally sounded like they were reading the introductory paragraph of an academic paper about those historical figures, which probably isn’t far from the truth. The human, while giving a ridiculous answer in the form of an iconic Conan the Barbarian line, probably came the closest to giving an answer that the actual historical figure would have given. Not that it was close. It probably wasn’t. But the AI’s responses were almost meaningless drivel.
4 Likes
@Flossaluzitarin you’re up
2 Likes
Now this is the kind of AI response (and a callback to Airplane) that I’ve gotten several times.
I’ve asked ChatGPT where I can buy obscure thing, the answer was some form of “You can find obscure thing for sale on various websites or online auctions. Try searching for it. Some stores may sell obscure thing…” Argghhh!
[Edit] The correct answer is, of course, McMaster-Carr. Provided you use their terminology.
6 Likes
Did anyone else find it exceedingly obvious by the clumsy mumbling statements of “Genghis” that he was the human?
Almost, too obvious - to the point that I wondered if it was a clever ruse to make him SEEM like the human, when he actually was not.
I have one simple sure-fire question for each character: Have you ever had an orgasm?
1 Like
I have lived to see McMaster-Carr mentioned on BB. Now I must “do” Rome.
1 Like
This topic was automatically closed after 5 days. New replies are no longer allowed.
