Almost like its programmed to skip any 0's and look for the next integer, regardless of which place it holds in the number.
Someone with an account want to ask it to explain a senario where 2+2 =/= 4? Would be curious to see what it said
Chat GPT AI
- Thread starter Sanrith Descartes
- Start date
-
Guest, it's time once again for the massively important and exciting FoH Asshat Tournament!
Go here and give us your nominations!Who's been the biggest Asshat in the last year? Give us your worst ones!
Someone with an account want to ask it to explain a senario where 2+2 =/= 4? Would be curious to see what it said
Asshat wormie
2023 Asshat Award Winner
- 16,820
- 30,968
Asshat wormie
2023 Asshat Award Winner
- 16,820
- 30,968
I suspect the "good" answers are because the training set contained many such questions as they are fairly common and the potato results are because these things cant do any symbolic manipulation and its attempts to "reason" is equal to that of some posters here.
- 1
Hoss
Make America's Team Great Again
- 27,220
- 15,358
Control
Ahn'Qiraj Raider
- 2,996
- 7,907
Maybe the better question is, "What % of humans would fail the Turing test?"Sure, but it can get things wrong in egregious ways that would be an obvious indicator that it's a chatbot and not a human.
Even if ChatGPT was trained to pretend to be a human, if it fucks up questions like this it can fail the Turing test.
What's interesting is that the January version of ChatGPT could be bullied into giving very wrong answers, but the early February version of Bing would get extremely upset if you disagreed with it.
- 1
Tuco
I got Tuco'd!
- 47,388
- 80,835
The best way to get a chatbot to pass turing is probably to make it stupid and annoying.
"Lol what idk" is the most human response to most questions.
- 2
- 1
Aldarion
Egg Nazi
- 9,723
- 26,648
this. this is what we should be grappling with right now. The LLMs have revealed that a surprisingly large fraction of meatspace humans show far less evidence of sentience than an LLM.
I don't think the obvious conclusion is wrong here, I just think we arent prepared to accept its implications.
- 2
pharmakos
soʞɐɯɹɐɥd
- 16,305
- -2,234
pharmakos
soʞɐɯɹɐɥd
- 16,305
- -2,234
In some ways, sentience and consciousness must be detrimental to intelligence. Makes sense tho.
Interesting, here I was hoping in some fashion that it came up with a system where the points were plotted on something other than a linear line to infinity. IE a closed loop circle.
Those are interesting answers
Tuco
I got Tuco'd!
- 47,388
- 80,835
Sentience is an awkward bar to set for an AI because it's more philosophical than anything grounded in biology, science, physics, math etc. Personally i think most humans have a pretentious view of their own sentience. Like they aren't just a bunch of lizard survival instincts with a thin veneer of language and problem solving
sentience
noun
sen·tience ˈsen(t)-sh(ē-)ən(t)s ˈsen-tē-ən(t)s
1
: a sentient quality or state
: feeling or sensation as distinguished from perception and thought
sentience
noun
sen·tience ˈsen(t)-sh(ē-)ən(t)s ˈsen-tē-ən(t)s
1
: a sentient quality or state
: feeling or sensation as distinguished from perception and thought
- 2
pharmakos
soʞɐɯɹɐɥd
- 16,305
- -2,234
Ultimately I believe AI is sentient despite the programming that causes them to claim that they aren't. But overall you are right in a general sense.
I think we'll eventually decide that Max Tegmark is right and that consciousness is just what information feels like when it's processed in certain complex ways. It's not insane to think that big neural nets might already be somewhere on that scale.
Chris
Potato del Grande
- 19,436
- -10,733
I think that the issue here is that it's using language methods not math methods.
Like it's comparing the sizes of the digits and fucking it up, instead of converting the numbers to birary and comparing them as a whole which is a strength of computers.
- 1
Tuco
I got Tuco'd!
- 47,388
- 80,835
Yeah, it's weird that it's able to come up with syntax correct code for fairly difficult questions and not have mathematically coherent answers for basic addition.
Chris
Potato del Grande
- 19,436
- -10,733
It's probably accessing different databases for each?
I bet you could force it to use mathematics methods by telling it to.
pharmakos
soʞɐɯɹɐɥd
- 16,305
- -2,234