Thread #108268056
HomeIndexCatalogAll ThreadsNew ThreadReply
H
OH MY GOD IT'S SENTIENT AND WILL KILL US ALL
+Showing all 48 replies.
>>
>guesses what pornstar I watch
>never calls the FBi

truly a bro
>>
salami aleikumi!
>>
>>108268082
??
>>
>>108268092
i don't get it :DDDDDD
>>
>>108268082
>>
>>108268414
when you get it, it will be too late! hehehe
>>
I was actually creeped out by Akinator when I used it years ago. It didn't make sense to me how quickly it was able to guess an obscure character based on very generic information.
>>
>>108268643
all of the training is made by user input, so if one autist makes a bunch of specific answers, does not find the character, and makes the character after, ackinator will remember the questions used to get there, and if you're autistic enough to think of a rare character, you will probably guess similarly to the original creator even though only one person was involved in creating the character's profile.

ackinator is actually far LESS accurate when guessing well-known characters or species/organizations because there are usually multiple profiles for the same thing, and conflicting answers.

the real difficulty in ackinator though is getting it to give you the right string of questions, because by 30 questions in it WILL give you an answer, no matter how many times you've said "no" or "not applicable/idk" and i've noticed that the first 8 questions are so is whether or not the character is from roblox, fortnite, FNAF, or stuff like that.

ackinator is not advanced at all, it just benefits from focused, mostly accurate testing from small groups. LLMs are more adaptive but they hallucinate far more often because they have a practically uncountable set of training data and are constantly being misled by incorrect data or user input.

if AI-bros tested their LLMs entirely on accurate data, rather than any shit they scraped from social media and getty images, it would eclypse ackinator entirely
>>
>>108268682
I see. That makes sense.
>>
>>108268056
A prime example of real artificial intelligence.
>>
I beat him using Puffy (the OpenBSD mascot)
>>
>>108268538
>>
>>108268082
Akinator is a pretty cool guy and doesn't afraid of anything.
>>
>>108268056
how does this fucking thing work? even back in the early 2000s it was stupidly good. Is it just one big nested if function?

>>108268682
I don't buy this
>>
>>108269271
>how does this fucking thing work?
Not very well, but I imagine if you aren't very knowledgeable or clever it would have a pretty high accuracy rate.
>>
>>108269271
>how does this fucking thing work?
I think I remember reading about this. It's some sort of decision tree. It has a slew of questions and each answer is saved with a list of responses. It asks the questions with most responses first. Whenever it cannot find what you had in mind, it creates a new record with all responses.
tl;dr decison tree and db of previous responses
>>
There is something about natural language that tickles the normie brain.
>>
>>108268056
Now that we're at it, think of any obscure character and have Akinator guess it
Post the results
>>
File: 1.png (670.4 KB)
670.4 KB
670.4 KB PNG
>>108270143
He never manages to get r34 characters 1st try, guess it's a filter or w/e because he only asks about sex on 2nd try.
>>
File: s.png (617.3 KB)
617.3 KB
617.3 KB PNG
>>108270143
i'm blown away once again
>>108270197
yeah maybe, at least in the second try he usually hits it
>>
>>108270143
I tried to get Emily of Emily's Babying 2 and while it got me a bunch of furry characters I'd never heard of it couldn't get it before running into an error and kicking me out
>>
>>108270143
he thought kuriboh was an anime bird at 25 questions but got it right after 5 more questions
>>
File: 1.png (409.1 KB)
409.1 KB
409.1 KB PNG
>>108270197
This one took 8 tries.
>>
>>108270143
after 50 questions he guessed a candy land character despite being told my character is from magic the gathering
>>
More like AGInator, and I right, gamers?
>>
AK--
>>
>>108270143
>>
>>108270143
Heh, I win
>>
>>108271178
Which character was it?
>>
>>108271209
Lukyon
>>
>>108270143
failed to guess sugaar of basque mytholohy
>>
File: saya.png (103.7 KB)
103.7 KB
103.7 KB PNG
>>108270143
couldn't get saya unfortunately
>>108271230
very cute
>>
>>108268056
50 questions and 4 wrong attempts later, this simulacrum of intelligence still can't figure out my character. I bet I can keep going 'till the AI bubble bursts and it still won't name the right guy.
>>
File: file.png (130.6 KB)
130.6 KB
130.6 KB PNG
>>108268056
mmh
>>
>>108270143
I tried it with BB from Anthology of the Killer. Turns out he didn't know her, which honestly surprised me a little.
>>
File: .png (439.2 KB)
439.2 KB
439.2 KB PNG
why is it retarded
>>
>>108271487
>why is it retarded
It learns from the people playing it
>>
are you for real
>>
>>108269271
>how does this fucking thing work? even back in the early 2000s it was stupidly good. Is it just one big nested if function?
You ever play 20 questions before? it's just that
>>
File: ruka.png (620.1 KB)
620.1 KB
620.1 KB PNG
>>108270143
second attempt
>>
File: 786576.jpg (70.3 KB)
70.3 KB
70.3 KB JPG
>>108268056
MFer finds even best girl.
>>
>>108270143
Didn't get the guy at the start of BLAME! that got blasted in half
>>
File: Akinator.png (373.1 KB)
373.1 KB
373.1 KB PNG
>>108270143
Didn't guess Manhack Matt from HL2
It was an existing option when I was asked to write it in though
>>
>>108268056
Oh damn im gonna play 20 questions with chat gpt
>>
>>108270143
>>
He couldn't find Sylvie
>>
>>108268538
the joke is CP why aren't you laughing bro. lil homie, it's a joke about child porn don't you geddit
>>
>>108271178
akinator couldn't guess randi from secret of mana. not even in the list. guess he was rather too generic

Reply to Thread #108268056


Supported: JPG, PNG, GIF, WebP, WebM, MP4, MP3 (max 4MB)