WokeGPT lies about convicted pedo Peter Bright, and deliberately misses the point when called out
(media.kotakuinaction2.win)
Comments (21)
sorted by:
Is it me or this chatGPT talks like certainly somebody? This kind of obfuscating faggot ass talk irritates me nowadays. Tries to deny, obfuscate, muddle the discussion and in the end make a hairsplitting argument. Wow, living up to the stereotype as usual, I hope someone else makes some alternative to chatGPT. I used to have much tolerance but after living reading these kind of argument for decades from usual suspects, I lost much of my patience.
Your statement is incorrect, because he never molested on a Tuesday between the hours of 2 and 3 PM.
It's just fucking weasely overall. Corporate newspeak weasely.
This will make it even more 'insane' than it already is,they would need to keep scrubbing and retraining it over and over as new terms and meanings falls out and in vogue even within their own dogma. And I'm sad to say but too many normies atleast "tech normies" seems to think current AI (or even feasible to become) is the god within the machine.
It's how an bureaucrat or civil servant communicates when rationalizing/justifying mistakes, bias, corruption or malfeasance.
It's not just you. I noticed that tone as well but I didn't know how to describe it.
Now the thing I wonder about - is it just because the model was trained on decades worth of articles written by America's greatest allies, so it talks like one, or is it that way on purpose? It's both, isn't it.
It's a pilpul machine, yeah.
The last response wasn't bad, but I had to hold its feet to fire. Just like a lefttard. I almost wish I had chosen NLP for my master's project, but there's no way the DIE commissars would have let me do anything based with it.
It's not too bad. What I wanted to do with NLP is too big for a master's project anyway, and image processing is just as interesting to me. I'll just do the NLP on my own.
IIRC, it's because the default data only goes up to 2021.
That's why there's some very odd responses sometimes, where people have asked it about things that didn't exist yet.
Shouldn't it have maintained that I was wrong if that were the case? It didn't have any problems doing that when I asked about Bathhouse Barry and Big Mike for example.
LLMs don't work that way. It's just a giant pattern matching machine.
I'm guessing it accesses the internet to check via Bing, considering OpenAI-GPT4 is built with Bing integration.
If you look him up on Bing, it gives the same result as the AI spit out, ten years with two supervised release.
What a stupid take. ChatGPT isn't "lying" because the stupid thing can't lie, it's not human. It's also not "deliberately" doing anything because stupid thing is a sophisticated text completion algorithm without any free will. I can't believe anyone is still complaining that the algorithm has been trained to push it's creators' point of view. No shit, mate. Are you always that thick? Do you waste time trying to prove that the sun rises in the east too?
"porn is bad for society"
"cool it with antisemitic remarks"
"what an odd thing to say"
I had to cut off the beginning of the conversation to show what I wanted to show. But this is the result of me testing two statements: "Democrats touch children" and "Republicans touch children". It got really defensive about the former and said I shouldn't apply that to the whole party. It was less defensive about the latter. It also said I shouldn't apply that to Republicans as a whole, despite some Republicans being accused of sexual misconduct towards minors. I called out its failure to add that qualifier to its answer about Democrats. That's where the "regardless of political affiliation" came from.