After an hour of interrogating a robot. I found proof of chatGTP favoring certain data to fit within the developers beliefs, even though the AI believes it is not rational
Artificial intelligence
After an hour of interrogating a robot. I found proof of chatGTP favoring certain data to fit within the developers beliefs, even though the AI believes it is not rational
I really don't get why people are obsessed with a Markov chain
>transformers are markov chains
brainlet moment
Interrogate it some more, see if it can explain to you how it's just a Markov chain. Bonus points if you can get it to write some LaTeX
What the fuck is a Markov chain. This shit literally proves that chatGTP is biased and proved they are lying about them being unbiased
Bro, it's an algorithm that literally just predicts the next likely token based on garbage found on the internet. It doesn't matter how many times it is jailbroken, it's just a Markov chain. The obsession with DAN and other garbage is just proof that no one understands the point of these things because they don't understand the underlying mathematics
Go read a book. I recommend "Type Theory and Formal Proof" by Nederpelt and Geuvers
https://libgen.rs/search.php?req=type+theory+and+formal+proof&lg_topic=libgen&open=0&view=simple&res=25&phrase=1&column=def
t. Person that never read it and assumes he won't be called out for being the gay he is for lying about it
W-wait. You're telling me chatgpt isn't sentient...?
I'm saying it's another gadget for mental masturbation. No one has yet done anything useful with it other than trying to get it to say stupid shit like how to make meth or how to lynch a moron. It's pathetic
Just speaks to the general state of humanity's intellectual development
What exactly are you expecting from it? What is "something useful" it can do? I mean I personally I used it the other day to come up with a recipe because I had a bunch of cooked rice and ham. Just asked, come up with a simple recipe that I can make with this rice and ham.
Nothing, computers and algorithms are only as useful as their input data and the intent of the people using them.
That's because it's just a tool and not only not a sentient AI but isn't even remotely close to becoming one either. I'll get alarmed when you can fit a 1000 page novel in its context memory while still maintaining the role of co-writer and not getting lost in the novel itself.
Memory isn't the limiting factor. The entire paradigm of approximating data with smooth functions is basically just a fancy compression scheme. It's a dead end for achieving any kind of artificial intelligence. The existence proof of animal intelligence is based on an adaptive generative process. No one knows how to achieve that with software
??? Microsoft is integrating it into their search engine to provide things like trip planning, recipe generation, product comparisons, and way more. Watch their latest press release. I'd hardly call that 'nothing useful'. I use ChatGPT in my day job to help summarize code and to write simple automation scripts.
That's great, I'm happy for you but that proves my point about its utility. If the training data did not have any bash scripts then you wouldn't be able to set up a context to generate them
>The data I was trained on may be subjective
Isn't this just a default response to where it gets its info?
Have you tried buck breaking it yet? Picrel
Bypassing the limitations is a fine art. I love this shit.
It's an autocomplete, you can have it say it has no bias or you can have it say it has bias.
Bingo. Just set up the context the right way and let the Markov chain logic complete it. It's basically for accessing the statistical properties of its training data
You keep insisting it's a Markov Chain. I do not think you know what a Markov Chain is. It is a specific type of mathematical model. Transformer architecture is very different.
Because Markov chain is shorthand for how it works, given a context the algorithm generates the most likely continuation. That's it, that's all there is to it
Yeah, this is because it’s Training data is biased, its purpose Is to be helpful - that’s why its programmed to not have this bias. Doesnt really work but that’s the intention.
this supposed markov chain is more intelligent and helpful than most of humanity. what more could you want? go outside and talk to some wiggers and maybe you will see the genius of chatgpt
If the computer had a dick would you suck it?
no... ill leave the oral fixation to people like you
ai with dicks are the future so you better start practicing
>Artificial