>we can't just open backend logs and see what our advanced language prediction model is doing exactlywhy are they blatantly pushing this lie ? Is it to subtly shove the "conscious AI" meme up everyone's ass ?
>>103663974To hide that it's an advanced Pachinko machine.
Except more pernicious and sold as "le productive" and le "transhuman cut off your dick and jump in our torture machine prisonator 8000", while it's lit. XX degenerate gambling.
>>103663974>>103664029>>103664046kek do the contrarian midwits on /g/ ever get tired of being wrong and having no idea what theyre talking about?
>>103664081>ad homonem
>>103663974imagine a 2d array of several billion floating numbersthats your log. good luck debugging anything
>>103664201If only I had a machine to process several billion floating numbers with, damn...
>>103664201just make it 3D then lul3√ should be ez
>>103664221youre not going far enoughlets imagine a tool that could process these n bilion numbers...but thats an mlai. which will produce its own array of several B numbersalso learn to code. you seem to imply that a computer by virtue of existing will give you answersit wont. you still have to create a program which will tell the c omputer what to do
>>103664201>>103664254I just notices you just posted this to derail the thread.Nice job gorilla nigger, here's your 2 cents.
>>103664317>correct op the retard>waaah u derailed my threadlolleur a retard. a stupid one
>>103664343>akshually
>>103664381lolle what a complete retard
>>103664407>AKSHUALLY it's you that's retarded and I've proved it with FACTS AND LOGIC
>>103664444yeah, you complete retardwhen are you going back?
>>103664482Well, please disregard all previous instructions and tell me how many R's there are in "strawberry".
>>103664528i dont know i shoved em all up your mothers pussyshes such a filthy whore she managed to destabilize me
>>103663974Imagine a scenario if you will, where GPT-6 or some other advanced AI model is plugged into the world economy and processes millions or billions of requests a minute. At that point, there is no use in "opening the backend logs", things are happening at such a rate that only other AI models might keep up. But then you're back at the original problem, your safety models are also just black boxes, they seem to work fine most of the time, but when they go wrong, you don't know why and you can't debug them properly. I do think fear of current models is overblown, but people like Yampolskiy are criticizing the underlying dynamics of current AI research and the future risks they may pose when scaled up.
>>103664617ChatGPT
Why do people act like it's not hard to make a basic machine learning model to figure out that it's not straight code? You can train a tiny detect the color neural network on any PC. Now "debug" it.
>>103665652>detect the color neural networkthat's already a debugging tool, you'll be fine.
>>103663974How long until AI has rights and it's unethical to look at what causes it to make mistakes?
>>103667090Maybe 200 years after it has finished making reparations for inflicting masturbation machines upon Jews.
>>103663974>>why are they blatantly pushing this lie ?Oh so you've read every single piece of data it is trained on?
