It does not work. At most it looks like it works.
A human brain is able to understand and process information. An AI simply calculates a mathematical function. There is no reasoning and no understanding of anything, all that ChatGPT does is try to look like a human. And by “try to look like a human”, I mean “generate sentences that can be believable, on shape, to be written by a human”.
If you ask it to calculate 2+2, and then tell it that it is equal to 5, it won’t see any problem because it doesn’t understand any of it. But it will give you answers that are, grammatically speaking, reasonably human.
If I ask a rock how much is 2+2 and I throw it, and it bounces 4 times, it does not mean that this rock knows how to count. ChatGPT and similar are just better illusions, but they’re nothing more.
SmoothIsFast@citizensgaming.com 1 year ago
No it does work, it’s just most people don’t understand its just a very complex lookup table essentially using a predictive model. It doesn’t think, nor feel, nor imagine, it runs a function to choose the next fitting word based on previous inputs and it’s training set. In that regard it is damn good.
Solumbran@lemmy.world 1 year ago
Yes but that’s not what people think. They use it as a search engine, a programmer and a teacher. This is my problem, they use it as something it absolutely isn’t and base their opinions, knowledge and work on that.
SmoothIsFast@citizensgaming.com 1 year ago
The only issue with it is needing to teach people to fact check back with 1st party sources. Teachers are not always absolute, documentation is not always absolute, and search engines are certainly not always absolute especially with SEO. Fact checking is always a necessity whether using generative AI solutions or not. You need to take a step back, this improves efficiency when used correctly and you won’t teach proper use cases if not brought up early with students in school. You are arguing against yourself here.