krion

I have to admit I survived a lot of hypes in IT industry for the last 30 years: Y2k, dot-com, Grid, Cloud, VR, Blockchain, BigData etc. Now we are inside of AI era with a lot of promises which AI can’t keep and which is just an illusion for normal user. Unfortunately all the hype tastes the same.

What people really need to understand is that LLMs are not capable of independent thought. They rely entirely on stochastic guessing, which isn't inherently bad – it has its use-cases. However, what frustrates me is the attempt to sell AI as a complete labor replacement, causing significant collateral damage both directly and indirectly. The truth is, AI tools often aren't robust enough to handle many jobs alone, and legal risks necessitate human oversight. But who cares, AI is easy to sell, you can always dupe management with “cutting costs” arguments.

There's a fundamental misunderstanding about ChatGPT and LLMs. They aren't truth oracles, filtering reality for perfect answers. That's a myth. LLMs don't possess genuine thought or reasoning, they're simply distillation of knowledge. They churn through mountains of human-generated data, spitting out answers that often appear convincing but lack the crucial element of verification. In essence, LLMs are adept at creating illusions of knowledge, not knowledge itself.