Member-only story
Major Challenges in Today’s Large Language Models
Is It Just Me, or Are AI Models Actually Stupid?
5 Issues I’m Facing with ChatGPT and Other LLMs
AI made me think of a future in which a Terminator T-800 roams around trying to find John Conor to stop him from preventing Judgment Day.
Boy, was I wrong!
At least, as of now, in 2025, I would say that future is unlikely to happen in the next decade, even if AGI becomes a thing.
That’s my speculation, and I’d love to be proven wrong.
This article is about the current state of large language models being stupid.
Yes, these so-called AI products (ChatGPT, DeepSeek, Mistral, etc) claiming to be revolutionary are, in fact, stupid.
I want to clarify what I mean by using such terminology.
Being stupid doesn’t mean they lack intelligence or practicality. On the contrary, they are more intelligent than I could imagine.
The problem is how badly they’ve been under-designed while over-prioritising developer work.
It seems like developers wanted to make LLMs a thing more than designers do, which produced products unsuited to use by humans.
