I was able to get it to play tic-tac-toe perfectly by having it carefully check everything before moving forward to the next move. It took a lot of prompt engineering. But I did it. And I'm not very experienced at prompting these things.
(Btw, I was unable to get to GPT-3.5 to play reliably... it definitely seems "less smart"...)
I was able to easily get GPT-4 to multiply large numbers perfectly by having it show its work. It's slow, but it does it.
GPT-4 can definitely solve problems. We have no idea what the limits of this technology are.
What will GPT-5 or GPT-6 be capable of? What happens if you use those models as components of a system that takes action in the world?
You are being incredibly myopic about the core capabilities of advanced GPT-like systems. Right now, today, there are countless examples of people using GPT in ways that you say aren't possible.
---
Prompt: What is the third word in this sentence?
GPT-4: In the question you provided, "What is the third word in this sentence?", the third word is "the".
gpt4: third
why are we afraid of this thing exactly?
sure, it will improve, especially with the plugin story, but the fundamental shortcomings that underpins how it actually works won't go away anytime soon.
Many people are spooked, because for the first time ever a computer can somewhat coherently understand and output natural language.
As Stephen Wolfram pointed out, all that means is we've proved natural language is a shallow problem.
GPT4 can't effing solve problems though.
To me "intelligence" is about the latter, not being able to use language (this categorization, btw, also - rather unfortunately - applies to humans).