Jiggawatts’ second rule: “Unless an opinion on LLM technology includes the specific phrase ‘GPT 4’, it can be dismissed.” The author tried older, thoroughly outdated models, and has decided to publicly state an opinion without bothering to check if it’s still valid or not. Ironically confirming that humans are just as susceptible to writing false statements as Chat AIs. Remember boys and girls: self driving cars don’…
Does, in your personal opinion, GPT-4 has an underlying model of the world?
No. GPT is a model of human writing. But, as the Box quote goes, "Essentially, all models are wrong, but some are useful.". It isn't writing the same way that we are, with the same thoughts or mental models. It's just amazingly good at imitating it. For tasks that can be achieved by just writing, the GPT model is so good at modeling writing that it performs as well as a normal person using their writing skills plus their mental model of the world. But GPT won't look up information unless asked to. It won't try something new to see if it works.
It's this distinction useful? Rarely. But it's one of those things users should remember, like leaks in an abstraction. When it doesn't do what you expected, you should know these gaps exist in the model.