Comment by charles_f

I find interesting the reaction of a lot of people to that paper, calling it out of touch, and bringing up that ChatGPT is super useful. I don't think such claims are made, rather Chomsky Robert and Watumul open with:

> These programs have been hailed as the first glimmers on the horizon of artificial general intelligence [...] that day may come, but its dawn is not yet breaking, contrary to what can be read in hyperbolic headlines and reckoned by injudicious investments.

The article is not claiming a lack of usefulness, but the fact that this is not a human-like intelligence as it's been claimed "in hyperbolic headlines".

What I get from it is that while the technology is suggesting a lot of enthusiasm, it remains a conversational tool rather than actual AI, and exhibits the limitations that come with it. It is in fact akin to a more advanced search engine, working probabilistically, mimicking what a conversation on a topic looks like. It is incapable of building a system of beliefs, of morality, or critical thinking. It is not really inventive but rather plagiarist. It cannot infer or deduce. It doesn't "learn" the same way as humans do.