We might be arguing past each other. I don't know any (serious, at least in my mind) AI researchers that would make that claim, no. But it is a distinction tha

We might be arguing past each other. I don't know any (serious, at least in my mind) AI researchers that would make that claim, no. But it is a distinction that is poorly communicated to ordinary people who use these models, and most of the time that you encounter "how do we know that humans reason" in the wild in non-researcher contexts, it usually is being used to argue that we just need to throw more processing power at models and get more data into them -- or to paper over the real criticism that the current generative models that are available to ordinary people are not useful for tasks that require something approximating "actual reasoning." It's a hard thing to talk about, because I recognize that it is absolutely unfair to criticize AI researchers that are not making these arguments in the first place, but also recognize that pointing out this kind of distinction is absolutely necessary for the general public on a public forum and that it is important to continue to hammer home the practical difference between pattern-matching and reasoning in current models. I don't know how to walk that line other than to point out that trynewideas's comment was reasonable and that attacking it from a philosophical perspective misses the point of what they're saying. But I don't mean to imply that researchers are wrong or misinformed about the state of AI, I'm just pointing out that the philosophical distinction doesn't really change anything about the practical argument (since non-researchers also frequent HN).