Great article explaining how LLMs works under the hood, especially for non-technical folks not getting (understandably) that it's largely a pre-trained, stateless "blob of floating point numbers" (as you note) and not a human.
I find it's really hard for people who haven't worked on this stuff themselves to get. I suppose there's still something to the Turing test (silly as it seems nowadays)—people aren't really able to conceptualize that something that sounds like a human doesn't actually work like one at all.
Loved your piece about mental models and common misconceptions about how AI “learns”. It notice the confusion, too. And it’s understandble, this stuff is mighty complex. But having a base level understanding really improves people’s ability to grasp what AI can and cannot do.
Great article explaining how LLMs works under the hood, especially for non-technical folks not getting (understandably) that it's largely a pre-trained, stateless "blob of floating point numbers" (as you note) and not a human.
I find it's really hard for people who haven't worked on this stuff themselves to get. I suppose there's still something to the Turing test (silly as it seems nowadays)—people aren't really able to conceptualize that something that sounds like a human doesn't actually work like one at all.
Loved your piece about mental models and common misconceptions about how AI “learns”. It notice the confusion, too. And it’s understandble, this stuff is mighty complex. But having a base level understanding really improves people’s ability to grasp what AI can and cannot do.