This is a great little article. I've been calling ChatGPT the "bullshit robot" since it came out; it really does fit Frankfurt's definition remarkably well. Bu

This is a great little article. I've been calling ChatGPT the "bullshit robot" since it came out; it really does fit Frankfurt's definition remarkably well. But that doesn't mean you need to be "anti-AI". I'm also also super excited about LLMs and their potential and am actively working in the field. It's important to understand their role. LLMs aren't the whole brain: they're just the linguistic cortex. That's something computers couldn't do before! It's pretty amazing that we now have a computers that can not only use language, but are pretty good at it! Incredible breakthrough. The problem comes when you let the linguistic brain babble without integration with a more structured reasoning system or cite-able fact database. That's exactly what anyone who is working on retrieval and tools for LLMs is doing, so we're not on the wrong path. But a lot of people still treat this language model as if it were a whole, reasoning brain, and it's really really not. (disclaimer: my comparisons to human mind/brain/intelligence are for analogy only. Real brains and transformer models are very different things.)