Knowing AI Limits

Published: 2023-05-23 10:38 |

Category: Technology | Tags: teaching, AI, LLM, productivity

I've spent the last couple of weeks taking deep dives into LLMs and how they work. Stephen Wolfram's essay (more of a mini-book) took me several days to read (and I still don't understand all of it) and other suggested reading by Dave Cormier and Ted Chiang were also very helpful in building a picture of what can actually be done with tools as they stand.

Turns out - not much.

The most interesting piece of the AI boom is the natural language processing that is possible. The way the models understand prompts is very, very cool even if the output is stale. The point is that knowing more about AI and how it works today has put a wet blanket on wanting to actually use it.

I think understanding context is critical for putting technology to use in meaningful ways. I want to be able to use assistants some day (privacy issues and data retention are another ball of worms) and I think the progression of AI to be able to do things dyncamically has huge implications. I would love to be able to say, "Schedule a meeting with Jim, Carrie, and Bill the next time we're all free."

I think after all of my reading and digesting, the potential for AI is there, but it's not something I'm going to put a whole lot of time or effort into using yet.

If I'm wrong, I'd love to get some insight. Send an email with what I should consider.

Comments are always open. You can get in touch by sending me an email at