The Challenges of Agent Based Automation Models

The challenge with that is that these models will hallucinate. And I think they will be hallucinating for a long period of time. Fundamentally, when transformer models do this, they predict the next token essentially in a sequence of text and then they do something with that. So what we say we use the models for build time, which is getting you from your intent to the actual automation script. Then once I have the automation script, it's a normal program. You can think about it like a program on your computer that it will always do the exact same thing every single time.

Transcript

Play full episode

The AI-powered Podcast Player

Save insights by tapping your headphones, chat with episodes, discover the best highlights - and more!

Get the app