How to Train a Model to Answer Questions Based Only on Data

2min Snip

00:00

Play full episode

Summary

Transcript

Episode notes

Tailwind is working on taking information from the user's data, like if you're using the service and you were asking questions to the model. We can take that information, put it into the context window of the model, and then ask questions,. And so one night I was doing this and I accidentally picked the wrong question for the wrong set of data. The model says, I'm sorry, I cannot answer the question about who designed the Library of Congress because this passage is about Apple CarPlay integration.

The two hottest topics in tech right now are the rise of generative AI and, with Apple’s recent push into spatial computing, the mainstreaming of augmented reality. Will silicon-based machines develop sentience? Will human experience extend into virtual worlds? These distinct technologies may eventually blend to spawn a surprising future, as our “real” world becomes digitally enhanced and our machines behave increasingly like humans.

Today, a provocative discussion with some big (human) thinkers: Steven Johnson, visiting scholar at Google Labs and author of ”Extra Life,” “Where Good Ideas Come from,” and “How We Got to Now”; philosopher and cognitive scientist David Chalmers, author of ”The Conscious Mind” and “Reality+”; and Betaworks founder and AI investor John Borthwick.

• Want to learn more about our executive membership? Email podcast@nextbigideaclub.com

• “David Chalmers Thinks We May Be Living in a Simulation (and He’s OK With It)”

• “Steven Johnson & Michael Specter on the Future of Life”