Glitch Tokens – Computerphile

Language Models’ Achilles heel: Rob Miles talks about “glitch” tokens, those mysterious words which, which result in gibberish when entered into some large language models.

More from Rob Miles: http://bit.ly/Rob_Miles_YouTube

The AI safety/alignment post: https://www.alignmentforum.org/posts/aPeJE8bSo6rAFoLqg/solidgoldmagikarp-plus-prompt-generation

https://www.facebook.com/computerphile
Tweets by computer_phile

This video was filmed and edited by Sean Riley.

Computer Science at the University of Nottingham: https://bit.ly/nottscomputer

Computerphile is a sister project to Brady Haran’s Numberphile. More at http://www.bradyharan.com