How to (not) spot LLM-generated text
2024-09-24 | tags: thoughts

Your periodic reminder: there is no sure-fire, clearly-visible indicator that text came out of a genAI bot (LLM).

Similar to Becca Chambers ✨ , most of the items in the screen cap align with how I -- and many people I know -- write! Plenty of people use Oxford commas, em dashes, and the word "delve" in their everyday writing. Plenty of people use Oxford commas, em dashes, and the word "delve" in their everyday writing.

Spotting generated images is a bit easier because we can see extra appendages or squiggly lines in lieu of words. You don't get that with generated text.

(To be fair, I understand why people want those easy tells for generated text. But we're not there yet: https://bsky.app/profile/qethanm.bsky.social/post/3l33q5ruze22j )

Complex Machinery 017: Stacking the deck

The latest issue of Complex Machinery: Machines conquered Wall Street, then learned to play a mean game of poker.

Complex Machinery 018: Robot jobs

The latest issue of Complex Machinery: Sometimes AI is a better fit for that task. Emphasis on "sometimes"