Do you use LLMs? And some podcast recommendations

Right now, I am using LLMs to (a few examples): Translation Given the transcript and/or script of an episode of my podcasts, generate a summary (show notes) for the episode Generate fluffy or bureaucratic text, filler text, based on a few main points Give me the main points of a text—but from a text I don’t need to read. This is not a replacement for a full reading. If the LLM didn’t exist, I simply wouldn’t read it. Ask questions in the classroom (I am a professor) just to see a summary of a topic and, perhaps, find some information I had forgotten—or, even better, a hallucination. Manipulations with copied texts from web pages. For example, suppose I copy something from the internet, and it comes with a lot of clutter. I ask the LLM to remove the clutter. Transcribe text from images to use as ALT on Bluesky. I mostly use Mistral Le Chat for this. Generate programming exercises. And more... I am NOT using them to: generate images (unless as a capability test) write blog posts for me (I did it a few times and the results were not great) Podcast Suggestions If you're using LLMs, I recommend listening to the following podcasts: Better Offline, by Ed Zitron System Crash, by Paris Marx and Brian Merchant I've listened to a few episodes of both, and they're excellent. I know there's another related podcast, but I haven't listened to any of its episodes yet: Mystery AI Hype Theater 3000 One of its hosts is Emily M. Bender, the first author of the well-known paper: On the Dangers of Stochastic Parrots: Can Language Models Be Too Big?

Apr 4, 2025 - 17:55
 0
Do you use LLMs? And some podcast recommendations

Right now, I am using LLMs to (a few examples):

  • Translation
  • Given the transcript and/or script of an episode of my podcasts, generate a summary (show notes) for the episode
  • Generate fluffy or bureaucratic text, filler text, based on a few main points
  • Give me the main points of a text—but from a text I don’t need to read. This is not a replacement for a full reading. If the LLM didn’t exist, I simply wouldn’t read it.
  • Ask questions in the classroom (I am a professor) just to see a summary of a topic and, perhaps, find some information I had forgotten—or, even better, a hallucination.
  • Manipulations with copied texts from web pages. For example, suppose I copy something from the internet, and it comes with a lot of clutter. I ask the LLM to remove the clutter.
  • Transcribe text from images to use as ALT on Bluesky. I mostly use Mistral Le Chat for this.
  • Generate programming exercises.

And more...

I am NOT using them to:

  • generate images (unless as a capability test)
  • write blog posts for me (I did it a few times and the results were not great)

Podcast Suggestions

If you're using LLMs, I recommend listening to the following podcasts:

I've listened to a few episodes of both, and they're excellent.

I know there's another related podcast, but I haven't listened to any of its episodes yet:

One of its hosts is Emily M. Bender, the first author of the well-known paper: