A very astute framing by Ted Chiang—large language models as a form of lossy compression for text.
When we’re dealing with sequences of words, lossy compression looks smarter than lossless compression.
A lot of uses have been proposed for large language models. Thinking about them as blurry JPEGs offers a way to evaluate what they might or might not be well suited for.
It doesn’t bother me much that bleeding-edge ML technology sometimes gets things wrong. It bothers me a lot when it gives no warnings, cites no sources, and provides no confidence interval.
Yes! Like I said:
Expose the wires. Show the workings-out.
Whatever the merit of the scientific aspirations originally encompassed by the term “artificial intelligence,” it’s a phrase that now functions in the vernacular primarily to obfuscate, alienate, and glamorize.
Do “cloud” next!
This could pair up nicely with the most dangerous writing app.
If you rely on Word, Evernote or Notion, for example, then you can’t work unless you have Word, Evernote, or Notion. You are helpless without them. You are dependent.
But if you only use plain text, you can use any program on any device, forever. It gives great flexibility and peace of mind.
Tired of using "lorem ipsum dolor..." for placeholder copy? Use real English words that, while apparently non-sensical, transform into stories when spoken aloud.
Please read Miriam’s latest blog post.
I’m trying to be open to changing my mind when presented with new evidence.
I listened to a day of talks on AI at UX Brighton, and I came away disappointed by what wasn’t mentioned.
It’s almost as though humans prefer to use post-hoc justifications rather than being rational actors.
Using generative large-language model tools? Sleeping well at night?