justin․searls․co

GPT 3.5 is a lot worse than GPT 4

It should shock no one to learn that Open AI's newer, better language model is an improvement over the old one, but if you aren't an active user of any of this newfangled AI stuff, it can be easy to lose track of just how much better things are getting and how quickly.

If you subscribe to GPT Plus, ChatGPT will also implement the ReAct pattern for requests it thinks can be formalized, which is one way to mitigate hallucinations.

Pictured here, asking "days between 12/10 and 2/11":

  1. GPT 3.5, which gets it completely wrong with a nonsensical reasoning
  2. GPT 4, which gets it right and even has a little terminal prompt link
  3. Opening up that link will actually show you a Python script GPT 4 used to compute the answer

Cool beans. 🫘

Got a taste for hot, fresh takes?

Then you're in luck, because you'll pay $0 for my 2¢ when you subscribe to my work, whether via RSS or your favorite social network.

I also have a monthly newsletter where I write high-tempo, thought-provoking essays about life, in case that's more your speed:

And if you'd rather give your eyes a rest and your ears a workout, might I suggest my long-form solo podcast, Breaking Change? Odd are, you haven't heard anything quite like it.