GPT-4’s Secret Weapon: Token Limits
I’m disappointed by GPT-4. Not because it’s bad, no it’s a lot better. But because it’s a lot more expensive than GPT-3.5.
And because of that I’m not interested in using GPT-4 in my apps. But then I started thinking… is GPT-4 even worth it?
If you’re using it for a business that’s making a lot of money then yes, it is worth it. Especially if you consider the alternative which is hiring an actual person. But sad to say most of my apps are still not bringing in a lot of money.
But then I realized, there is a feature in GPT-4 that is not talked about very much. No, it’s not multimodel inputs, or it’s improved reasoning. It’s token limits.
Token Limits
When I first used ChatGPT I thought it remembered the entire conversation thread. I thought how it works is when you type something it takes what you say as input and saves the internal state. So next time it can just append more information to the end.
No, apparently OpenAI chose not to do it this way. Apparently ChatGPT will take in only…
