AI-powered
podcast player
Listen to all your favourite podcasts with AI-powered features
How Many Tricks and Hacks Are Left Out?
The difference between GPT say three, 3.5 and four, is it mostly a bandwidth thing where it's given basically the token window or the ability to send it more information at once? Is that the primary difference besides like these little hacks and shortcuts and efficiencies? No, I mean, so from a user perspective at inference time, a big difference will be the size of the token window.