After decades of using free online services, and seeing the consequences of being the product and not the customer (see Soshana Zuboff “The Age of Survailance Capitalism”, or the Netflix documentary “The social dilemma”.), I’m getting less stingy and I try to pay for services that are truly useful to me, especially for my work.
That said.
ChatGPT gives completely different results if we use the free model (GPT-3.5-turbo) or the paid model (GPT-4 and GPT-4 plus plug-ins). Since I’ve had access to GPT-4 I use it daily for many tasks, and often my documents are the result of a dialogue with the machine. If I think of GPT-4 as an intern, at intern rates it pays for itself in a single day, or in a single morning.
On the other hand, our students will surely use the totally free version, as we would have at their age. And we need to understand how both versions work and what they can do.
In any case, using the GPT-4 version gives you a sense of the quality of what’s coming. My impression is that the jump from GPT-3.5 to GPT-4 is like going from nothing to GPT-3.5.