On First Drafts
Kevin Kelly: Yes, ChatGPT, Bing, and stuff, and Google. For me, I’ve always had problems making that first draft. It’s just a killer.
Tim Ferriss: Yes. And I know the feeling.
Kevin Kelly: You know the feeling. I find it helpful in making the first draft.
Tim Ferriss: How do you prompt it? What would be an example approach?
Kevin Kelly: All different ways. And I’ve been collecting this. And here’s the thing about —
Tim Ferriss: Your book of spells.
Kevin Kelly: Yes. Here’s the thing about it, is that this is an important lesson about technology, is that we have to use it to figure it out. There’s something I call thinkism, which is this reliance on trying to solve problems by thinking about them, which is very appealing to people who like to think. And you can only go so far with thinkism because all the things we’re discovering about this, none of the inventors of this had any idea that they could be used this way.
Tim Ferriss: That’s cool.
Kevin Kelly: Right? And so we’re discovering, we collectively, by using it are discovering its capabilities and eventually its harms. But that’s important because this is how we steer the things. And so the problem with trying to prohibit or turn it off or ban it, is that you don’t get to steer anything. Going back to that metaphor.
So right now through use we’re uncovering all these things. And I’ve been trying to track how people are actually using them. For instance, chats. There’s a couple prompts. So here’s the thing, these chat models, basically what they generate are wisdom of the crowd kind of knowledge. The wisdom of the crowd is very famous, counting the jelly beans. If you average all the attempts by humans to count the number of the jelly beans in a bottle, the best guess, the most accurate, was the average of it. And that’s what we’re getting with the chat. It’s taking everything as written, the plus and the minuses, the geniuses and the jerks, and it’s averaging out. And that’s what it’s giving you is an average.
So most of the content generated by the chats is broadly correct, very average, very bland. And a lot of what you’re doing with the intern is pressing them. So one of the tricks is that you can ask for it to be a little bit snarkier or more professional.
Tim Ferriss: So let’s say you’re starting tabula rasa. Idea popped into your head in the shower. Okay, I want to give a rough draft a shot. What is the step number one, step number two?
Kevin Kelly: So it depends, but I might ask it to do a summary of what’s known about this. Tell me everything that it knows about it. And then maybe write a first draft with bullets, five bullet points. And then I might —
Tim Ferriss: Could you give a real example or a example you might use?
Kevin Kelly: I’m trying to think of the last one I did.
Tim Ferriss: How the Egyptians influenced Roman architecture. I don’t know, making that up.
Kevin Kelly: Exactly right. You could do that. Give me the five bullet points and stuff. And then you could have questions about some parts you didn’t understand or expand bullet one citing more sources or give me an example of a day in the life of this or 10 more examples of how this might play out. You could expand it that way. You could also shrink it in terms of summarizing, making bullet points or what’s the key takeaway or how about if I wanted to have a teachable moment out of this. And so you would have all these kinds of things flowing around. And then, again, it’s the intern at work. It’s good but you’re not going to use it. You’re probably rewriting it. It maybe gives me some ideas I didn’t have or maybe the structure of how it organized it, that’s pretty good for four of them. And so it’s a start.
No comments:
Post a Comment