David Boyle 1 16 23 video1616765699

David, what prompted you first of all to get so interested in Chad cpt?

Well, I was just fascinated by its potential. I just kept seeing weird and wonderful uses on the Internet, mostly like wacky uses. And I just thought, well, if it can do all these crazy creative things, then maybe it can help me to do my job. And so the book came about from a very selfish attempt at saying, how can it help me to do my job? And we just went through like activity by activity that we do every day and said, how can we make this tool helpful? And sure enough, it turned out to be helpful in most of those ways. And then we decided, after very selfishly trying to work out how to use it, we decided then to write that all down and share it with everybody else in the hope that it would be useful. Everybody else.

Okay, so it has many uses. I haven't really up until here, heard anybody really talk about its research capabilities and such. But talk to me about the time frame between you first started to get, I'll use the word obsessed with Chad CPT and actually deciding to write a book about it.

I guess it was a few days of really messing around with it before we really decided to write the book. And it really was just a few days though. And in those few days there were so many interesting examples. Like I would work out, oh, it can help do this thing. And then Richard would send a message saying, oh, hang on, but also it can help doing this thing. And so it was a few days, but a few days of really rapid iteration. And then we said, look, there's three weeks until Christmas. And I think if we keep going at this pace, like if we keep finding these interesting seeing ideas and if we keep writing down how to do it, I think we could string that together. I think that'd be a pretty valuable handbook for people like us. And so we looked each other in the eye and said, look, are we crazy or does this feel important enough to stop what we're doing, stop all of our other work and just focus on this for two and a half weeks? And we looked at each other and said, yes, this is a life changing moment. This technology really changes everything about how we work and we want to help spread that word. So we dropped everything and wrote a book.

So one of the questions I normally have asked over 25 years when I see something new and dramatic and dynamic like this is if it were to take hold in market research, what would it cannibalize that already is pretty heavily in use.

That's a really good question. I Think of it slightly differently, though. I think it makes everything a little bit easier and something a lot easier. I think it makes whether you're trying to come up with hypothesis at the start of the project, whether you're trying to broadly understand an industry before you get stuck in, whether you're trying to think about what audience segments might exist, or whether you're starting to think about what marketing strategies might be good responses to those audience segments, like it doesn't matter where you are in that process. This helps you get to an answer like quicker, easier, richer, and in a way that's better communicated than you otherwise would have done. So my hypothesis is that actually whether you're super junior and whether you're doing very tactical work, or whether you're super senior and doing very strategic work, this tool actually just helps you to do what you're doing quicker, better and better communicated than it otherwise would have been. So I don't see it replacing anybody. I think it just gives everybody a superpower.

You've told me that it is, in your mind, faster and less expensive, but it may not or probably is not better. So delve a little bit more into what you just told me about exactly what it can do to provide a superpower to a typical researcher.

Yeah, that's a great point, because it's definitely cheaper than any other method because it's free and any other method takes at least time. So even if we're saying hypothesis about audience segments, I could do that myself, I always do that myself. But this is quicker because right there at the push of a button, almost instantly it's given me some hypothesis. It's quicker than I could have done, therefore cheaper, because it would have cost my time and attention to do it, or an intern's time of attention. So it's quicker and it's cheaper because it's free and it's so radically quicker and so radically cheaper that even though it's not better than the ones I would have come up with, it's very good and therefore I can apply it in many more ways than I otherwise would have done, I can take on many more projects, work with many more brands, work in many more industries than I other would otherwise would have done. And therefore it's better because what I have to do now is to polish and refine and edit each of those different projects, rather than having to do all of the end to end work on them. So the end project's better, even though ChatGPT isn't better, but it's so radically quicker and so radically cheaper that it's gets to the same answer.

So from a standpoint of faster what, what I have heard is that it's, it eliminates your first go around in dealing with any research issue that you're confronted with, from which point you can then take the material and your self improvement. So you've taken a big savings in time from the time you ordinarily might have had to invest in that first step, whether it be 20 minutes or an hour or even more than that. Is that a fairly accurate way of looking at it and kind of express it that way for me?

David, let's say you're coming up with hypothesis for different category related needs. I'm studying peanut butter and I want to understand what are the category related needs that drive peanut butter behavior. Now in the old days, how would we have done that? Well, me and my team plus the client would have had a brainstorm and we would have come up with needs. We'd have done some desk research and over about a week or two, practically speaking, we would have come up with a really comprehensive, really thorough list of all the category related needs to start our research with. Now, in the new world and chatgpt world, I type in chatgpt, hey, I'm interested in what are the category related needs of peanut butter? Please give me 10, let's say and instantly it will give me 10. Now in my experience, probably three or four of them are really good, really spot on and you know, it's great to see. Probably would have come up with them anyway. Two or three of them, probably you've not quite read the brief. You, you're not quite thinking the way I'm thinking. I'm going to ignore those. And two or three of them might be just plain wrong, you know, not, not, not necessary for this project. Not out of scope in some way, but there's some gold in there as well that you otherwise wouldn't have thought of. There's always something you're like, oh, it would have taken me a long time to think of that, that, that really niche one. So it'll surprise you in interesting ways. It does require editing and tweaking and it will get you to where you would have got after a week or two of collaborative work. It'll get you there almost instantly. You meaning the collaborative work. And that time can be spent enhancing, enriching and going on to the next step of the project. So it saves you lots of work in lots of ways. It's better in lots of ways. And it does require editing and tweaking though.

And your own version of chat CPT that you use, from what I understand, learns every time you have it do same kind of task. Is that true? And obviously there would be great benefits to that.

Yeah, it learns in a couple of different ways. So the first way is across all of the users all around the world, the things that people use it for, it learns. The responses that work, that don't get edited or that end up getting used, it learns. Ah, that must be what people were looking for. So collectively it's learning and getting smarter all the time. But also it's personally learning. In any one chat session it remembers who you are and what questions you've asked. And so in any one chat session it's learning about your style and what you want as well. So whenever I do a project, the first thing I do is to tell it who I am. Next I tell it who my organization is, audience strategies, how we think about the world, our approach. And then I ask it for, let's say category related needs. Then it knows exactly who I am and how I like to think about it, how my organization thinks about it, what language we use, and therefore it will give me an answer that's in our house style and our house way of thinking and that matches our kind of brief for the project almost. So yeah, you can personally trade it within any one chat session, which is a really valuable way to get much better results out of.

Would seem to me, and we didn't talk about this the other day, but I did think about it over the weekend, that the runaway enthusiasm and use of chat CPT has to have a lot of AI people out there trying to come up with something of their own that will be markedly, maybe even significantly superior and available in the short term. What do you think of that concept?

There are so many different versions of this technology already available. So OpenAI who created ChatGPT have like four or five other versions available that they're slightly different for different use cases. Google have their own version, the biggest and best, which is not released yet. There'll be new versions coming out from all sorts of people. So I think I would recommend not worrying about the specific technology that we're talking about here. And in many ways when we use the chase, the phrase chatgpt, what we mean are sufficiently advanced chatbots and lots of what's in the book. For example, you can apply to other chatbots as well. So yeah, there'll be other ones coming. There are already other ones out there, there'll be other ones that are even better. What I'm saying I think is don't worry about what the future holds, don't worry about that promise, don't worry about that potential. What's much more useful is to say this one chat GPT is available right now for free for everybody and it can do at least all these things really, really well. So your time is probably very well spent working out how to use what we have already. And then after that is a really good time to worry about what's coming next. But master what's already available and it can already change everything about how you work. Master that now and worry about the future later.

And the current version, which is good through 2021 as far as its capabilities, is due for a revise, isn't it?

Yeah. The current version is trained on data on the Internet and the state of human knowledge up to about the end of late 2021. Actually for most use cases in the book, it doesn't require current knowledge of the world. You think about again, category related needs, which is at the heart of a lot of what and how we think about audiences. Category related needs don't change that much over a period of many years. Tactically how you execute marketing might change a little bit each year, but the strategy and the underlying audience needs, they don't change that much. So actually, even though it's old, because it's data is only up to the late 2021, it doesn't impact any of the use cases in the book itself. But yeah, that's going to get retrained on new data or get updated. And there's lots of applications available that you could kind of plug into it to update it with modern information as well. For example, just right now, what you could do is take a news article or report about the current state of, let's say peanut butter marketing, you can paste that into your chat session to train it and then you can ask ChatGPT to summarize the themes or the topics that are current in the world of peanut butter marketing and it will do a great job for you. So bring your own data in if the data you want isn't in the training set.