You are currently viewing Meet your AI creative collaborator

Meet your AI creative collaborator

Meet your AI creative collaborator

How to work with AI

Selecting transcript lines in this section will navigate to timestamp in the video

(keyboard clicking) (bell dinging) – Hello and welcome to the course. Over the next few lessons, I’m going to be showing you how to use ChatGPT to improve your research and your writing. If you’ve not done so already, then head along here and sign up to ChatGPT. It just takes a couple of minutes. AI is important because it’s a seismic shift in computing. In the space of just a couple of months, ChatGPT is shaking up the world of tech and impacting the market valuations of some of the world’s biggest companies. So why is that? Let’s look at it from the human perspective. The way that we add value in our work is either through thinking or doing. Doing has been increasingly automated for a couple of hundred years. Weaving by hand became less relevant once engineers created machines that could do it faster and better, and now many automotive production lines are entirely staffed by robots. This was all work that involved repetition rather than thinking, so it’s only now with the rise of AI that thinking is being automated, and that’s why it’s making a change. You can see this change as a threat, or you can see it as an opportunity, and believe me, it is a huge opportunity. This course covers input, process, and output, where input is research, writing is the process, and output is the text. We’ll be identifying parts of that process where AI can help us, but it can’t do everything well. You still need to add your own thinking and apply your judgment, expertise, and abilities. Let me start by talking about how we should work with generative AI and ChatGPT in particular. First, let’s talk about your attitude. You should see ChatGPT as a collaborator. Both sides need to be involved throughout the process to get a great result. There are three parts to this process. There’s preparation, there’s interaction, and there’s continuation. Preparation is vital. You should know exactly what kind of result you’re after before you start asking anything. When you define the output first it helps you ask the right questions. And once you know this, you can write your prompt. When you write a prompt, that is your brief to the AI. Most people start by writing a short sentence like, “Write me a poem about sentient tennis shoes.” That’s fun and everything, but if you want AI to help you with your work, you have to be a lot more specific. The prompts I’ll be showing you are paragraphs rather than sentences. They require you to add context and define what the output should look like. The truth is, if you give the AI less information than you’d give a human doing the same task, you can’t expect great results. Now we’ve got interaction. This is your conversation with ChatGPT. Your job is to direct that conversation and keep it going in the right direction until you have the response that you’re after. Finally, we’ve got the continuation. You may have been thinking that this should just be the result, but I think this is where you have to continue the work. You use your judgment. You challenge the output to see if you can improve it. You refine the words to make them better. In general, you add the human elements that even the best AI can’t quite nail yet. It’s not about AI replacing human thinking. This is about AI augmenting it to help you achieve more. In the next lesson, I’ll be showing you how to write a good prompt, so stick with me. (bell dinging) (keyboard clicking)

How to write an effective prompt for AI

Selecting transcript lines in this section will navigate to timestamp in the video

(keys typing) (bell dings) – There is an art to writing a good prompt and I’ll be sharing an easy framework that will help you become a better prompt engineer. That’s the term for it. Let’s understand that ChatGPT is drawing on a massive dataset taken from numerous sources. This covers lots of topics and ranges from great quality to low quality. The job of a great prompt is to focus in on a specific topic and to try and narrow in on the good quality data. I’ll explain how we can go about doing that. It just so happens that the framework follows an easy mnemonic, CREATE. Let me take you through each of the letters in turn. C is for character. It’s a good idea to tell ChatGPT what role you want it to play. So I’ll often say something like, you are a talented copywriter with 20 years experience writing high-performing sales copy. That tells the AI what information it should draw on, not just zeroing in on sales copy, but drawing on the best sales copy. So imagine that ChatGPT is an actor and you’re letting it know its role and motivation. Now we’ve got R for request. This is where we tell the AI what we want it to do. I recommend that you’re specific here so it’s not just write me a sales email for a sports car. You know, it should be giving context and useful information. It’d be much better to say something like, write me a persuasive and enthusiastic sales email for the Cougar Hyper Sport, totally made up. This is a high-performing electric sports car with the best acceleration in its category. The sales letter should tell people they need to experience it for themselves, find out what it feels like to be pinned to your seat, to feel the heart-thumping exhilaration of cornering, and to enjoy the experience of everyone looking to see who’s in the driving seat. That’s quite a lot, isn’t it? But don’t skimp here. This is what defines the quality of the response. Now we’re onto E for examples. This bit is optional. If you want to give ChatGPT some better direction, it might be worth adding some examples, maybe some examples of previous headlines or an example of the tone of voice. ChatGPT is great at learning from these and delivering a response in a similar style. Next, we’ve got A for adjustments. This is when you’ve tried a prompt and it’s not quite giving you exactly what you want as a response. These are the refinements. For me they’re often phrases like, don’t use bullet points, or use subheads to break up the text. This is where the noodling happens to get things just right. And now we’ve got T for type of output. Tell ChatGPT exactly how you want it to format its response. An example might be, deliver your response as a 500-word article with a headline and a conclusion. The type of output can include articles, bullet points, tables, poems, jokes, scripts, just about anything that can be written. And finally, we’ve got E for extras. These are the little additions that can really make a big difference. I’m going to give you three examples here. The first is ignore everything before this prompt. This is useful if you don’t want to draw on any of the conversation that you’ve had up until this point. I will often use this at the beginning of a prompt when I’m changing tack. Next we’ve got, ask me questions before you answer. Ugh, I often use a variant on

How to work with the responses from a generative AI chatbot

Selecting transcript lines in this section will navigate to timestamp in the video

How to work with the responses from a generative AI chatbot

Selecting transcript lines in this section will navigate to timestamp in the video

(keys typing) (bell dings) – Working with an AI shouldn’t be about helping you do less work or helping you produce more work. It should be about helping you achieve better. And the way that we do that is by adding our own human abilities on top of the AI response. It’s important that you don’t make the mistake of thinking of AI in the same way as a search engine. With a search engine, you put in a query and it returns results. Boom, you’re done. With AI, it’s more of a journey. You need to refine your request to get closer to what you were after. It’s a back and forth. It’s a conversation. That’s why ChatGPT starts with the word chat. You should think about the AI as a collaborator rather than as a servant. You need to give as well as take. What you put into the conversation defines what you get out of it. If you don’t like what you get, give feedback so that the AI knows exactly how to improve its response. If it gives you something you like, say so and ask the AI to expand on that thinking. If you get an idea during the conversation, share it with the AI and ask it to explore that concept. The results you get are defined as much by your input as by the AI’s output. Don’t see the results it gives you as a finished product. The AI is not human, although it does a pretty good job at mimicking human output. Your humanity is important. You need to apply your own abilities to the output to make it even better, to take it from adequate to excellent. I also recommend that you double-check the accuracy of the output. You can get output that sounds entirely plausible but is in fact completely made up and you don’t want to be working with inaccuracies and lies unless, of course, you work in politics. And finally, be honest if you’re using AI. Don’t try to pass off AI-generated work as your own. There is nothing to be ashamed of in using AI to produce better results. In fact, that makes you really smart because it gives you an advantage and benefits the organization you’re working for. In the next lesson we’re going to start playing with actual prompts. So get ChatGPT loaded up on a browser and let’s get going. (keys typing)

Using AI to summarize complex information

Selecting transcript lines in this section will navigate to timestamp in the video

(keyboard clacking) (bell ringing) – If you want to come up with great ideas, with great solutions, then you need to start with great research. The brain is a processor. You’ve got input, process, output. People tend to think that the ones who are amazing at coming up with ideas have got much better processing, they’ve got stuff they do in their brains that is just special and magical. Actually, the people who I know who are the best at coming up with ideas have got amazing input. So the breadth of your input actually dictates the opportunity for your output. So we’re going to start looking at how we can use ChatGPT to help us with our research so that we can get more information and better information to work with. So let’s start with our first prompt here, that again, use our create framework, starting with our character, and then looking at our request, and moving on from there. So let me read out here the first prompt that we’re going to work with. “You are a highly experienced writer who writes concise and readable text without stop words, filler words or jargon. I want you to summarize the following text, highlighting the most important concepts. Deliver this as a short paragraph of 100 words. Then list the most important points as bullet points. Finally, follow it with a one sentence summary. The text I want you to summarize is…” Now the text that will pick, let’s think, something random, maybe text from this book here. This is a fantastic book, it’s written… Oh, it’s written by me. There we go. So I wrote a chapter in this where I talked about how in the future we will be using AI as our creative collaborators. And look, here we are doing a course on it right now. So let me take part of a chapter and paste it in here. This is quite a lot of text. We’re going to paste in and ChatGPT is going to give us a summary. So here we go. It’s working on it. “The text is about the increasing role of teamwork in the publishing of new ideas and the evolution of the concept of creativity.” And there we go, as I read through it here, it’s doing a pretty good job of summarizing this text. So it’s done the paragraph there at the front. Now it’s going through the bullet points. And here’s the summary. So our final one sentence summary at the end, let’s see how accurate that is. “The text is about the evolution of teamwork and creativity, including the increasing role of teamwork in publishing and the debate around whether computers can be considered creative, with a conclusion that the definition of creativity may evolve to include non-human creations.” And I believe it will. This has done a really good job of summarizing this information here in a very digestible way. So there’s other stuff that we can summarize. Now, what about taking a book that’s already out there? We’re not going to paste any text in this time. We’re actually just going to rely on the information that ChatGPT already has. So “you are a highly experienced writer who writes concise and readable text without stop words, filler words or jargon.” Just like before. “I want you to give me a summary of the book.” Let’s again, just pick a book at random here. What about “Iconic Advantage” by… Who are the authors of “Iconic Advantage?” Soon Yu and Dave Birss. Oh yes, here we go. There’s another book that I helped to write. So let’s put that in there, Soon Yu and Dave Birss. So this is now using the information that it has, and it’s got information on this book, and it’s now going to summarize this book based on what is in its dataset. And I’m just going to read through these and check that they’re actually accurate. Yeah, that actually looks pretty accurate. So this isn’t bad. Make sure that it’s a book that is done quite well. If it’s a rarer book, a more unusual book, then it won’t have the information to do the summary for you. The other thing is that it will sometimes make up information. So it will sometimes add stuff in there that’s not entirely true. So maybe be a little bit careful, but on the whole, this does a fantastic job of summarizing books. So let’s now think about other stuff we might want summarized. And our final prompt here, we’re going to look at academic thinking. So sometimes you just want to know what is the academic thinking on a particular topic, please summarize it for me so that I can get a basic understanding of it. So here we go, again, we’re starting with this, “you are a highly experienced write who writes concise, readable text without stop words, filler words or jargon.” Okay, now, for a request, “I want you to give me a summary of current academic thinking around the field of…” I’ll put the field in there in a second. “Highlighting the most important concepts. Deliver this as a list of bullet points and follow it with a one sentence summary.” So the topic that we shall go for, let’s see here, neuroscience of creativity. So this is something that I know quite a bit about. So I’m able to tell you if this is accurate or not. Let’s see what happens when we read through this text. This is actually great. So I was worried at the second point here where it said “the right hemisphere of the brain is commonly associated with creative abilities while the left hemisphere is linked with logical and an analytical thinking” because neuroscientists have debunked this. In the next point, it goes on to say, “however, recent research suggests that both hemispheres are involved in creative processes and work together to produce novel and useful ideas.” Thank goodness it saved itself there but everything it’s said here is really good. So these are three prompts that you can use to give yourself information that will help you come up with better ideas. (bell ringing) (keyboard clacking)