Loading...
RUNEUAI 5-1 Efficient Prompts
Almudena
00:10Hello, Hello,
hiramb
00:13good evening. Hey,
Bernard Goldbach
00:15Hiram, how you doing?
hiramb
00:16I'm okay. How are you?
Bernard Goldbach
00:21It's over a very wet week last week, so feeling good, yeah.
hiramb
00:30How are ish? Here it is dry.
Bernard Goldbach
00:33Very good. I know Maria said she's going to join us. Eleanor won't. Looks like a good cross section of people, Vanessa should be here as well soon.
Bernard Goldbach
00:51Folks who are asking about problems getting into the Google Drive, I'm trying to allow people who have a Google mail address to also have an identity to go in there some schools seem to restrict access to Google Drive and Google Gemini. So the only work around appears to be trying to have a Google address or an address that they know about. Hiram, I'm making you a co host of this as well. So if you don't mind letting people in to the room, I expect a few people will be here. Two or three more people will be here. So looking forward to that, I'm going to walk us through a series of slides that take us into some of my own personal shortcuts for the use of effective prompts. And then we'll do a few things as well with a couple of different services, which I hope will make people feel comfortable with not only newly learned prompt skills, but other perspectives, which should give you a higher level of confidence with the services we're introducing you to. So this is I'm calling it. I believe I'm calling this session five, which may be different than the 
normal number of times we met. But just to keep my own accounting correct, I want to start with a short campfire session and then walk into the main session. Filipe won't be with us today. He's meeting some people from running you at the during this same timeframe, so I'm hoping that means good things. I'm just reemphasizing the fact that these are the AI tools, Gemini from Google Cloud, AI and tropic chat, GPT, which a lot of people use, and it's actually five now for Omn, the is actually the model in the background that people are using, but five shows up in your browser, copilot from Microsoft and Perplexity, which we'll use tonight, I talked to my AI's, and with the visit of a few people from America, my cousins from America, who stayed in Ireland last week, I know they were literally talking using their own language, their own conversation with The microphone option with their phones to get feedback from Google, Gemini audio feedback. And it is not uncommon for people to develop a bit of a relationship and a shortcut certain phrases they use to to get actions back from an AI. Ethan Malik makes this point in his book co intelligence. So I guess the cautionary tale is if you're not able to make a distinction between your car and a real person or your horse, so the human and you talk to things like your blender in your kitchen as it were a living object, well, be aware that you could also have the same behavior pattern with an AI, and that's not always healthy, but it is happening. I need to mention that workflow is what we're picking up when we go through this module, this training program, authenticity while working with these tools is important, and your ability to create practical workflows with the help of AI is critical. So this week, I'm just mentioning in this small little slide that I normally start with handwritten notes. I know Maria and I work in the same building, so she might have seen me scrolling on a piece of paper about what I would like to cover tonight, I use electronic updates that come to me through two apps that are AI helped though AI helps fabric, then AI helps Instapaper, and then I use cross checking with a variety 
of other tools to make sure I'm not totally out of out of step, and that the AI inside of Microsoft helps me improve the accessibility of this slide deck. So for example, this is what would have shown up this morning in my fabric, I would have saved something about a service standard called the atmosphere prototype. It allows you to connect between different service providers like x, well, not x, but actually Mastodon and blue sky and a few other services. You'll see something from Grace, Leung about Comet, the AI agent inside Perplexity, and there are 15 hacks to know about managing different AI. So I'll share with them, share you, share with you that particular piece of information. And then from instant paper, I'm getting a lot of stuff recently about, hey, look, if the stock market gets too hot because the AI services are maybe overvalued. There could be a problem. But there's plenty of excitement in this space around AI, and your enrollment on this module seems to prove. Look, it percolates all the way down to people who are going to be graduates, recent graduates in the economy, as well as mid career professionals. Today, I want to cover this. I want to if Vanessa's on the call, want to get something from her, Maria, I know you are so quick. Thing from you. We did not listen to the entire clip, unless everybody on the call says, You have I want to play that job. Oppolis Jobpocalypse discussion with the Financial Times, just because it does hit home with a big concern I have, which is there are issues about people the jobs they would hope to normally transition into, and how AI might be a stumbling block now, specifically the AI 
an AI love affair, a lot of big companies might be preventing people from getting the jobs. We use Perplexity to challenge what the Financial Times is saying about the Jobpocalypse, and I can't say the word, but the turmoil in the job market, there are some prompting techniques, which, in fairness, I should probably put at the top of this deck, but I'm not, because we use these prompting techniques. So I teach you about using Google Gemini to explore two places. So a quick campfire thing. This is, this is what, why I want to do this now in the course objectives laid out by run EU, I mentioned the word campfire, but I might not have explained exactly where it fits. Now, in the past, I use the word campfire with people to say, please explain to us casually about how you might be doing something with a service or a capability you've learned in terms of this module, I want to make sure you can also suggest how what you're doing might fit a workflow. So it's important that you're the framework of how you work and the system by which you have to do the work. It might be a subscription service your school has, or might be a preferred application your company or your organization has. Those things may use an AI or maybe not. But the idea is that you list what you need to do as an outcome, kind of explain the workflow that goes into that. And maybe say, and here's the app I'm using, and maybe some AI I'm using. So this is how I hope that you could break it down. So Maria, you know the task you have to do, why you might use Canva for it, a sample workflow possible AI that might be on it in your flow of work, and then we'll give you some feedback about it. So Maria, you're on the call, so if you want to take the screen, you should be able to grab it. And just if you have something to say, otherwise, just go full screen and say, right, this is me and how I do it. So go ahead.
+2
normal number of times we met. But just to keep my own accounting correct, I want to start with a short campfire session and then walk into the main session. Filipe won't be with us today. He's meeting some people from running you at the during this same timeframe, so I'm hoping that means good things. I'm just reemphasizing the fact that these are the AI tools, Gemini from Google Cloud, AI and tropic chat, GPT, which a lot of people use, and it's actually five now for Omn, the is actually the model in the background that people are using, but five shows up in your browser, copilot from Microsoft and Perplexity, which we'll use tonight, I talked to my AI's, and with the visit of a few people from America, my cousins from America, who stayed in Ireland last week, I know they were literally talking using their own language, their own conversation with The microphone option with their phones to get feedback from Google, Gemini audio feedback. And it is not uncommon for people to develop a bit of a relationship and a shortcut certain phrases they use to to get actions back from an AI. Ethan Malik makes this point in his book co intelligence. So I guess the cautionary tale is if you're not able to make a distinction between your car and a real person or your horse, so the human and you talk to things like your blender in your kitchen as it were a living object, well, be aware that you could also have the same behavior pattern with an AI, and that's not always healthy, but it is happening. I need to mention that workflow is what we're picking up when we go through this module, this training program, authenticity while working with these tools is important, and your ability to create practical workflows with the help of AI is critical. So this week, I'm just mentioning in this small little slide that I normally start with handwritten notes. I know Maria and I work in the same building, so she might have seen me scrolling on a piece of paper about what I would like to cover tonight, I use electronic updates that come to me through two apps that are AI helped though AI helps fabric, then AI helps Instapaper, and then I use cross checking with a variety +2
of other tools to make sure I'm not totally out of out of step, and that the AI inside of Microsoft helps me improve the accessibility of this slide deck. So for example, this is what would have shown up this morning in my fabric, I would have saved something about a service standard called the atmosphere prototype. It allows you to connect between different service providers like x, well, not x, but actually Mastodon and blue sky and a few other services. You'll see something from Grace, Leung about Comet, the AI agent inside Perplexity, and there are 15 hacks to know about managing different AI. So I'll share with them, share you, share with you that particular piece of information. And then from instant paper, I'm getting a lot of stuff recently about, hey, look, if the stock market gets too hot because the AI services are maybe overvalued. There could be a problem. But there's plenty of excitement in this space around AI, and your enrollment on this module seems to prove. Look, it percolates all the way down to people who are going to be graduates, recent graduates in the economy, as well as mid career professionals. Today, I want to cover this. I want to if Vanessa's on the call, want to get something from her, Maria, I know you are so quick. Thing from you. We did not listen to the entire clip, unless everybody on the call says, You have I want to play that job. Oppolis Jobpocalypse discussion with the Financial Times, just because it does hit home with a big concern I have, which is there are issues about people the jobs they would hope to normally transition into, and how AI might be a stumbling block now, specifically the AI +1
an AI love affair, a lot of big companies might be preventing people from getting the jobs. We use Perplexity to challenge what the Financial Times is saying about the Jobpocalypse, and I can't say the word, but the turmoil in the job market, there are some prompting techniques, which, in fairness, I should probably put at the top of this deck, but I'm not, because we use these prompting techniques. So I teach you about using Google Gemini to explore two places. So a quick campfire thing. This is, this is what, why I want to do this now in the course objectives laid out by run EU, I mentioned the word campfire, but I might not have explained exactly where it fits. Now, in the past, I use the word campfire with people to say, please explain to us casually about how you might be doing something with a service or a capability you've learned in terms of this module, I want to make sure you can also suggest how what you're doing might fit a workflow. So it's important that you're the framework of how you work and the system by which you have to do the work. It might be a subscription service your school has, or might be a preferred application your company or your organization has. Those things may use an AI or maybe not. But the idea is that you list what you need to do as an outcome, kind of explain the workflow that goes into that. And maybe say, and here's the app I'm using, and maybe some AI I'm using. So this is how I hope that you could break it down. So Maria, you know the task you have to do, why you might use Canva for it, a sample workflow possible AI that might be on it in your flow of work, and then we'll give you some feedback about it. So Maria, you're on the call, so if you want to take the screen, you should be able to grab it. And just if you have something to say, otherwise, just go full screen and say, right, this is me and how I do it. So go ahead. Maria O'Brien
08:37Okay, the problem, I ran into Bernie because I would use Canva for work. We would have different events, Lunch and Learn series, and I would use Canva to create the flyers. However, the difference is my work. One is a pro Canva. It's the paid for version. So I'd have things like the background remover features that aren't available on the free version. So because I'm here on my own laptop at home, and I can't bring in my work account into this to show you, I'm restricted on the features that I can use. But I suppose I mean, people are probably familiar with the basic free version anyway, and and the capabilities of it are, are not, I don't know.
Bernard Goldbach
09:26Okay, well, people didn't call like, Leti you. You've used Canva before, right? Okay, so And like, I don't know. Almudeah, have you done it yourself? Have you used that, that image creation program called Canva? Has anyone else used it? No, okay, no, I
Almudena
09:41use PowerPoint usually,
Bernard Goldbach
09:43okay. Maria, the difference between PowerPoint and image creation and Canva and its image creation because, you know, I have a prejudice about what Canva can do. I'm kind of amazed at what it can do. So give us an idea in terms of a workflow. Maria, say, like, if you had a PowerPoint deck to develop and you opted to use Canva to make it nicer? Explain?
Maria O'Brien
10:04I just think you can personalize Canva a lot more. There's a lot of features. You have, the access to all of the the content online. So you know, they'll make suggestions on your templates and on your designs. There's shapes that you can incorporate. So while you might go in with a set idea of what you want to design, there's always other designs that you can borrow from. You're not plagiarizing. It's just that there are suggestions there that can improve it and make it look better. It's very easy to just, you know, change the color of the background, change the font size. If the font you're you want isn't there, you can upload it, you know, so you can, you can have your own brand kit on Canva. That's really, really useful, because at one stage, I was going between two different brand kits for a variety of reasons. So the templates or the the fonts were different, the sizes were different, and yet I could just jump between the two brand kits very, very easily, just in the one app you can you can use it on the go, because, as I said, just there, I would use it on my workstation. However, there have been times where I would monitor it on the app, and it's still my work account on the app. And then, personally, I can just switch between both of those accounts as well.
Bernard Goldbach
11:41Okay, background removal. Do you use it for that?
Maria O'Brien
11:48I'd use it
Bernard Goldbach
11:49a lot. Tell me why? Why would you want to remove background?
Maria O'Brien
11:53It depends on like, if you're using an image and you have a photograph, if it's a template and you just don't want the background that they've provided, you can remove it. If it's a photograph, it's from a previous session, or it's like you we I say, if I took a screenshot of you and I want to remove Tipperary buzz quarter, yeah, I can do that with the background remover. So it's just It speeds up the process, you know, being able to reuse files that you already have. It just speeds things up.
Bernard Goldbach
12:26Yeah, so almedina that there is a free version of it. Most people get addicted to Canva, so they pay for the pro version, you know. And if your organization likes the facility, how fast you can do the work with it. And that's my thing, is, because Maria kind of mentioned this, once you build a deck, once you build a brand identity, brand, like in her case, might be the institutional identity. There might be a thing called an event identity, a brand called an event. There might be an event called or a brand that's called the financing, the financier, the people that pay for an event. So you can actually quickly modify the same content with a different look, with colors, logos, fonts that match the identity of the people that funded the event, or the executives that are looking for compliance on the way the money was spent, and it's just a subtle little way to make it look like you're make it look like you actually have several designers who are working for you, just because you've been able to master the way Canva works. Now, there are other ways of getting all that stuff done inside the Microsoft suite of services and inside of other services, but Canva is a thing that I know Maria and I know at least two other people on this program are using. So, yeah, Leti, would you want to add anything to? Since you are a user, you want to add anything to what Maria said, maybe a pro tip or observation. Okay, Maria, you're the expert.
Maria O'Brien
14:08Then, not really. But what I do find is it's, it's so user friendly, Bernie, I mean you literally, it will do the job of what previously would have taken a marketing team, content creation team. Even publishers like you can design your flyers and just have them ready to print. You know, whereas before it would have been somebody in a publishing house might have had to do publishing for you. So it's incorporated elements from publishing, from marketing, from content creation teams, yeah, so it's streamlined a lot of processes into one,
Bernard Goldbach
14:47no, and that's workflow. You know, your workflow is compressed nicely. You can do more in a shorter space of time. And then the other thing that I mentioned, there's a really simple way of sharing your work inside of the paid version of the program. You can see past templates, past projects, and it's very handy. So if your workflow is accelerated because of the cloud collaboration that you can bring to the task, that's good. So it's a big win. I don't know if anyone else on the call wants to add anything, or have any questions. Anybody want 
to add anything? Okay, I'm going to go over. I know Vanessa's normally here for this stuff, so this is, let me just put some words on the screen that relate to a service that she's using. So Vanessa's building product inside of a collaborative space called the Kunst matrix, the task she has is grab content that people share in a specific space. It should be why can't? Why couldn't. So she's using a space to share, showcase our work, the possible, the workflow she has, and the AI that she uses, I should probably show you, she would grab things from Google Drive so she would look at stuff coming into maybe the images folder. So several of you would have done an image of an avatar, maybe an avatar with images like food on it. So she would have this stuff. I can see Firoz, thank for putting that stuff in. So she she needs to put some of these images up on the wall. She also may pull these images in from food, which we'll use in another session to find restaurants that work with everybody. And the goal would be, could she? Could she use an AI to more quickly produce what she has to do? So, speaking for Vanessa, what she has to do is grab an image, make a minor resize of some of it, maybe change perspective of it. And it was when she appears at another session I'll ask her specifically about. Is she doing any post production work with it, with an with an AI, or with a tool like Photoshop or some image editing so she may modify the image slightly, just to make sure there's continuity. Image Size are the same and maybe the perspective is the same, and then it would be handy for her to know when she would go into and look for new images. So behind the scenes, we're developing a workflow, and I'm hoping that she does that, creates a workflow which simply gives her an email saying there are two new images inside the o2 images folder, or inside the o2 images Kuntz matrix folder, and then she would know, right, it's time to look at that stuff and hang it. I don't know when it's going to be possible for her to use an AI to actually hang things on the wall of that exhibitation, that exhibition space called the Kunst matrix. But I'll let her explain the next time we're together about how she might accelerate what she's doing. And do appreciate folks that put your images up there from last time when you weren't here, because she will work those images into the space called Kunst and she shows later on the call or ask her to run around the exhibition space, she'll show you that. So last week, I had a problem with the connectivity I was using to talk to the internet. So you didn't have to go through that pain with me, Maria last week, but not every. I don't know if everyone saw the Financial Times video, and it is worth watching. It's approximately 10 minutes long, because it does put an accent point next to something which is a big concern of mine. So I would have worked for decades with young people as they progress into the jobs market, and then work with recent graduates as they moved up one level of expertise in the job that they earned after they earned a university degree. And now there's some real pressure at the lower end of the jobs market across the world, because some companies are saying, we spent this money for these AI tools. So maybe we can make let the middle level managers, instead of hiring an assistant, a person, maybe we can train them to use an AI, and that will be them, their their virtual personal assistant. Now this is happening, and it is creating a tension of sorts, for young graduates to get on the first rung, the ladder of employment. And one of the workarounds for that is to be able to show to a prospective employer that you as a 20 something have the skills that the person in their mid 30s, early 40s would have claimed to have because they've just been upskilled with AI. So that's hopefully what this program is doing. It's helping somebody positioning into the jobs market to say, I have these skills, newly learned under a program that has academic credentials, and I should be able to enter the jobs market as a young graduate could. So anyway, the Financial Times investigates this issue. What they've done is they've made a short video, and what I wanted to do is 
play the thing again for folks and then use Perplexity after the fact to fact check it. So the journalists talk about how some jobs are disappearing because they're automated now there's economic uncertainty in the markets, constraining revenue companies have, so they're leveraging the expertise inside without having to go outside to to find a graduate. And I'll play this like it's about 14 minutes long. I'm pretty sure I keyed in the sound, so if this does not play for sound. Let me know it's the job buckets, as explained
+2
to add anything? Okay, I'm going to go over. I know Vanessa's normally here for this stuff, so this is, let me just put some words on the screen that relate to a service that she's using. So Vanessa's building product inside of a collaborative space called the Kunst matrix, the task she has is grab content that people share in a specific space. It should be why can't? Why couldn't. So she's using a space to share, showcase our work, the possible, the workflow she has, and the AI that she uses, I should probably show you, she would grab things from Google Drive so she would look at stuff coming into maybe the images folder. So several of you would have done an image of an avatar, maybe an avatar with images like food on it. So she would have this stuff. I can see Firoz, thank for putting that stuff in. So she she needs to put some of these images up on the wall. She also may pull these images in from food, which we'll use in another session to find restaurants that work with everybody. And the goal would be, could she? Could she use an AI to more quickly produce what she has to do? So, speaking for Vanessa, what she has to do is grab an image, make a minor resize of some of it, maybe change perspective of it. And it was when she appears at another session I'll ask her specifically about. Is she doing any post production work with it, with an with an AI, or with a tool like Photoshop or some image editing so she may modify the image slightly, just to make sure there's continuity. Image Size are the same and maybe the perspective is the same, and then it would be handy for her to know when she would go into and look for new images. So behind the scenes, we're developing a workflow, and I'm hoping that she does that, creates a workflow which simply gives her an email saying there are two new images inside the o2 images folder, or inside the o2 images Kuntz matrix folder, and then she would know, right, it's time to look at that stuff and hang it. I don't know when it's going to be possible for her to use an AI to actually hang things on the wall of that exhibitation, that exhibition space called the Kunst matrix. But I'll let her explain the next time we're together about how she might accelerate what she's doing. And do appreciate folks that put your images up there from last time when you weren't here, because she will work those images into the space called Kunst and she shows later on the call or ask her to run around the exhibition space, she'll show you that. So last week, I had a problem with the connectivity I was using to talk to the internet. So you didn't have to go through that pain with me, Maria last week, but not every. I don't know if everyone saw the Financial Times video, and it is worth watching. It's approximately 10 minutes long, because it does put an accent point next to something which is a big concern of mine. So I would have worked for decades with young people as they progress into the jobs market, and then work with recent graduates as they moved up one level of expertise in the job that they earned after they earned a university degree. And now there's some real pressure at the lower end of the jobs market across the world, because some companies are saying, we spent this money for these AI tools. So maybe we can make let the middle level managers, instead of hiring an assistant, a person, maybe we can train them to use an AI, and that will be them, their their virtual personal assistant. Now this is happening, and it is creating a tension of sorts, for young graduates to get on the first rung, the ladder of employment. And one of the workarounds for that is to be able to show to a prospective employer that you as a 20 something have the skills that the person in their mid 30s, early 40s would have claimed to have because they've just been upskilled with AI. So that's hopefully what this program is doing. It's helping somebody positioning into the jobs market to say, I have these skills, newly learned under a program that has academic credentials, and I should be able to enter the jobs market as a young graduate could. So anyway, the Financial Times investigates this issue. What they've done is they've made a short video, and what I wanted to do is +1
play the thing again for folks and then use Perplexity after the fact to fact check it. So the journalists talk about how some jobs are disappearing because they're automated now there's economic uncertainty in the markets, constraining revenue companies have, so they're leveraging the expertise inside without having to go outside to to find a graduate. And I'll play this like it's about 14 minutes long. I'm pretty sure I keyed in the sound, so if this does not play for sound. Let me know it's the job buckets, as explained Speaker 1
21:04by, entry level jobs are vanishing at an alarming rate.
Speaker 2
21:09It's likely to be one of the most challenging times in history for graduates to get a job. Today, the
Speaker 1
21:14perfect storm of AI political instability and economic uncertainty are causing what some are calling the Jobpocalypse.
Speaker 3
21:23If you don't hire people right now, the company will cease to exist in 15 years.
Speaker 1
21:28If 
the first rung of the job ladder is disappearing, what does that mean for the future of work? I'm Isobel Barrick. I lead the FTS, working at Rand, speaking, presenting and writing about management, leadership, workplaces. In this series, I'll explore some of the most pressing issues around the future of work, and talk to senior leaders about how they are making work better. Human beings are wired to resist change for everyone, the the
+1
the first rung of the job ladder is disappearing, what does that mean for the future of work? I'm Isobel Barrick. I lead the FTS, working at Rand, speaking, presenting and writing about management, leadership, workplaces. In this series, I'll explore some of the most pressing issues around the future of work, and talk to senior leaders about how they are making work better. Human beings are wired to resist change for everyone, the the Speaker 1
22:07job 
openings for graduates are at an all time low. Job listings in both the US and UK are plummeting, and for the first time, on record, levels of joblessness amongst graduates is above the overall unemployment rate. When I met with industry experts a year ago, we talked about the big problem facing recruiters, the skills shortage and the AI arms race, changing the way people applied for roles. A lot has changed in this period, and not for the better. The issues with recruitment are still there, but what has changed is the rise in big graduate employers really cutting the number of jobs they're offering.
+1
openings for graduates are at an all time low. Job listings in both the US and UK are plummeting, and for the first time, on record, levels of joblessness amongst graduates is above the overall unemployment rate. When I met with industry experts a year ago, we talked about the big problem facing recruiters, the skills shortage and the AI arms race, changing the way people applied for roles. A lot has changed in this period, and not for the better. The issues with recruitment are still there, but what has changed is the rise in big graduate employers really cutting the number of jobs they're offering. Speaker 4
22:52There's some data from indeed, which is one of the big job platforms that says that the number of job vacancies for new graduates has dropped about 33% about a third just over the past year. So that just means more people, because we know that more and more people are going to university, chasing fewer and fewer jobs. Some people talk about this idea of a diamond shaped structure, where you might have your top people, your CEOs, at the top, but then a lot of people in the middle, but not many people at the bottom. I struggle to figure out how that's going to work, because there's the very obvious problem that anyone can think of, that, where are the people in the middle going to come from, if you're only hiring a few at the bottom,
Speaker 1
23:29Chris Eldridge, chief executive at recruitment firm, Robert Walters, met me at the FT Video Studio. So what are you seeing in the market for graduates at MD? What's the big picture.
Speaker 2
23:40I think if you zoom out, it is, it's likely to be one of the most challenging times in history for graduates to get a job. Today, the one consistent factor is the economy. I think companies want to hire when they've got some foresight into the economy. And today, it's a really challenging time to be able to forecast what's going to happen next? We see that then in hiring numbers, companies just don't want to risk their capital right now, and therefore they're, they're just pausing the the graduate intake. But that pause really has now lasted for two, two and a half years.
Speaker 1
24:14So what's the knock on effect of a two or three year pause in early career hiring? You know, further up in workforce development, I think
Speaker 2
24:23what we'll end up doing is we'll end up finding, in two or three years time, a kind of void of talent, if you're not bringing in new skills. Now, what you tend to do is hang on to the existing staff you've got, and then you promote them. You potentially over promote them. I think particularly when you look at the US market, what they're looking for when they say is that kind of new applicants to the market, recent graduates. They really they don't mean someone's coming out of university today. What they're really looking for is somebody with one or two years experience.
Speaker 1
24:58I went to Paris to meet Xiaomi glider, co founder and CEO of AI powered recruitment platform. Welcome to the jungle, which attracts large numbers of new graduates looking for their first jobs.
Speaker 3
25:11Just on our platform since 2024 I think you've seen like a very sharp decline on entry level jobs like posted by companies. And even more specifically, you see that this decline has been like, three times faster and higher than other positions for more senior people. So basically, what does it mean is that companies are really like, waiting to see if like, what? What is impact of AI? Does it mean that we don't have to recruit like entry level positions, and most importantly, what are we going to do with more senior roles? So a lot of people, in fact, are waiting, and because companies are waiting, you see this decline going faster and
Speaker 1
25:52faster. Do you think that employers realize that they have this kind of if we say social duty, maybe it's more than that. You know, imperative to train people from the start, otherwise they'll have no pipeline. Or are they just focused on the money?
Speaker 3
26:06I think it's hard for companies to think about like social duty without like putting their money problems first, but I think they know that it's going to be a big, long term problems. Just look, for instance, at two specific industries, which are like accounting and law firms. Because probably with AI, we can do a lot of stuff, like maybe 80% of what junior lawyers do. But the problem is, is also, for instance, the average age in this law firm is 55 and so if you don't hire people right now, the company will cease to exist in 15 years. I don't think AI will replace everything. And anyway, we don't know. And so the problem is that we don't know who is making a mistake.
Speaker 1
26:55The graduate jobs crisis is much more nuanced than just being down to AI, but there are companies that have gone all in on artificial intelligence and have cut jobs, and the results have been varied.
Speaker 4
27:09We're in the very early days of figuring out how robust this technology is in the real world of the workplace and outside of a few specialized things, and so I think it's quite possible that we'll see some companies getting carried away and laying off swathes of their staff and saying they'll replace them with AI agents, and then discovering that it's actually much more difficult than they realize to make this work in a way that's reliable, accurate, cyber, secure, and so you might See some wrong turns and missteps along the way.
Speaker 1
27:42And 
legal profession is a sector that's likely to be radically reshaped by AI, certainly in terms of cutting the work typically done by graduate recruits. I met with Julian Taylor, senior partner of international law firm Simmons, and Simmons, who have fully committed to an AI future. So looking ahead, what kind of role will this kind of AI tool play in future? We're just at the very beginning in the legal
+1
legal profession is a sector that's likely to be radically reshaped by AI, certainly in terms of cutting the work typically done by graduate recruits. I met with Julian Taylor, senior partner of international law firm Simmons, and Simmons, who have fully committed to an AI future. So looking ahead, what kind of role will this kind of AI tool play in future? We're just at the very beginning in the legal Speaker 5
28:08profession. I think it's going to do a huge amount of some of the more commoditized work that takes place at the moment, the more data oriented, the analytical work, administrative. Vast amount of that will be done by AI. I've got no doubt about that, three, five years from now, I think things will look quite different.
Speaker 1
28:27Will it change the career structure of a lawyer? Will it change the org chart? Yeah.
Speaker 5
28:33I mean, some organizations have been coming out and saying, Well, they're going to hire fewer, particularly, you know, graduate lawyers. I think that's just really dangerous. I think if you if you start cutting your numbers, you need to think about what's your pipeline like for the future. And if we get too narrow at the entry point, I think that will create problems further down the line, it's much easier to take your talent from your own trainees will need to think differently about how we train our people, but I think it's really important to retain that pipeline.
Speaker 1
29:08So we've talked about the structure of how a junior lawyer would work. At the moment, with it, there's a lot of repetition, and what probably in the past would be called drudge work, if that's going or is there's much less of it. What happens next for them in the new era,
Speaker 5
29:22we'll need to find ways, if AI's doing that, of ensuring that they still learn. You learn from maybe the third or fourth time that you do that stuff, you're not still learning in quite the same way after the 20th or 25th time that you're doing it. I think there's a real opportunity to move our trainees, who are highly smart, highly skilled, kind of further up the value chain in terms of the work that they're doing. I think a lot of the time in big law firms, more junior lawyers and trainees are protected for quite a long time from that higher value work.
Speaker 1
29:53In 2023 Simmons and Simmons launched their in house Gen AI tool called Percy.
Speaker 6
30:01It's a retrieval, augmented generation engine, and it's a way to process documents with a particular technique in the background to make them more accessible.
Speaker 1
30:11The tool assists their employees to crunch through masses of documents and make sure they cover all the important points in their decision making.
Speaker 7
30:19So we've asked it to help us draft some interview questions or potential questions that we might want to kind of get more information about the complaints he's raised here. You can see that Percy has kind of spat out some questions by theme.
Speaker 1
30:33Do you think you've lost anything by not having to read through reams and reams of documents in the same way?
Speaker 7
30:39No, I don't think so. I mean, I'm still reading all the documents. I'll have read every single line of all the interview notes before I've produced any outcome, because, again, you're never going to rely 100% on what Percy tells you. So I've still done that. But then what it's allowed me to do is kind of just ensure that the output that I'm creating is just that bit more accurate. So there's
Speaker 1
31:03been some research that shows that humans tend to cognitively offload when we're using AI, we don't engage with it in quite at quite the same intensity as when we're processing stuff that we're reading and writing about. How do you get past that? Because you have to, as
Speaker 7
31:19lawyers, where you might see the cognitive offload happen is where people kind of shortcut their prompts, and naturally what you get back is also not as good. Here you can see that there's times been taken out to actually produce the prompt, and that actually requires you to actively engage in what the output that you want is. So until you do that, you're then not actually going to get what you want from from the AI.
Speaker 1
31:45So your prompting skills are improving along with your legal skills. Yeah, absolutely. Being adept at writing prompts is currently key to using AI effectively in law firms and elsewhere. It requires us to communicate complex ideas effectively and clearly, and will be a skill most of us need for the future. Expensive talent laws still continue in the legal sector, but could the lure of AI training and job augmentation be what sways the best talent
Speaker 5
32:20you've got to be comfortable using AI. I've moved my messaging a bit internally when I'm doing town halls, from trying to reassure people to kind of, I think people have to be a little bit scared, because I think jobs will disappear. We don't quite know how things will be in a few years time. It's all changing pretty rapidly, so they need to be able to be flexible with whatever's taking place.
Speaker 1
32:44As 
employees, we have to accept that our bosses are in a period of experimentation, and we're part of it. Some of them are going all in on AI at the expense of jobs, and some of them are maintaining hiring, especially at the junior level. But for students and graduates, it raises the question of whether University is even worth it anymore. My feeling is yes, if you've got a degree from a top university, that's always going to start your career off on the right foot, but overall, the education system, both in schools and universities, is not equipping graduates and school leavers with the skills and also the critical thinking that they'll need for a future proof career in the workplaces of tomorrow.
+1
employees, we have to accept that our bosses are in a period of experimentation, and we're part of it. Some of them are going all in on AI at the expense of jobs, and some of them are maintaining hiring, especially at the junior level. But for students and graduates, it raises the question of whether University is even worth it anymore. My feeling is yes, if you've got a degree from a top university, that's always going to start your career off on the right foot, but overall, the education system, both in schools and universities, is not equipping graduates and school leavers with the skills and also the critical thinking that they'll need for a future proof career in the workplaces of tomorrow. Speaker 4
33:30So 
if you look back sort of hundreds of years to medieval times, one of the ways in which young people learned their skills was that they would be apprenticed to a master, and actually often, their middle class families would save up their money and pay that master craftsman to take their young person on. And you know, I have talked to some people who think that maybe that's going to be a model for the future, because if a big company can't really monetize the skills of their more junior people, rather than not hire them at all, maybe, maybe they'll be willing to take them on in exchange for some sort of fee. The alternative would be, you know, a return of kind of indentured labor, which also happened with apprenticeships, whereby, you know, a company can say, Well, look, we'll take you on and train you, even though we can't really sell your skills just yet. But in return, you know, once you're fully trained, you're going to need to stick with us for 10 years or whatever. It's interesting to see how these sort of super modern problems are having people reaching into the history books for potential solutions.
+2
if you look back sort of hundreds of years to medieval times, one of the ways in which young people learned their skills was that they would be apprenticed to a master, and actually often, their middle class families would save up their money and pay that master craftsman to take their young person on. And you know, I have talked to some people who think that maybe that's going to be a model for the future, because if a big company can't really monetize the skills of their more junior people, rather than not hire them at all, maybe, maybe they'll be willing to take them on in exchange for some sort of fee. The alternative would be, you know, a return of kind of indentured labor, which also happened with apprenticeships, whereby, you know, a company can say, Well, look, we'll take you on and train you, even though we can't really sell your skills just yet. But in return, you know, once you're fully trained, you're going to need to stick with us for 10 years or whatever. It's interesting to see how these sort of super modern problems are having people reaching into the history books for potential solutions. Speaker 1
34:31AI evangelists are saying that the skills that will be needed for entry level jobs in future are changing fast, and some of them are even guessing what skills are going to be needed for jobs that don't exist yet, but in the meantime, we're cutting those entry level jobs and the drudgery that goes with it. But are we throwing out something valuable when we do that? Those entry level jobs create the pipeline for the future, and without it, we may all be adrift. You.
Bernard Goldbach
35:03Way, the purpose of showing you that is to kind of examine what Perplexity would say as as a as a means of fact checking what's going on there. So I think what I might have let me see if I've done this while the video was running, I was trying to have a look at what, what was happening in another space. So I went to, I went to, I went to Perplexity and asked the question a couple different questions, and then asked fabric to save it and we'll see if I can share that with you.
Bernard Goldbach
35:48So I typed with my phone, so there's some misspelling, but I asked it like, what major successes have companies enjoyed by using AI agents? The goal is to kind of fact check the story that the Financial Times have has by using Perplexity that the search engine and it'll even even the free version will will find the word Job, Job pockets, job Puck, pocket lips. I think is what they lips, hey, I think it's the job. Pocket lips, yeah. Poco lips, yeah. If you use that term and you ask Perplexity to find holes in the argument and what some of the things, if you took notes, you'd know that the journalist is trying to find an alarming trend, which is companies are using AI to find higher levels of productivity at a lower cost. And one of the significant lower costs is hiring somebody in Europe, where you have obligations to pay them if they're let go, and you have social insurance obligations or training obligations in the way the company structure exists. So it's not just hiring somebody and providing them a wage. There's these other incidental things with training and social insurance and provisions that are normally in the mix unless, of course, they can, a company can hire a freelancer to do the job. So there's several things about pressing workplace issues, the capability of what an AI can do, the track record of agentic or AI agents. And I want to try to see, can you power up an instance of perplexity.com and ask a couple of questions, and when you do, put put the prompt you're using into the into the Zoom chat. So for example, one of the things that I think I asked was, you know, what does the AI arms race mean, for recent university graduates. So that's, that's a prompt that all I'll use in Perplexity, and I'll do it live, as I'm hoping that you folks can do. So I'll share my screen. I'll go back on over to a browser. I'll run the browser. I'll run my browser in Perplexity, and my my screen says that it's pause. I'll stop sharing again and try to share. Try to get the demonstration screen in place so it's showing now the sharing, and now the first thing that's popping up is it's saying, hey, Comet, download that and 
use it. I'm just going to show you another video where comet is going to be something to say, download. I'll download it, and it tells me how to install it. So I'll let it, it is downloaded, and I'll do that later. But comment is an agent, and I can run the agent on the back of my Perplexity and it can can do things. So now I think I have an account with Google. If not, I'll just sign in as myself as with Google account perplexity.ai I actually typed in perplexity.com but perplexity.ai where I am, and I'll use a run that same prompt. You know, what does the AI arms race mean? It's telling me, if I'm a student, I can verify and get free pro so if you folks are on this service and you have a anyone, any of the EU are run, eu addresses should work for a free pro account and student features to verify, I had to click my verify now and then. Have to go over to a student URL. But this is showing me that I have a pro account already from a previous login, and now it's saying that the AR arms race is doing the following things. I'll do a summary here and put it in so I can share this with that URL, and I'm going to pop that URL into the Zoom chat, so theoretically anybody could go to that URL and see the result that I got back. Okay, so I want folks, I want you for the sake of running this task, fire up an instance of Perplexity, put a relevant prompt into your Perplexity as a means of fact checking what you just heard from the Financial Times, or trying to Find more information about an item that you heard mentioned in a previous 14 minutes that raised your concern or that you want to have verified. If you're being cantankerous, then you can put something in the prompt that takes issue with the reporting standard or the authenticity of the job recruiters or whatever. So you throw that in. Now, if you're well read, you may know there are some holes in the research that was presented by the Financial Times. Not big problems, but there are always, there are often more than one side, there's more than one side to an argument or a premise. So the Financial Times is going down the pathway of saying there are going to be entry level job problems exacerbated by the rise of AI. And they went out to prove that point. If you want to take the opposite track, then you'd be looking for proof that the jobs market hasn't been affected by this, or that AI is actually on the way to a gigantic explosion of failure. So a few prompts into the Zoom chat, please use Perplexity to get an answer. If, when you get an answer, you can share the URL. That'd be, that would be, that would be brilliant. Not everyone's going to be able to share the URL, in my experience, but see what we can do. And I want to learn from the prompts you put in. Later on in this session, I'll discuss how to maybe make a better prompt, or how to make a really better follow up prompt, but at this point, we're just going to bang in a couple of things. What was your question, though, summarize, summarize the video. Okay, all right, did it find the video? Or did you use the URL of the video?
+2
use it. I'm just going to show you another video where comet is going to be something to say, download. I'll download it, and it tells me how to install it. So I'll let it, it is downloaded, and I'll do that later. But comment is an agent, and I can run the agent on the back of my Perplexity and it can can do things. So now I think I have an account with Google. If not, I'll just sign in as myself as with Google account perplexity.ai I actually typed in perplexity.com but perplexity.ai where I am, and I'll use a run that same prompt. You know, what does the AI arms race mean? It's telling me, if I'm a student, I can verify and get free pro so if you folks are on this service and you have a anyone, any of the EU are run, eu addresses should work for a free pro account and student features to verify, I had to click my verify now and then. Have to go over to a student URL. But this is showing me that I have a pro account already from a previous login, and now it's saying that the AR arms race is doing the following things. I'll do a summary here and put it in so I can share this with that URL, and I'm going to pop that URL into the Zoom chat, so theoretically anybody could go to that URL and see the result that I got back. Okay, so I want folks, I want you for the sake of running this task, fire up an instance of Perplexity, put a relevant prompt into your Perplexity as a means of fact checking what you just heard from the Financial Times, or trying to Find more information about an item that you heard mentioned in a previous 14 minutes that raised your concern or that you want to have verified. If you're being cantankerous, then you can put something in the prompt that takes issue with the reporting standard or the authenticity of the job recruiters or whatever. So you throw that in. Now, if you're well read, you may know there are some holes in the research that was presented by the Financial Times. Not big problems, but there are always, there are often more than one side, there's more than one side to an argument or a premise. So the Financial Times is going down the pathway of saying there are going to be entry level job problems exacerbated by the rise of AI. And they went out to prove that point. If you want to take the opposite track, then you'd be looking for proof that the jobs market hasn't been affected by this, or that AI is actually on the way to a gigantic explosion of failure. So a few prompts into the Zoom chat, please use Perplexity to get an answer. If, when you get an answer, you can share the URL. That'd be, that would be, that would be brilliant. Not everyone's going to be able to share the URL, in my experience, but see what we can do. And I want to learn from the prompts you put in. Later on in this session, I'll discuss how to maybe make a better prompt, or how to make a really better follow up prompt, but at this point, we're just going to bang in a couple of things. What was your question, though, summarize, summarize the video. Okay, all right, did it find the video? Or did you use the URL of the video? Maria O'Brien
42:29Maria, no, it found the video in the first place. And then the next prompt I asked, what were the social implications? Okay, the long term social implications of a diamond shaped workforce, and it tagged it on, onto that original URL.
Bernard Goldbach
42:46Okay, so you did a follow up prompt, yeah. Okay, all right, good. So I'm going to ask another question. I
Bernard Goldbach
43:08go in English. I'm going to ask Perplexity to see whether it can actually go back in history. So I'm going to be asking, is it the most is this the most challenging time for graduate to be hired? If anyone else could put up a prompt to be handy. That'd be nice to know. So I'm gonna, I'm asking it. I could have used the microphone option to ask the question. So in my case, that found a lot of information across the world, 28 different sources. It's looking at the current job market, unemployment rates, the challenges, and it identifies the impact of AI and automation as a second challenge, the first one being stalling hiring plans because Unknown market pressures so and Perplexity
hiramb
43:59says that
Bernard Goldbach
44:02says what we thought might be the case. So
hiramb
44:13what I did is I asked Perplexity if it can compare the job apocalypse, got it
Bernard Goldbach
44:23right first time. No, you got it right.
hiramb
44:27If this, if this applies as well for Flanders I got already from chat GPT, that situation is slightly different. We do see, we usually see some changes, but not that as drastically as what the Financial Times reports. Perplexity also indicates that this is typical US and UK.
Almudena
44:50Apparently.
Bernard Goldbach
44:51Ah, so the market trends in the EU are different. Do you think is that what your chat GPT suggesting that it's different than other parts of the market.
hiramb
45:03Well, yes, and, and likewise this Perplexity, it's it says you should be able to look at the prompt, at the prompt, and at the result, yeah. It says, While finders does not face a wholesale job or collapse apocalypse, identical to the US or the UK. The region does show clear science of Flanders Show does show clear signs of stress and job job market for young people and graduates. I think there is also a difference in what graduate means in the UK and what we see as graduates, because at the moment, we have since 2020 an increase of, I think I've written in the chat, 75% more graduate students, and even the numbers show that we can expect an increase of 40% compared to last academic year.
Bernard Goldbach
46:01Yes, yeah, that's true. Maddy, aren't you in the jobs market within the next two years? Because if so, what's your what's your prompt, What's your concern? Can you answer that? Are you?
Maddy
46:14Yes, I will probably be in the job market in the next two years. Actually have to do an internship next year.
Maria O'Brien
46:21Good, Oh, yeah.
Maddy
46:24Could you repeat the second part of the question?
Bernard Goldbach
46:26No, no, but then, then, would you mind asking an AI to confirm your concerns? Say, look, I'm entering the job market next two years. What kind of do you? Would you anticipate exceptional pressure for in the industry against hiring a new graduate. I mean, because that's what Hiram saying is it may be more accentuated, more dismal in America than in the European Union, but on the other side of the ocean here in Europe, there are, there definitely is a higher number of students progressing from a standard honors degree, a four year degree into a master's degree immediately after graduation. So theoretically, what that means is higher number of graduates. I mean more people with grad actual graduate degrees, not just university degrees, which may compound the problem, you know? Yeah, I just want to, I don't want to create stress for people in the call. I'm just want to try to use the power of AI Perplexity AI to cross check the accuracy of the reporting and the documentation offered by the Financial Times in their story, the story is getting repeated in different versions in other elements of the trade press that I read. So anyway, Perplexity is a tool. So does anyone want to add to this? Any other people that are on the call want to, you know, add any comments to this. If not, I can, I can move on to to a to another part of the this session that I wanted to show you.
hiramb
48:19So what can I ask you a question before you move
Bernard Goldbach
48:22Hiram, yes.
hiramb
48:24From Perplexity. My question would be, is there any reason why you would use Perplexity for any kind of job more more particular than any other
Bernard Goldbach
48:38originally, Hiram, the service offered the best way of sourcing information. At the very top, it gave you, here are the sources I'm looking at, and it would and it exposed small text strings telling you how it was thinking where it was going. So originally, before Google Gemini was unleashed, Perplexity had the best, most transparent way of showing both an AI's LLM answer as well as the current information that had just trawled from online, trustworthy news sources. That's why there? There are other add ons to Perplexity, for example, that this comet agentic search capability is something that is clever. I have another video queued up that I will play only part of because of the interest several people on this call have expressed for the uses of AI and and, but that answers the question for you, I think, is where
hiramb
49:47Thank you. Okay,
Bernard Goldbach
49:48so now there's a woman, Grace Leong, who is, I believe, a Chinese American, and her, her language register isn't the same as the people on this call. So there are two specific parts of the 16 minute video, which I want to point out. So I will play the video for a few pay a few minutes. One I'll go into that's two minutes in talking about a research agent, because there are at least three people on this call involved in research. And she'll explain how, when you install comet, then you get the research agent and what it can do. And nine minutes in, she talks about the AI's search agent. I won't show that, but to say that it is a very clever way of using Perplexity on top of Google search to improve the focus of the search. There are other things in the video. It's broken down by chapters one talks about automated business processes. So I know that would be of concern of Vanessa's she's previously expressed 15 minutes in there are privacy considerations, which you may want to consider when looking at any kind of an AI. So I'll play forward just a few of the minutes of this, because the video will be in a link that I will attach to this deck when I revise the deck. So I'll start this video and I'll scrub forward to two minutes. She talks about Comet
Speaker 8
51:40Perplexity AI's no longer just an AI search engine. It's now an AI agent. Perplexity comments the AI browser not just find information, but really executes the task for you.
Bernard Goldbach
51:52And then she talks about the research agent. Part of
Speaker 8
51:58it, one we all know the pain of research, dozens of clutter tapes and information chaos. So I will show you how commerce can turn that mouse into an organized workspace, letting the agent read everything at once to deliver the powerful insights. So usually when we do research, we have so many tabs open and they look very cluttered. So a cool way is to ask Perplexity to get relevant tapes into what it calls a dedicated workspace as it can understand the context for all the open tapes, or you can just ask it to assign any open tapes in the relevant workspace so you can make everything so organized the agent is
Bernard Goldbach
52:34actually looking at Your entire browser with all the confusion and memory usage and just a lot of information that it's able to parse through. It's a decent video. I'll go forward to her issues of privacy concerns, which are very relevant to everybody here.
Speaker 8
52:59MD, 
tasks you can ask Perplexity agent to do for you. Now, I know privacy is a big hesitation when it comes to Perplexity AI browser. I think first, we all need to understand that for an AI browser to be useful and effective, it needs context from a browsing and that means it has to access to certain extent of the data. Though, I can see that Perplexity Commerce does collect more data by default, but there are ways you can minimize it.
+1
tasks you can ask Perplexity agent to do for you. Now, I know privacy is a big hesitation when it comes to Perplexity AI browser. I think first, we all need to understand that for an AI browser to be useful and effective, it needs context from a browsing and that means it has to access to certain extent of the data. Though, I can see that Perplexity Commerce does collect more data by default, but there are ways you can minimize it. Bernard Goldbach
53:25So what I'll let you do is watch the video to figure out how you if you're going to use comet as your virtual assistant, which I think everyone ought to experiment with, but if you're using any AI agent on a free version, not a paid enterprise version for work. Just know that what you allow that agent to see, that would be your emails, your calendars, your browser history, your search history, what you allow it to look at builds a really interesting profile of self, or profile your company, or a very specific demographic, the kind of work you're doing for the task you're attempting. So I don't mean to get people paranoid in just saying to have it working well, just assume the agent is like a person who knows what's inside your purse, knows where you park your car, knows what you had for dinner last night, and what kind of coffee you prefer 10 o'clock or 11 o'clock in the morning. It it can be unnerving when this happens. Said another way there are high end spyware companies paid by the Trump administration to build these profiles on people now using the back end the back door technologies wherever you browse and wherever you look. So just be concerned about that. When, before the program started, I was recommending people use the little signal app, because that can't be probed, can't be looked at, because that's a very trustworthy, highly encrypted program. Anyway, Grace has great film, and that video does talk about what comet does, how it can help you as a marketing student or as a business analyst, as somebody who can manage an agent as a service that can manage your workflow during the day, taking practical steps you allow it to 
do with the permissions you give, but there could be issues as you do it. So I'm not I'm just saying, be aware. This is one of the things that happens when you activate a service, and then you let the service so first you can ramp it or run unhinged on your work. Okay, so I was we did this. We asked Perplexity of how to do the work. We shared your prompts in the Zoom chat, and maybe did a few other things. Okay, some things about prompting. I'll give you 10 prompting techniques worthwhile. And next year, next week, I'll give you some other perspectives on it. If you're going to prompt, you need to tell it what you want, identify clearly. You need to summarize something, generate ideas, solve a problem, if you offer a very precise goal that's going to give the AI clear purpose, and it's foundational to prompt effectiveness. Be clear about what you want. You create a practical workflow for me getting a job in the era of AI with a company that is not hiring new graduates at the moment, help me specify my skills that can ensure I get approached 
for an interview. Second thing prompt frameworks are important, so I'll talk to you in a later lesson about how to set a goal. Provide context to the AI. How to source where you want it, to source the material and the expectation that you have for the result. Another acronym is a star, the situation you have, the task you have, the task you need to accomplish, the action you want to activate, and the result that you hope so if you have a framework for the prompt engine to analyze, for the LLM to execute, that reduces a lot of the complexity and simply allows layered and conditional logic to work on what you're asking. Number three, be specific, be descriptive. So if you need to have text that you're going to use to generate images, if you're in your and you're vague, you're going to get unpredictable results if you're clear and detailed with the descriptions, especially by offering descriptions that define what you want, the style in which you want it, and The context that you want to have in the result, the text result, the image result, or the video result, you're going to get significantly better outputs. Be specific, be descriptive. I rarely stop at the first prompt. I iterate by offering multiple prompts to an AI. It's rarely a one shot process before I get a result that I trust. So it's like I talked to someone I just met. I might be at an Irish pub, going back and forth, prompting, talking to them, finding out what kind of work they do. I rarely stop at the first question, hey, what do you do? If the person says, I'm a solar panel installer, I'll start asking that solar panel installer about what kind of panels, what kind of wiring, what kind of roof layout, what kind of inverters to use, because I want to refine my understanding about what that person is same way talking to an AI, you're going to follow up with a prompt, look at the initial outputs, and then refine where you want the output to go. This is iterative. It's conversational. It allows me to zero in on what I really want to get from the AI without necessarily knowing when I started what I need, so I make successive adjustments which should improve the accuracy of 
the information and the relevance of what the AI tells me iterate with multiple prompts. This is one reason why I asked people to maintain a prompts notebook so it can see that you actually did iterate. You weren't helping. Happy with the first result, yes, the follow up, and maybe use a different AI for the follow up. I specify what kind of result I want, how I want it styled, bullet points, list, table, paragraph, academic research. There's academicians vote and Sharma say, if you highlight the kind of structure that you want back from the AI, you're going to get better formatting, clear result, and potentially the kind of thing you can copy and paste into a form for an abstract for a funding mechanism, or for like bullet points on an agenda for a meeting, or some kind of a format, like an academic abstract that you need for professional paper. Number six, give the AI a role and a scene for context. So I tell it a chat. GPT pretend you're the marketing manager. You need to get a brief for a Christmas exhibition in a multi purpose venue that's going to attract the tour operators and casual visitors. So it's thinking now as a marketing manager, it's going to contextualize. It's going to listen to the contextualization I gave in the prompt to allow this thing called conversational pedagogy, which means more meaningful interaction with the AI. Set the role marketing manager, set the context where they're going to have the result, apply the information that the AI comes back with role scene context number seven,
+2
do with the permissions you give, but there could be issues as you do it. So I'm not I'm just saying, be aware. This is one of the things that happens when you activate a service, and then you let the service so first you can ramp it or run unhinged on your work. Okay, so I was we did this. We asked Perplexity of how to do the work. We shared your prompts in the Zoom chat, and maybe did a few other things. Okay, some things about prompting. I'll give you 10 prompting techniques worthwhile. And next year, next week, I'll give you some other perspectives on it. If you're going to prompt, you need to tell it what you want, identify clearly. You need to summarize something, generate ideas, solve a problem, if you offer a very precise goal that's going to give the AI clear purpose, and it's foundational to prompt effectiveness. Be clear about what you want. You create a practical workflow for me getting a job in the era of AI with a company that is not hiring new graduates at the moment, help me specify my skills that can ensure I get approached +1
for an interview. Second thing prompt frameworks are important, so I'll talk to you in a later lesson about how to set a goal. Provide context to the AI. How to source where you want it, to source the material and the expectation that you have for the result. Another acronym is a star, the situation you have, the task you have, the task you need to accomplish, the action you want to activate, and the result that you hope so if you have a framework for the prompt engine to analyze, for the LLM to execute, that reduces a lot of the complexity and simply allows layered and conditional logic to work on what you're asking. Number three, be specific, be descriptive. So if you need to have text that you're going to use to generate images, if you're in your and you're vague, you're going to get unpredictable results if you're clear and detailed with the descriptions, especially by offering descriptions that define what you want, the style in which you want it, and The context that you want to have in the result, the text result, the image result, or the video result, you're going to get significantly better outputs. Be specific, be descriptive. I rarely stop at the first prompt. I iterate by offering multiple prompts to an AI. It's rarely a one shot process before I get a result that I trust. So it's like I talked to someone I just met. I might be at an Irish pub, going back and forth, prompting, talking to them, finding out what kind of work they do. I rarely stop at the first question, hey, what do you do? If the person says, I'm a solar panel installer, I'll start asking that solar panel installer about what kind of panels, what kind of wiring, what kind of roof layout, what kind of inverters to use, because I want to refine my understanding about what that person is same way talking to an AI, you're going to follow up with a prompt, look at the initial outputs, and then refine where you want the output to go. This is iterative. It's conversational. It allows me to zero in on what I really want to get from the AI without necessarily knowing when I started what I need, so I make successive adjustments which should improve the accuracy of +1
the information and the relevance of what the AI tells me iterate with multiple prompts. This is one reason why I asked people to maintain a prompts notebook so it can see that you actually did iterate. You weren't helping. Happy with the first result, yes, the follow up, and maybe use a different AI for the follow up. I specify what kind of result I want, how I want it styled, bullet points, list, table, paragraph, academic research. There's academicians vote and Sharma say, if you highlight the kind of structure that you want back from the AI, you're going to get better formatting, clear result, and potentially the kind of thing you can copy and paste into a form for an abstract for a funding mechanism, or for like bullet points on an agenda for a meeting, or some kind of a format, like an academic abstract that you need for professional paper. Number six, give the AI a role and a scene for context. So I tell it a chat. GPT pretend you're the marketing manager. You need to get a brief for a Christmas exhibition in a multi purpose venue that's going to attract the tour operators and casual visitors. So it's thinking now as a marketing manager, it's going to contextualize. It's going to listen to the contextualization I gave in the prompt to allow this thing called conversational pedagogy, which means more meaningful interaction with the AI. Set the role marketing manager, set the context where they're going to have the result, apply the information that the AI comes back with role scene context number seven, Bernard Goldbach
01:01:24just 
be aware of the fact that the response you get may not be most current thing. That's okay. You can use another AI Gemini and Perplexity normally gives you back stuff that a simple chat GPT prompt won't come back with a chat GPT has gotten a lot better. Copilot is not not as good as chat GPD for relevant stuff, but is relevant, but not necessarily for yesterday's news. If you can think by that you want to get a browser and an LLM combined to get an answer, then just know that the information you get back might need another piece of technology like Grace Leung was saying, may need comet to look inside of the tab structure you have for follow up information, to get the most current responses and references, just so just know that what you get back may not be as good as reading yesterday's news or financial times. It may be two weeks old, month old, a year old, depending upon where it got. The information you need to check the sources, and the AI should provide you the sources. So chat GPT will invent responses. Chat GPT may make up answers you're responsible for the information your AI comes back with. So if you're given something forward as a 50,000 euro answer, and you haven't cross checked it against another large language model or Google Scholar or some other mechanism, you're the one that's going to feel the heat. I have been misled by information that chat GPT said, that I wrote and published. I have seen chat GPT tell me things I did not say, because it's just the way the AI works. It starts to fill in the blanks in strange ways. Just know you're you are responsible for checking 
the sources, and something that comes back seems really outlandish, cross check it number nine. Don't overload the poor, AI. So if you're deep into a conversation, by that, I mean more than like 10 prompts trying to follow up and get information, you may confuse the model. So what will happen is something that Brandon Robinson from the university Adelaide, point out, that too complex of a prompts sequence starts to cause either hallucinations or this affirmation mechanism, which the AI starts talking to you as, oh, you're, you're the God, and you know so much, and you're really good, it's it's sycophantic. So just know that if you overload the AI with too much of a of info, too much prompting, you could result in getting stupid answers coming back. And finally, experimental studies have shown that, you know, participants can write prompts Well, it's they have learned how to do this. Well they end up, they end up being someone whose skills, prompt engineering skills, through practice, exposure and iterative learning, learn lead them well down the pathway of future employment and better results. So this is learnable prompt engineering. It's a skill. We'll go beyond prompt 
engineering in this particular course. Okay, what I want to try to do is use Gemini together, doing a couple of things we're going to run right up against. We'll go take this all the way up to 715 here's what I want to do. So the first task is this, use, use Google Gemini to look at the information that's inside of Google Drive. Okay, so it means you need to know where the Google Drive is. I'll put it into the chat prompt, and you simply use Gemini to summarize the material and then share one thing that you that you discovered. We'll do a second thing together on this prompting tool, I'll pop the Google Gemini is URL into the Google Chat. In case you don't know where that easily, you can't easily find Google Drive. Okay? Firoz, thanks for that input you had on the chat. So that's the Google Drive. And what I want to do is I'll show you see if I can share my screen, I can show you me going into it. So go into Google Drive, you'll see me sharing the screen to Google Drive. The top of mind, it says, Ask Gemini. So Gemini can summarize the folder. I'm just going to tap it summarize this folder. This folder. Quick tip, I can ask it about anything. It says, it says it contains 90 files, mostly images, folders, PDFs, and it tells me all the stuff that's in it, and a whole bunch of other stuff. Okay, so you don't have to use that. You don't have to use that same prompt. I'm interested in knowing. I just asked Bernie asked.
+1
be aware of the fact that the response you get may not be most current thing. That's okay. You can use another AI Gemini and Perplexity normally gives you back stuff that a simple chat GPT prompt won't come back with a chat GPT has gotten a lot better. Copilot is not not as good as chat GPD for relevant stuff, but is relevant, but not necessarily for yesterday's news. If you can think by that you want to get a browser and an LLM combined to get an answer, then just know that the information you get back might need another piece of technology like Grace Leung was saying, may need comet to look inside of the tab structure you have for follow up information, to get the most current responses and references, just so just know that what you get back may not be as good as reading yesterday's news or financial times. It may be two weeks old, month old, a year old, depending upon where it got. The information you need to check the sources, and the AI should provide you the sources. So chat GPT will invent responses. Chat GPT may make up answers you're responsible for the information your AI comes back with. So if you're given something forward as a 50,000 euro answer, and you haven't cross checked it against another large language model or Google Scholar or some other mechanism, you're the one that's going to feel the heat. I have been misled by information that chat GPT said, that I wrote and published. I have seen chat GPT tell me things I did not say, because it's just the way the AI works. It starts to fill in the blanks in strange ways. Just know you're you are responsible for checking +1
the sources, and something that comes back seems really outlandish, cross check it number nine. Don't overload the poor, AI. So if you're deep into a conversation, by that, I mean more than like 10 prompts trying to follow up and get information, you may confuse the model. So what will happen is something that Brandon Robinson from the university Adelaide, point out, that too complex of a prompts sequence starts to cause either hallucinations or this affirmation mechanism, which the AI starts talking to you as, oh, you're, you're the God, and you know so much, and you're really good, it's it's sycophantic. So just know that if you overload the AI with too much of a of info, too much prompting, you could result in getting stupid answers coming back. And finally, experimental studies have shown that, you know, participants can write prompts Well, it's they have learned how to do this. Well they end up, they end up being someone whose skills, prompt engineering skills, through practice, exposure and iterative learning, learn lead them well down the pathway of future employment and better results. So this is learnable prompt engineering. It's a skill. We'll go beyond prompt +2
engineering in this particular course. Okay, what I want to try to do is use Gemini together, doing a couple of things we're going to run right up against. We'll go take this all the way up to 715 here's what I want to do. So the first task is this, use, use Google Gemini to look at the information that's inside of Google Drive. Okay, so it means you need to know where the Google Drive is. I'll put it into the chat prompt, and you simply use Gemini to summarize the material and then share one thing that you that you discovered. We'll do a second thing together on this prompting tool, I'll pop the Google Gemini is URL into the Google Chat. In case you don't know where that easily, you can't easily find Google Drive. Okay? Firoz, thanks for that input you had on the chat. So that's the Google Drive. And what I want to do is I'll show you see if I can share my screen, I can show you me going into it. So go into Google Drive, you'll see me sharing the screen to Google Drive. The top of mind, it says, Ask Gemini. So Gemini can summarize the folder. I'm just going to tap it summarize this folder. This folder. Quick tip, I can ask it about anything. It says, it says it contains 90 files, mostly images, folders, PDFs, and it tells me all the stuff that's in it, and a whole bunch of other stuff. Okay, so you don't have to use that. You don't have to use that same prompt. I'm interested in knowing. I just asked Bernie asked. Bernard Goldbach
01:06:29So I asked, summarize folder, okay, I put a bunch of need to have different people put stuff in that Zoom. Chat about what you asked Gemini to do. It could be something like, is there anything in here about Maria O'Brien? Stick your name in and see if your name comes out, or, you know, and if it does, it's okay. What's it? Know about me, and see what happens, but see how good Gemini is. If you want to use it on your phone, if your phone has Google Gemini on it, you can, you can go to the prompt in the browser, and as Gemini live, summarize this folder, and it'll talk back to you about the folder, but otherwise, just put the prompt in there that you're asking Gemini. And then I like people to tell me, Hey, here's what came back. This is a surprise, or here's what came back. I think it's wrong. I'd like to find that Gemini didn't discover something accurate, but it actually made something up that has not happened to me yet. Chat GPD has made stuff up, but Gemini hasn't summarized the folder. With that skill, you're going to be able to accomplish a lot of other things on this module. So in Portugal and on this in a meeting in October, I'll have a complex skill that you're meant to accomplish. I wouldn't be surprised that somebody on this call within five minutes can answer a very complex set of questions just by letting Gemini answer the questions by looking at the material all down to your prompting skills. So any of you folks have something you want to put in the chat that you've asked
Almudena
01:08:11Gemini, excuse me, in my last word, don't apply. Don't either can't see you can't get
Bernard Goldbach
01:08:22Almedia, can you give me? Give me your Google mail address again? Can you put that in the chat and
Almudena
01:08:29see I gave you in the chat?
Bernard Goldbach
01:08:31You can, yeah, I'll stick it in the object. I'll stick it into the Google Drive, and then, yeah, see, I thought I put that in, sorry. What I'll do is I'll go, I'll tell you how she can. I'll show you what I'm doing as I do this.
Almudena
01:08:49But now in the I can see the folder, Rune the folder, but I can see the use to use Jiminy.
Bernard Goldbach
01:09:01Okay, okay, let me. Let me just try to fix this here.
Bernard Goldbach
01:09:12Darn
Bernard Goldbach
01:09:18way 
too many folders, as you can tell,
+1
too many folders, as you can tell, Bernard Goldbach
01:09:45okay, now, how do you know, if you log into your browser as your Gmail, then your Gemini should work now, and if anyone else on this call has that same problem, just let me know, and I'll add you to that drive with the Gmail identity. It'll work if you have a Yahoo identity, a Hotmail identity, it'll something that Google has registered as a as an email that is allowed to use the Google services. So ubu.es isn't letting that happen. I think standard might not let it happen either. So yeah, so you should be able to see it now with generally, and so just stick some kind of a prompt in there. Leti, do you? Did you ask it for anything at all? Or you did somebody,
Letizia Cenni
01:10:32I had a problem could be the same. Maybe I'm a dinner. I had to go into the My Workspace settings. And enable, what was the name, some kind of smart connection between Gemini and others. Maybe that's your problem.
Bernard Goldbach
01:10:53What browser are you using? Chrome? In my case, yeah, on the top right of Chrome, there's these identities and profiles that I have, so just got to toggle the identity for Gmail, and normally I can, I can use the Google services that way. The other thing some people have done is they've used an incognito browser and seem to sniff in. I don't know if that would work for for you, Leti, but I want to see, can you use previously on a course similar to the one that you're on, students have been able to simply right click on their mouse and ask questions, and it's brilliant, how fast they get information. It's just amazing. And I want to make sure that people can see the drive and can ask questions. Now it may be that using that shareable link, and I haven't tried this using the shareable link, so when you go into the Google Drive and I think the link is, yeah, it's further up here on my Zoom chat, just using that link inside of Perplexity might do the same thing, but I'm not sure. So what I should do is grab that link and see if I can do that, because I'm trying to figure out a way of having this work with other services. So I'm going to go into, well here Perplexity installed on my system, so I'm going to go, and probably can't, I'll just do it.
Bernard Goldbach
01:12:35So I'm going to try to use the browser. I think you can see the browsers. I'm trying to do it here.
Bernard Goldbach
01:12:57It's behind a login. Okay, let's see. Can I change the
Unknown Speaker
01:13:04service?
Bernard Goldbach
01:13:18Anybody can view it. Anybody Copy Link, try it again.
Bernard Goldbach
01:13:35Doesn't see I'm there to get in that way, but it's not funny me, so Perplexity is not being let in either. So okay, so Gemini, if Gemini cannot see it, because you can't get in with Gemini. Okay, Maria, did it answer that question
Maria O'Brien
01:13:55for you. I got in with Gemini, and I just tried it in Perplexity, and it said there's no directly accessible content from Google Drive link. I did do it, but Perplexity didn't
Bernard Goldbach
01:14:08right. So did it find your name?
Maria O'Brien
01:14:11Yep, yep, okay,
Bernard Goldbach
01:14:13derogatory or inflammatory? No.
Maria O'Brien
01:14:17It was the opposite. No, okay. It summarized. It said I was a participant in the running EU AI session. And during the session, I discussed my my goals related to the program,
Maddy
01:14:31okay,
Maria O'Brien
01:14:31outlining that I was interested in sustainability, but not an expert in the area, and I need to utilize AI to help implement certain measures. So it kind of summarized what I had spoken about on week one.
Bernard Goldbach
01:14:47Yeah. All right, I'll try one other thing here. Share my screen to too many browsers open. Do this?
Bernard Goldbach
01:15:13Try one other thing, because I'm now in a space which I did not have, having a problem which we did not have previously. So let's see if I can.
hiramb
01:15:29Nope, it's not the park screen.
Bernard Goldbach
01:15:35Trying 
to, okay, let's go back and forth here, trying to share into to now Sharing in Microsoft Edge. I'm going to go to copilot.
+1
to, okay, let's go back and forth here, trying to share into to now Sharing in Microsoft Edge. I'm going to go to copilot. Bernard Goldbach
01:16:03Going to summary this page. All right, so now I'm in copilot, which is finding this Okay, so now I'm going to ask another thing here, find information in this location, about
Bernard Goldbach
01:16:32Okay, so it's asking me to use Google Drive to find Maria O'Brien. I
Bernard Goldbach
01:16:45I ask and just send you. Okay, so it's limited, and hence, the reason I'm asking about using Gemini is because I know Gemini, when you have permissions to get in, Gemini can do it. Okay, we're beyond the time I want to go to. So I want to stop and then rejoin and ask a follow on. So if you decide what you want to do is cheat and go to the next thing I want to ask, we'll go back in and we'll do a second task, which is using the same tools run inside the thing called the thicker digital transformation group and ask it this the same kind of kind of questions you would ask it, you know, it's a tell it, it's URL, and summarize what's in it, and then document some support material that's that's in that flicker digital transformation group. But we'll do that when we get back together in about 15 minutes time. So I have a few questions, a few slides at the end of this deck, which I said I'm going to modify to make sure the links are there, and we'll open up the Trello boards, and I need to explain to you what you're seeing there inside those Trello boards. I'll stay on this call for another few minutes if anyone has questions. Otherwise, just plan to come back in the same zoom URL in about 15 minutes time. Okay, and so you don't find it in Gemini. Thanks, Firoz. Okay. What I'll do is I'll add that to the I'll add your URL to the to the shared space in Google Drive. Firoz. If it works, it works. If it doesn't, okay. We're learning there's a limit. So I'll stay on this call. If anyone else has any questions or whatever. Just shout them to me. I'm adding Firoz to the
hiramb
01:18:47space.
Bernard Goldbach
01:18:55Okay, almedina, are you? Okay?
Almudena
01:18:59Yes, I'm Gemini.
Bernard Goldbach
01:19:03Yeah, Gemini, okay, it's okay. Here's where I go. I'll put it in a chat. If you I run it as an app in my service. I downloaded the thing that but you can go to Gemini
Almudena
01:19:13inside the drive. I can see Gemini. I ask and don't understand before.
Bernard Goldbach
01:19:21Can you can you see the URL I put in zoom chat, gemini.google.com
Maddy
01:19:27What
Bernard Goldbach
01:19:29in zoom in the Zoom chat I put in the URL, gemini.google.com can you see that?
Almudena
01:19:38Excuse me? Oh, yes.
Bernard Goldbach
01:19:42Okay. Going to do is I'm going to go there. I'm going to share my screen as I go there, I think. So I'll go to gemini.google.com, and now a minute, I'll say,
Bernard Goldbach
01:20:10need you to Please summarize.
Almudena
01:20:15Oh, okay, can you do it now? I
Bernard Goldbach
01:20:27you. I'm all over the place here.
Almudena
01:20:34Okay,
Bernard Goldbach
01:20:41that's probably too many windows, so we'll just try this one here, please. I don't know whether this is going to work inside of the chrome bro, the browser called for Microsoft, called Edge Omn.
Bernard Goldbach
01:21:09So I'm in Edge, the screen should be showing edge, asking to summarize the URL, see if it does.
Bernard Goldbach
01:21:26Gemini is looking at the question. It's analyzing it. I should summarize it, gemini.google.com, should do this now.
hiramb
01:21:39Okay, so
Bernard Goldbach
01:21:47So I'm asking, what's it know about Almudena?
Bernard Goldbach
01:21:59Okay, so let's see in all the work you've done makes it a lot easier for me to grade you say she's got an image I gave you compliments about your image and the behavior and so how about that? So I went to Gemini to overcome sorted It's another tool. Okay, so what's happening is you can in another browser use you have another google.com, and you should be in the browser called Edge, Microsoft, your institution might let you use edge, as long as you can see gemini.google.com, then you're into the AI through a browser. You give it a place to look that is allowed under Google's rules for people to view now you can see it. Okay, I tried, you try it. No, this is like, this is a, yeah, yeah. So now if, if you're, you're, I don't know in Google, I don't know in Microsoft, edge, how to make it think that I am big Goldbach achievement.com because it edge knows me as different names, but somehow the Edge browser is letting me in anyway. Let's experiment with it. We'll see other people on this call have the same problem. So I know Firoz I just gave him access with his Gmail, let's see when he comes back on the call, whether he can do this I'm going to end this call, start another one. Okay, so we have two sessions that are recorded. Okay, about 10 minutes. Okay, okay.