Episode transcript The transcript is generated automatically by Podscribe, Sonix, Otter and other electronic transcription services.
Hello, everyone; welcome to the Five Minutes Podcast! Today, I like to continue the podcast, the series about artificial intelligence in projects. If you go back to last week, I spent most of the time of the podcast talking about what is AI, the uses of AI and the benefits that you can help you to reduce your admin tasks, help you to forecast budget and schedule, understand how you can fit the right team that will, for example, reduce the risks, improve your budget precision, improve your schedule. I spoke about risk identification in that you can speak or talk or write to a chatbot and have your risk register populated based on some of your perceptions about the risks. However, it's not that easy. There are many challenges, and as I finished my last podcast, I said data. Data is king. One of the biggest challenges we face in our projects is the lack of proper data. First, you have too little data to analyze. You have three projects in your historical information, and you want to use AI. It's not like that. You need to have thousands and thousands and thousands, if possible, millions and millions and millions for you to have data that you can really, in a strong way, make a clear declaration that that pattern is relevant for budget or that pattern is relevant; for the team, etc. The second is you don't have reliable data, and this is a very big difference between, for example, trying to create an AI to play go and trying to create an AI to manage a project.
Go or chase their games. They are very complex games with endless possibilities. But if you explain the rules and the computer plays against itself, millions and millions of times at some point, the game learn how to play. What is the best move based on all these scenarios? But the rules are very clear, right? Very clear. The Queen in a chessboard can move in such a specific way, but in our project, you know, do you know what happens in our project? The Queen comes to you and says, I resign because now I was hired for another chessboard and the Queen just leave, and you say, but in my game, this is a real project. So it's far more complex for you to model that because there are millions of variables that you need to be in place. I know that you will say to me that a self-driven car is the same. Yes, but there are, for example, driving rules. There are some very specific and the eyes, for example, cameras in this that will help you to do some basic functions. But in a project, the environment is so complex, is so creative. That is very hard for you to have an AI project manager or an AI scrum master. You can have an AI to do admin tasks or to help you with forecasts, or to give you advice on what is the best composition of your team but to manage.
It's very, very hard, and this drives me to the second challenge. The first one is about data. The second one is patterns for human behaviour are very hard to identify. It's so hard to identify that John had a poor performance today because he's not motivated because when he was walking to work, it started raining, and he fell, and he hurt his knees. How did it feel? The parameters are pretty much endless, and human behaviour like motivation, like leadership, like conflict, like power, they are not rules by simple rules. It's far more chaotic. So how do you plan to map that? How do you plan to get this information? Do you plan to build this database by doing what? By going and interviewing each individual that works for you and trying to understand their motivation every single day until you have 10 million data? This is the difference between first and you managing a project and, for example, Facebook, because Facebook, what is Facebook doing? It collects every single information, every like or dislike, every message, every single connection you have. In overtime and billions, billions, I'm not even talking millions, Facebook starts to predict your behaviour. For example, you like a cat; then I send another photo of a cat. Then a friend of yours sends a photo of a cat.
Then I know that you may like cats because this is your behaviour, so you map these over millions. Imagine if you were able to create a Facebook for all the projects on Earth. Oh, then it would be very different. But inside your company, this is very, very challenging because finding these patents are extremely hard work. And last but not least, it doesn't matter too much how strong you develop your AI mechanism. It's still very hard for machines to modulate and to model ethical behaviour and ethical aspects. I'm not talking about the law because even the law, the law, it's not 100 percent clear because if the law was 100 percent clear, you don't need to have a judge; you just need a spreadsheet. And based on the answers on the spreadsheet, I would say, Oh, you are condemned to 10 years in jail. Why do you have all the proceedings and all the judgment? Because it's not obvious it's a human activity to interpret that ethical aspects are the same. Imagine I was watching an interview in a conversation between Michael Sanders from Harvard Law School and Yuval Harari from Sapiens. The book and talking about, I would say, all the trends today and these ethical dilemmas are extremely hard for us humans because of cultural aspects. Imagine for a machine, imagine you in a self-driving car that will decide if you should turn and hit someone and kill someone, or if you should go straight and hit a tree and kill three people.
You know, this is their sense of what is called utilitarianism. So how do you model a computer to do that? How do you model a computer, for example, to pilot a plane and leave the plane to start crashing to decide which house I should crash in house A or in house B? What would be the criteria? And imagine that you are listening to this podcast is an inhabitant of one of these houses. So these ethical aspects, I'm not saying this would never happen. No 15 years ago. I never thought that a car would be self-driven. These were science fiction movies. Maybe in the future, we will be able to model machines that will be very similar to us. But there is still a good way to go, and these are big challenges. So every time you see the opportunity to apply AI, you just need to understand that all the benefits they face are these challenges, and it's not very easy to overcome them. But maybe in the future, we will. And what is important for all of us is that we, as professionals, we need to stay tuned for what is happening with AI. What is the potential of AI? Because I'm pretty sure that these will play a massive role in the way we deliver and manage projects in the future. I hope you enjoy this podcast and see you next week with another Five Minutes podcast.