Demystifying AI or: How I learned to stop worrying and love the LLM | Henri Terho

Hello, everybody. I'm going to talk a little bit about demystifying AI - and what it actually is, and with the nice Dr Strangelove quote, - or how I learned to stop worrying and love the LLM. So, everybody of us is talking about AI and LLMs and much of that. So, I'm going to open up that box a little bit. Warning, there will be a little bit of math, - and there will be a test at the end, of course, as is for any university lecture, you have to have that. So, a little bit of what we're going to do. Who am I? A little bit of AI, a little bit about the types of AI slash machine learning - and how do we actually work, and what does this mean to all of you? Going into who am I, I'm actually a biologist. I'm not an IT guy. I jumped from the field of biology to IT - because I was sitting in a closet in the university doing data cleaning, - doing data processing and all of this stuff and deciding that - nobody's ever going to pay me enough to do this whole data janitor stuff. And here I am, still doing the same stuff 15 years later, - basically on IT sector, but not just doing the biology side. I worked a lot with the AI tools and doing stuff around AI and testing. And I think that's one key point that happens with AI now - is because we can generate so much code, - testing is actually going to become the number one thing, - that has been for a long time the second thing, - and the first one that we typically take savings from, - hey, let's not test as much, but now because it's so easy - to generate code, that's going to happen a lot. And now we can get to the AI part of it as well. AI everywhere. Everybody's talking about it. Every single moment. How did we get there? How is everything going to change? I didn't think everything's going to change. This change has already been going on for the last 60 years, maybe. And I think this is one of the points that I'm trying to make to prove that. This is pre-packaged food from Finland. It's leverlåda, so it's liver casserole, pretty much. This was my first AI project that I did 15 years ago - on building AI-optimised recipes - for creating a liver casserole at scale in these huge factories that exist. Fifteen years ago. And exactly the same kinds of problems with neural nets, - optimising recipes and all of that. So, this is not new. This has been going on for a long time, but now it feels like it's all exploding just right now. And as I said, neural networks are 70 years old. We invented all of this technology 70 years ago. Why is it happening now? And we're using this technology that is really old, - and everything's changing in a really fast pace right now. And I think it's just because in the base of it, it's just math. How these things work, it's not really anything super complex - or let's say, complicated, - but it gets complex because of the size of these things. So, because of all of this evolution that's been happening - with all of the math and vectors, that's all that is. It's just doing matrix multiplication. AI does nothing else than matrix multiplication. That's it. And for some reason, that gives us really good and powerful results - on predicting text, predicting recipes, - predicting a lot of other things, just doing this. And this is, I think, one of the key points that I want to make, - it's just math, it just works. But you have to understand that it's just math. It doesn't do any kind of reasoning, it doesn't think, - It doesn't do anything like that. And with these caveats, the reason why we think it's now exploding - is because we, as humans, are really bad at forecasting exponential change. And most of the stuff that's happening in the society is exponential. But when we think about, oh, when is AI going to change our way of life? Okay, I think in linear way, maybe in 10 years. When is fusion power going to come? In the next 15 years. When is something happening? It's always in the linear way that we are thinking about it. But suddenly, it explodes, and we get all the results.