When Risto Siilasmaa realized that Artificial Intelligence (AI) would transform the technology industry, he decided he had to learn what it was all about. Now the chair of Nokia’s board of directors is making sure all 100,000 of the company’s employees understand AI, Machine Learning (ML) and big data. In an interview in Nokia’s seaside villa in Helsinki, Mr Siilasmaa explained why Companies who fail to adapt to AI will be left behind.
Why did you become interested in AI and ML technologies?
I’m an engineer by background and I’ve followed new tech since I was a kid of 12 years old, and the fascination of intelligent computers has never left me from the days I read science fiction books. Now that we’ve started seeing real life examples of things that machines can do better than the best human experts, that of course triggered one’s imagination. It’s becoming obvious that any Business will draw much of its competitiveness from ML technologies in the future, and of course, I should understand what this means for the companies I work with.
Why should company leaders understand these technologies, rather than leave it to data scientists?
If I talk to an audience of CEOs, I often start by asking which ones feel that, in five years’ time, ML will be a critical piece of their competitiveness and all of them will raise their hands. Then I ask how many of them really understand how ML works and maybe one, two, three per cent of them raise their hands. This is exactly where I was. I believed a few years back that ML would be a key source of our competitive advantage, but I didn’t understand why, and I didn’t understand how it works. So, I could not ask the right questions when people came and talked about what we were working on.
If it is so strategically important for the company, I should understand and we all should understand, at least enough to ask the right questions, so that led me to a sort of wake-up moment that I don’t have to wait for others to explain this to me, I can actually move my butt and go back to school myself.
What is your advice for company leaders in the position you were in?
The problem with many leaders like myself is that we get used to people explaining things for us, we sort of delegate learning to others. Then we just get the gist of it from a very concise summary, but that doesn’t transfer real learning and understanding to us.
We should wake up from that paralysis that others do our thinking and learning for us, and then when there is something truly critical, we should feel that we can go back to school. Of course, we can get really top teachers who can condense the essentials for us, but we should ask them to truly go deep enough so that we can understand how this technology works. This is an attitude that I like to see in the companies I work with, throughout the company. It’s also an attitude of being brave enough to admit that I don’t understand something and there’s nothing wrong with it. There’s something wrong with claiming to understand something that I truly don’t. That can be it’s dangerous. It may lead us to making the wrong conclusions, so let’s just admit that we’re all learners, we’re all eternal students, and there’s nothing wrong with asking stupid questions.
What parts of the business is it changing?
We are in the process of getting transformed. We have a large number of ML projects underway in our internal functions, for improving the quality of our work and augmenting our people so that they become their better selves. For external purposes, we are adding new competitiveness and new functionality to our products and services.
We have also launched a program to educate all 100,000 Nokia employees, who will go through simple ML training, just like a code of conduct program, that is mandatory for everyone. We believe that our employees appreciate the fact we want them to learn. It’s important for them to be at the top of their profession and to understand broadly what’s happening, and it’s important for us that they develop themselves as human beings, that they know their expertise is appreciated, and that we’re investing in their development across the board.
Do you have any predictions for AI and ML in 2019?
There is of course research being done on ML widely, but regardless of the new findings and inventions the big bang will come from the already existing technology being applied widely. The technology itself is not very complicated, so companies broadly speaking will be experimenting with the datasets that they have, looking for new ways of using that data and getting productivity and so forth from that data.
But they may also at times find something uniquely valuable and surprising from the data, in the same way that we have been playing chess for thousands of years and really smart people have been writing books about chess, and not only have we lost to computers for the last decade but now we have had to acknowledge that our understanding of chess strategy has been flawed, it’s like there is another continent on earth that we didn’t know about and ML found it for us. This same thing can happen to companies as they start doing this work, they may realize something uniquely valuable that they were never even thinking about.
What do companies need to do to adapt successfully?
Adapting ML widely takes time because you need to educate a lot of people and you need to take a different approach to how you think about business problems. If you want to be an AI player, one of
the knee-jerk reactions that you must have is to acquire data. If you have a problem, the first thought is where do I get the data that I need to solve this problem, and then you buy data, you buy companies for the data that they have access to, you may buy datasets or do R&D work that you give away to people for free so that you get data in return. Then you need to reorganise, to structure your business and your organization in such a way that this tool can be effectively used, and this is a long transition, it’s not easy. I’m not saying that every company should become an AI company—not all can—but the ones who want to be need to think deeply about it.
Is it important that companies buy in external datasets rather than just use their own data?
It is important that companies think strategically about data, both the data that they have or have access to, but especially about data that they can foresee needing in a few years’ time.
Why is data valuable and what sort of datasets have the most value?
Well that depends on the business of the company, but data is the food that most ML needs, that’s the way they are trained. In simplest terms the way ML systems work is that you have a certain set of training data that you use for the training and then you have weights in the system that reflect what data is meaningful in what context. In the training process, those weights are adjusted so that the mistakes the system makes are minimized. You do that training many times, perhaps thousands of times or tens of thousands of times, and each time the mistakes that the system makes are getting smaller and smaller, so the real value is in the trained weights. If you have bad data you will get bad weights, and the system will make mistakes; it can’t answer questions correctly. If you have high quality data that fits the problem that you are trying to solve, you may get excellent results, far beyond human capabilities in these narrow fields of problem solving.
How can firms ensure the data they use is high quality and relevant?
They need experience to do that; it’s not easy. The problem is that we all have biases and sometimes we don’t understand our biases. Some can be pretty simple—when universities build systems, if the researchers are all white Caucasian males, then they may forget that there are other types of people. They don’t have any intention to skew the system so that it doesn’t deal fairly with people of color, it just happens. They may only find out after they launch the system for public use, and then it’s a big PR crisis. So of course you can measure the quality of data in many ways—mechanically, you test it—but then there may be deeper level thoughts or missing subsets of data that you only realize way afterwards, so you really need to approach it very thoughtfully and it’s another thing that people need to train for, it’s not something we automatically can do well. This is a new frontier; there are lots of things to learn and get adjusted to.
What external datasets are most important to a company like Nokia?
That purely depends on the business and the problem. Let’s say if we want to automate accounting then we need data of accounting sentries and of course all companies have lots of that data because they do accounting themselves. It’s only a question of if they want to use it and to automate accounting themselves or if they use a third party who is specialized in that and builds the systems and helps them to use it. Then there are some things that only one particular company does, and they have access to that type of data and then they will have to do something themselves, they can’t just resort to third parties. Sometimes the problem is when the company doesn’t have access to suitable data and they just have to figure out where to get that.
So how would you summarize the benefit to companies of investing in AI?
I think all companies probably start with attempting to maintain at least their current competitive advantage because everybody is investing in machine learning, especially in the tech space, so you have to run to stay in your current place. But of course, if we are better at this, if we are more innovative, if we actually come up with something new or apply it in an area that others don’t, our products will be cheaper, faster, better quality, they will make fewer mistakes, they will be more intuitive, cheaper to build and cheaper to operate. Those are the opportunities and advantages. In addition, there may be some things that would be completely impossible without machine learning and those are allegorical to situations like AI finding new chess opening strategies or Alpha Go making moves that no human being has ever played, or curing illnesses that were never curable before.
