- For a long time, computer vision represented the cutting edge of artificial intelligence research. Now, the advent of large language models has set the ball in motion and multiplied the number of AI researchers.
- There’s fierce competition for researchers, and universities are rarely able to match the offers made by big companies. Universities can’t afford to spend as much money on researchers and computing power as the tech giants.
- That’s not to say universities have nothing to offer. It’s just that companies, especially in Finland, don’t really know how to fully maximise the benefits of scientific research, says Joni Kämäräinen, Professor of Signal Processing at Tampere University. He urges European states to pool their resources to drive AI development and to make sure European data stays in Europe.
You’ve been studying computer vision for a long time, but now you’re more interested in robotics. Why the change of heart?
Before the deep learning tsunami, there were small, 5–10-person laboratories around the world and a few thousand AI researchers in total, compared to the hundreds of thousands of today. Suddenly we are competing with giants like Google who have 200 people from top universities working on things like object recognition.
An annoying pattern started to emerge – whenever we came up with a great idea for research, it didn’t take more than a few weeks before a student came to us, saying “Google has already published a paper on this topic.” Instead of universities, big companies are now the leading research facilities.
So, rather than focusing on computer vision, I started to think about AI and how most of its achievements only exist in the virtual realm. It is unable to do things in the physical world. If we could create a robot that uses artificial intelligence, it could do good in the physical world. That said, no doubt someone will invent evil uses as well. But the ultimate dream of AI research is to create a robot that helps humans.
What are the main issues you are attempting to resolve?
In robotics, there are two big problems, or mysteries: how to make a robot navigate autonomously, and how to make it pick up and use tools. Our focus is on navigation. The tests we are currently conducting in the Hervanta area involve a human initially showing a robot which route to take, after which the robot is able to take the same route and navigate around obstacles. Our robots bear some resemblance to the food delivery robots that grocery stores use, but ours are a little more autonomous. An essential difference is that instead of GPS tracking we use image-based navigation. It works amazingly well, and the process has reignited my childlike curiosity and passion for research.
Artificial intelligence can write poems and essays, but when it comes to operating tools, machines are hopelessly clumsy. In the past, the purpose of automation was specifically to reduce manual labour. Where do you think this development will lead?
It was generally considered that artistic and creative jobs would be the last ones to be taken over by artificial intelligence. It was envisaged that the first jobs to go would be those that require the use of hands and feet. But it’s beginning to look like the opposite is true: it will be easier to replace doctors than practical nurses who provide physical care to patients. Not many could see this coming.
When you write scientific reports, your subheadings often contain references to works like The Lord of the Rings, and to popular Finnish artists such as Matti and Teppo and Ismo Alanko. Do you think artificial intelligence will ever be able to create anything like that?
I believe it will. It may be hard for us humans to get our head around the fact that AI can generate something so seemingly creative. In the not-too-distant future a person sitting at home in, say, Hervanta, will be able to create a Hollywood film with all the special effects, simply by using artificial intelligence. While this is not pure artificial intelligence, AI has been used to remove multiple bottlenecks that previously would have cost a passionate filmmaker about a hundred million euros. Pretty soon artificial intelligence will write scripts, decide which camera shot angle and colour scheme to use, and so on. This is inevitable. But the problem is that AI can produce an obscene amount of material. If we talk about music, it can generate something like a billion songs a day. Who will listen to them?
You’ve said that companies of all kinds should adopt artificial intelligence. What would AI adoption mean for a Finnish SME?
It may not be necessary for a viable company to adopt AI, but if there’s a risk of being steamrolled by others, it’s better to be one of the early adopters. That does not mean companies have to develop their own language models; they can buy a licence to use a commercial version.
The first thing Finnish companies should understand is that they store a large amount of data that is valuable either to their own or for someone else’s business. China and the US are much savvier when it comes to storing existing customer data for future purposes and for creating better products. In short, the methods enabled by AI are available to everyone, but the one who holds the data is the winner. This is something Finnish companies have not quite understood. When I see what’s going on in the research centres of Finnish companies, it looks pretty amateurish to me.
What do you mean by amateurish?
Well, let’s say a big company has read an article in the newspaper about AI being a big thing, then they contact us and say “hey, this is big, we want to do something, too”. When we ask them what it is they want to do, they have no idea. What they could say is “You are the lab we want to partner with because we make big moving machinery that relies on computer vision. We want to support dissertations that interest us, we want to hire your graduates, and in exchange have some of our scientists come and work for you, as a sabbatical of sorts. And we want to go all in on this collaboration.”
A couple of million euros a year is peanuts for a big company, compared to all the nonsense they spend much more money on. But for some reason companies struggle to invest heavily in a Finnish university or research institute. Don’t the chief technology officers of Finnish companies have any decision-making power? Big international companies are not afraid of bold initiatives. Here, companies ask if a government organisation like Business Finland could finance everything — don’t they have any money of their own?
What can Finnish companies learn from the US and China?
American companies hire young people who are passionate about their work. They know that when a person is really fired up about something, it makes sense to have them run their own company and give them freedom, and hire team leaders who share their passion. They also know that once they channel some of that zeal and energy towards their own business, they will most certainly stand to gain something. Finland is not very good at this, unlike countries such as China, where the ethos of hard work is steeped in the culture. People in Western countries need to be motivated, they need good leadership.
In traditional industries, let’s say in a Finnish company that builds big machines, future success will be increasingly linked to software and artificial intelligence, but most of them don’t have departments focused on software and AI. I mean, with all due respect, people in the traditional automation industry tend to be clean-cut, very disciplined people who show up for work no later than at seven in the morning. I can almost see them saluting their supervisors. That doesn’t work with creative software people; they would laugh their heads off. Software developers operate in a creative world, and there’s no point in trying to control their chaos, because chaos is creative. The more a company needs software, the more it needs to change its traditional corporate culture.
Right now Finnish universities are on tenterhooks as they wait for the government’s decision on how it plans to allocate the money earmarked for research and development between universities and the business sector. How do you think the money should be allocated?
I would invest strongly in doctoral studies and in ensuring that graduates end up working in the R&D departments of companies. Writing a dissertation is a bit like reaching for the stars, and for anyone working in R&D, that experience is very useful. They know how difficult it is, but they have seen people pull it off. I would also support research centres operated jointly by companies and universities. These facilities would train doctors, and companies would offer financial incentives to research topics of special interest to them. This R&D funding could help Finnish companies learn the ways of the world, or in other words learn to have faith in work that might seem chaotic. That’s what you have to do if you want to stay in the game.
Tampere University runs the Doctoral School of Industry Innovations, DSII. Would that be a good example of collaboration between businesses and universities?
It is, in my opinion, quite possibly the best research-related idea ever developed in Tampere. If I was a research director in a big company, I would use DSII to take over the best laboratories at Tampere University. Then I would say I’ll pay for four dissertations a year, as long as you focus on topics that are important to our company. It hasn’t dawned on companies yet that this would be an effective way to outsource their R&D activities. What’s more, it’s not the research we do here that counts; it’s the people we train. Companies that have hired our top talent get more than value for their money: they get a much easier recruitment process and a highly skilled employee for four years. This should become an established system of training the workforce for the industrial sector.
What else can Finnish universities offer in the global AI race?
Robotics is an area full of promise and opportunities, which Finland would be wise to seize. When we talk about AI, there’s a huge gap between creating the software and building an actual device. We train very few people who understand devices. Perhaps we should take a leap of faith and start a new discipline called robotics.
When I first brought this up a while back and searched on Google for places offering such a discipline, the only hits were two Japanese universities. Now more and more of them are popping up at a steady rate, including one at the Technical University of Munich. This is an area where Finland could, for once, lead the way. If we focus exclusively on software development, even the poorest countries will be our competitors, because almost anyone can buy a laptop, even in a poor country. But if you need to build a robot or an earthmoving machine and figure out how to make them autonomous, the number of competitors drops dramatically.
You’ve said that Europe has made too many “sweet” and unselfish decisions. What do you mean by that?
An example of a sweet decision is the general data protection regulation, GDPR, which has in no way stopped data from falling into the hands of big corporations that transfer data outside Europe. The idea looked nice on paper, but the only thing that has actually been achieved is creating a nuisance for end users. Sometimes we need to think business first. By this I mean we need to build businesses in Europe in such a way that we can ensure not a single European person’s data is transferred outside Europe.
We really should make a decision to create a European language model that will beat the American and Chinese language models hands down. And this is by no means a fantasy. In this increasingly weird world, Europe is becoming an attractive place to work for talented people from North and South America, Africa and Asia. This is a pretty nice place to live – in fact it’s the nicest place in the world at the moment. Sometimes people turn this against us and say “You guys have so much fun that you never get anything done.” I would turn it the other way around: we get all kinds of interesting things done because we have so much fun. We need to have more faith in ourselves and in what we do, and pool our resources more.
How can Finland get its piece of the AI pie? Should we start developing our own language model?
No. We have nowhere near enough money for that. Training the first language model cost about a million euros, the next one ten million euros, and with the current ones we are probably talking about a hundred million euros. Setting up an adequate research organisation would cost between 200 and 300 million euros, and the Finnish government has typically invested millions in research infrastructure, or, in a spate of generosity, tens of millions. The European Union is the minimum scale. In terms of GDP, the EU can compete against China and the United States, but no individual member state can. Sadly, the growing nationalist movement is only playing into the hands of the big players. If we fail to understand that to be a winner in this game we need to think from the European and not the Finnish perspective, then we are simply helping bigger countries reap their profits at our expense.
What will happen to Europe if we fail to pull together?
The more heavily the data economy starts to dominate economic activity and money flows, the more money we will lose to China and the United States, if nothing changes. If we allow these countries to keep us on a tight leash, the piece of the pie available to Europe will become smaller and smaller. What I perceive as a threat is that one day we will buy services that we no longer understand. It would be inconceivable – just think about a situation where we bought a nuclear power plant but had no clue how to operate it.
There are naturally many ethical problems, too, but if we want to sell the idea of stronger European integration to European countries, we must focus on the economy: the risk of money flowing away from Europe and the impoverishment of European countries. In my opinion, in a nation of five million people, protectionism and isolation will inevitably lead to destruction. The first to suffer from nationalist politics will be Finland’s key tech companies and universities, in other words the very places where future success and prosperity is created.
Finnish text by Tuomo Tamminen
Translation by Leni Vapaavuori and Nick Moon