We wanted to understand what’s happening in Artificial Intelligence (AI) and machine learning and set out on a journey. In a two parts blog we take a look at what AI is, it’s effects, and then explore the state of AI in the Nordics. This is Part 2, about the Nordics. Welcome!
In our previous blog we took a deep dive and explored AI and what’s going on in the AI-world, and particularly in the US. We then continued our AI-journey in the Nordics, and have been a bit surprised by what we found, and did not find. So, is there intelligence in the Nordics? Yes, but what about the artificial kind? The short answer is yes, too, but various AI initiatives are scattered all over the place, often under different headlines than “AI” and there’s no clear agenda and only partial interest from universities, established enterprises or politicians. However, AI is booming in the startup community, and among students.
We wanted to answer a few questions. What’s happening in the Nordic academic community, in AI-research and who are the leading academics? Which investors, companies and entrepreneurs are active in AI? Which are the leading AI-hotspots in the Nordics? And how does the Nordics compare to leading AI-nations?
We talked to a lot of people and researched 129 companies across the Nordics. We divided this blog into two sections; first the academics, research institutions and scientific organisations, and second, the business and tech community side; companies, investors, entrepreneurs and hotspots. Since AI is a research intensive field, let’s start with the academics.
The Academics, Research Institutions and Scientific Organisations
Beginning with Sweden, we turned to Stefan Carlsson, a professor of Robotic Perception at the department of Computer Vision and Active Perception, (Robotics Perception and Learning, RPL) at the Royal Institute of Technology in Stockholm (KTH). Stefan has the simplest definition of AI; “It does what humans do”.
Stefan Carlsson paints a bleak picture of AI in Sweden. “It’s a big difference between traditional, established enterprises and the startups”, he says. “I get invitations to AI startups and meet-ups in Stockholm all the time.”
Meanwhile there’s not much interest from established companies, he thinks. Maybe they are focused on their current business, and productivity in traditional industry. We have seen few demands from big Swedish companies regarding our competence or hiring our students. The explanation is probably that the applications that are possible with AI comes more naturally for new startups and software companies. However, with more focus on self-driving, autonomous vehicles, there is increasing attention from Volvo and Scania who have shown interest for KTH’s research and students. There are also a handful of newer companies like Context Vision, Tobii and Electronic Arts who have been in touch.”
“It’s a striking difference compared to what the students are interested in”, he continues. “It’s really a boom for AI courses and a huge interest. The reason is probably that students are looking at 40+ years careers in technology and engineering, and want to go for areas that are future-proof. And they don’t seem interested in old companies like Ericsson and ABB.”
This implies that we are likely to see many new AI-startups in the coming years, started or joined by talented students that want to explore opportunities in AI, outside established enterprise. But what about the actual research at the universities?
“There is not much financing for AI research in Sweden. There are a few breakthrough areas though, like computer vision, natural language recognition and robotics. But in general, it seems to take a long time for both the academic world and traditional enterprise to make a transition, it’s not really happening. Compare to Google, Apple and Facebook who have made a very quick switch to AI. In research, there are long funding cycles, five years for projects, and the money tends to go to already existing areas, not new ones. It’s a general problem that Europe is falling behind the US.”
And rest of the Nordics? Stefan Carlsson thinks it’s probably just as bad in Norway, Finland and Denmark. A big push in AI doesn’t seem to be happening anywhere in the Nordics.
To explore further, we turned to Fredrik Heintz, Associate professor of computer science at Linköping University, director of the graduate school of the Wallenberg Autonomous Systems Program (WASP), and president of the Swedish AI Society (SAIS). WASP is Sweden’s largest individual research program with 1,8 billion SEK allocated over ten years, supporting strategic basic research in autonomous systems and software development.
The ambition is to “advance Sweden into an internationally recognized and leading position in these two areas.” Sounds like an amazing AI-project, but is it really about AI? “Not explicitly”, says Fredrik “which I think is unfortunate. Very few researchers in WASP have a background in AI so there is a risk that results from AI will be missed. At the same time much of the research will probably touch on AI-related questions, for example machine learning.”
Fredrik confirms that AI is not a unified research area in Sweden, but divided across several topics such as natural language processing, information retrieval, machine learning, multi-agent systems, and computer vision. “It’s a lot of activity, but it’s not always labelled AI-research”, says Fredrik.
Furthermore, “AI is not a very popular term”, says Fredrik. “One reason is that there was a backlash from so called Expert Systems back in the 80s. These systems tried to encode expert knowledge to provide decision support that were supposed to revolutionize sectors such as healthcare, but did not live up to the exaggerated expectations. Today similar systems are routinely being productively used, but without the AI label. The modern version is usually called decision support tools or question-answering systems and the most well known example is IBM Watson. The perceived failure has tainted the term AI in Sweden.”
“Another reason for the slow development of AI in Sweden is that traditional Swedish industry has set the agenda for academic research, and they have generally not seen a major need for AI. The Swedish government has also focused on industries that can hire thousands of workers, rather than high tech companies, such as Spotify and Klarna, with significantly fewer employees.”
“The focus on more traditional industries has also resulted in more emphasis on electrical engineering than on computer science at the universities. This is however changing, with electrical engineering and computer science converging. It would be great if AI could be a driving force in this unification.”
Just as with KTH, at Linköping University there is now a boom for AI, over 200 students are now taking an AI course and several new courses related to machine learning.
AI seems more of a cross-discipline uniting various fields, than it’s own domain. So, in all, there’s a lot going on, but no common umbrella for everything AI. “It might also be that in the US they are better at talking about AI, while we are actually better at doing it”, concludes Fredrik. “We have the competence at the universities and research institutions, and with the right direction we could easily become a leading nation with the right incentives and appropriate funding”.
Also, he says, stay tuned for at three major international conferences in AI, happening in Stockholm in 2018; International Conference on Machine Learning (ICML), Autonomous Agents and Multi Agent Systems (AAMAS) and International Joint Conference on Artificial Intelligence (IJCAI).
In Finland, we turned to Ville Kyrki who is an Associate Professor the School of Electrical Engineering at Aalto University. He serves as the head of the Intelligent Robotics research group. “Finland is probably one of the most (artificially) intelligent countries in the Nordics”, says Ville. “Robotics is a strong field and there is a long history of machine learning.” For example, professor Teuvo Kohonen is a pioneer in artificial neural networks and self-organising maps, sometimes called Kohonen maps. Ville can also see several spin-off companies using machine learning tools, most are in startup phases. Regarding intelligence in robots there is an exciting company called Zen Robotics, which is in scaling phase. Ville points out there are two strands of research. One is autonomous systems, machines and vehicles (which is not AI as such, but are using AI tools and robotics). This field is rather well established. The other is robotics, and there are a number smaller robotics companies, like Generic Intelligent Machines (GIM). The research at Aalto University is centered around two areas; 1) autonomous machines/vehicles, and 2) service robots, especially in healthcare. In general there is much research directed at traditional industry, like cars, mining automation and forestry automation, typical Nordic businesses.
There are a few other notable Nordic academics, active in Artificial Intelligence, robotics and machine learning in various ways:
- Finn Verner Jensen, professor emeritus at the Department of Computer Science, Aalborg University, Denmark
- Agnar Aamodt, Professor in Computer Science and Artificial Intelligence, and Deputy Head of Department, Norwegian University of Science and Technology (NTNU), Norway
- Erik Sandewall, Professor in the Chair of Computer Science at Linköping University, Sweden
- Danica Kragic Jensfelt, Professor at the School of Computer Science and Communication at the Royal Institute of Technology (KTH), Sweden
- Devdatt Dubhashi, Professor at the Department of Computer Science and Engineering, Chalmers University of Technology, Sweden
- Teuvo Kohonen, prominent Finnish academic and researcher, currently professor emeritus of the Academy of Finland.
In Sweden, there are a number of organisations that organise scientific research in Sweden and the interaction with business and enterprise:
- Royal Swedish Academy of Engineering Sciences (IVA) is an independent organisation promoting exchange between business, research, and government, in Sweden and internationally.
- Vinnova is Sweden’s Innovation Agency, funding needs-driven research and stimulating collaborations between companies, universities, research institutes and the public sector.
- The Swedish Research Council (Vetenskapsrådet) is a public agency under the Ministry of Education and Research, with the purpose to develop Swedish research of the highest scientific quality, contributing to the development of society.
- Swedish Foundation for Strategic Research (SSF), supports research in science, engineering and medicine for the purpose of strengthening Sweden’s future competitiveness.
- Research Institutes of Sweden (RISE) is a group of research and technology organisations, in global co-operation with academia, enterprise and society.
- SICS Swedish ICT (SICS) is owned by the Swedish government and a leading research institute for applied information and communication technology.
We contacted these organisation to ask if there are any specific projects within Artificial Intelligence. There seems to be many activities related to computer sciences, AI and machine learning, however it’s seldom labelled “Artificial Intelligence”, as Fredrik Heintz pointed out earlier.
Niklas Rudemo from SICS, says that their work in the field of AI mostly concerns the practical application of AI in industrial settings. While the theory behind AI is complex, it is even more complex to understand which problems are suited for a particular technology. By working with its customers in Swedish industry and with international research partners, SICS has gained pool of talent and experience in this field.
Finally, it does not seem to be any push at all from the political side. The Swedish government has launched projects like “Smart Industry” and various general digital initiatives, but nothing in Artificial Intelligence. The reason, in Sweden anyway, is that the government don’t want to promote any specific technology, it’s up to the enterprises. But maybe they should. A strong edge in AI might quickly turn into a national advantage.
The Companies, Investors, Entrepreneurs and Hotspots
Slush is the Nordics’ and one of the world’s largest tech conferences, gathering thousands of entrepreneurs, investors and startups. According to their database there was 241 AI-enabled startups at Slush, and of these 113 are from the Nordics. Searching Nordic Tech List (NTL), an industry database, there are currently 56 companies tagged as “Artificial Intelligence”. Crunchbase is listing 33 Nordic companies in the Artificial Intelligence category. Pooling together these lists, with our own research, we came up with an extrapolated list of 129 companies. You can find the list here. What did we learn?
Our impression is that there are really two kinds of AI-companies. The “Pure-AI” startups rely on long research, access to huge amounts of data and advanced technology. They were usually started as AI-companies, to explore a certain field within the vast domain that is AI. The other kind we can call “Wannabe-AI” startups. They might have started as more basic software companies and have then added smart features and more data, or they are simply just taking advantage of the AI-boom for better PR. In any case, we have not really distinguished between the two, any company that identify themselves as “AI” is on the list (we have however removed some cases that are not even remotely close to what we think is an AI-company).
Where are the companies based? There’s a strong dominance in Sweden and Finland (together 84%). It might be because we used a Finnish database (Slush) and a Swedish one (NTL), but we also cross-checked with the international Crunchbase to even out any bias. However, here’s the data:
- 57 AI-companies are from Finland (44%)
- 51 are from Sweden (40%)
- 11 are from Denmark (8%)
- 10 from Norway (8%)
And in which cities are they based? Also here Finland (Helsinki), and Sweden (Stockholm) are the main hubs, with the top-4 Nordic cities being:
- 34 AI companies are based in Helsinki
- 32 are in Stockholm
- 9 are in Copenhagen
- 8 are in Gothenburg
What about investments? Atomico’s latest report “The State of European Tech 2016” analyses deals made 2011-2016 in “Deep Tech” (a term that includes a number of advanced technology categories such as Artificial Intelligence, Internet of Things, Virtual and Augmented Reality and Frontier Hardware). The report found that in recent years (2014-2016) there were 950 Deep Tech startups founded in Europe, compared to 1,252 in the US. Looking at AI-companies, 43% of these got funded, compared to 53% of AI-companies getting funded in the US. In total, there was $2.3 billion invested in Deep Tech companies in Europe since the start of 2015. How much were invested in the Nordics? In Sweden there was a total of $198 million invested in Deep Tech during 2011-2016, Finland comes next ($137 million) and then Norway ($56 million). There is no data for Denmark.
Making a round of calls to some leading investors in Stockholm, gives the impression that there are not massive investments in AI. Industrifonden has made an AI-related investments in Now Interact. Creandum made an investment in a German AI-company. And there has been some acquisitions, for example Ebay bought ExpertMaker.
Bjarke Staun-Olsen, a Dane living in Stockholm and Investment Manager at VC firm Creandum, is focused on Artificial Intelligence and Fintech sectors. With a background in Machine Learning, Bjarke was an entrepreneur before becoming an investor and built an AI-startup, Alipes Capital, that uses machine learning to predict the financial markets. So what does he think about AI from an investment perspective?
“The technology itself is insanely promising”, he says. “It’s massive how much have changed, just compared to a year ago. However, it’s not obvious for a startup, since a successful AI-company needs much data, and the competition (Google and others) usually have the data. There’s a strong incentive for Google and other digital giants to offer open source algorithms and promote innovations, since the real value is in the data and what you can do with it, not the algorithms. The way to do a machine learning startup, is to find a very narrow vertical where there is limited competition and you can get a unique dataset and a data advantage. The algorithms themselves are not enough, but necessary.”
And what about investments in AI? There is probably the same ratio of investments in AI compared to investments in other sectors in the Nordics, as anywhere else. Bjarke can’t see other nations investing proportionally more in AI. And, he can’t think of any AI companies that have raised huge amounts of money anyway. However, there are large tech startups that have got much capital, that depend on AI without being AI-companies. For example Spotify relies on AI for recommendation engines, Klarna uses AI for detecting fraud and credit assessment and Uber has algorithms for predicting logistics. They benefit from AI because they have the data.
Finally, in terms of hotspots, Bjarke thinks the main Nordic tech hubs like Stockholm, Copenhagen and Helsinki probably rank as well in AI as New York or Berlin. However, they are all behind the two main AI-hubs in the world, San Francisco in the US and London in Europe. With that, we turn to another key influencer in Nordic AI.
Roelof Pieters is an academic and PhD at the Computer Science and Communication department at KTH, turned entrepreneur. He has a background as tech director of R&D at Vionlabs and has now started his own company, Creative.ai.
He says that the Swedish and Nordic AI markets are very fragmented, and that students are normally also separated from the startup crowd. People that want to do something move somewhere else. If you are a student and interested in startups in AI, the best place in Europe is the UK. He confirms that London is the strongest tech hub for AI development in Europe today. And top talents at Nordic universities (mostly from Asia, Russia, China and India) get headhunted by Google, Microsoft, IBM and others that spend significant money on recruiting and attracting students and whisk them abroad. The students are usually identified by their final thesis, anything related to AI gets the attention from the recruiters who make sweet offers. Canada is an interesting case, says Roelof, and there’s now an AI research lab in Toronto. Earlier, the Canadian talent was mostly being sucked up by the US, but now students stay, and research is heavily funded by the government.
What about AI-startups and investments? Roelof thinks that few companies actually starts out as AI companies (as we pointed out earlier). It’s rather a process where software companies start doing the basic stuff then gradually add more intelligence with algorithms as the data grows. He sees a lack of expertise from investors in AI and thinks that the Nordics is generally more risk averse, and that are reasons why we there are few investments in AI in the Nordics. The main problem for AI-startups is getting funding.
Roelof has helped organising the AI-community in Stockholm, first by initiating the Stockholm Deep Learning Meetup, then merging into the newly founded Stockholm.ai, part of Nordic.ai (run by Pia Elle Elmegård of Corti Labs), which in turn is under the umbrella of the intrenational City.ai iniative, a global community for applied Artificial Intelligence. Roelof is a Stockholm AI ambassador. There is also a Malmö.ai and the city.ai movement is now spreading to Oslo and Helsinki. The purpose of the meetups is to bring together students, researchers, techies, business people, investors and startups, all interested in AI.
AI is now also getting the attention of the city of Stockholm. Joseph Michael, Head of Startup and Tech for Stockholm, recently wrote a blog post about Artificial Intelligence in Stockholm.
In Finland we met with Marko Laiho (CEO, co-founder) and Eero Hammais (Product Manager) at Headai. The founder of Headai is Harri Ketamo, an independent researcher and start-up entrepreneur with over 15 years of experience in Learning Sciences, Data Mining and Artificial Intelligence. He is a Senior Fellow at University of Turku and Adjunct Professor at Tampere University of Technology.
An exciting company that was founded in 2015 based on ten years of pioneering research Learning Sciences, Data Mining and Artificial Intelligence, Headai is developing teachable software robots (bots) to automate various everyday intellectual jobs. These bots are fast learners, interactive and reliable assistants. They can boost data discovery, help to make better decisions on data, automate complex business processes and much more. Headai’s solution is a combination of semantic neurocomputing and learning algorithms following human behaviour. The bots use unstructured or given data to create semantic neural networks which can offer up to 1000x more deep insight compared to traditional ontologies (i.e. structure of languages).
Headai’s focus is primarily semantic neural computing, analysing massive datasets of text, words and how words are connected, the context of sentences. The bots can start reading any language, through understanding what are the connections of words, the semantics. The bot creates an image of the languages used for reading news, finding similar news and comparing with the text DNA of the surrounding world. Headai’s vision is that companies and individuals can have their own, personalised life-long bots, that keep on learning and helping us.
To understand in practical terms the significance of this, we asked about a current hot topic: Could Headai help Facebook identify fake news and other disinformation used to undermine democracy? Yes, absolutely, says Marko Laiho. Headai would create a DNA of the text on the site (a “fingerprint of news”), compare it with global master DNA and show the probability that the news is fake if it doesn’t find a similar DNA anywhere else. The master DNA of the text is big data (all the news in the world), compared to a small, specific text DNA (a piece of fake news on Facebook, for example “the Pope supports Trump”).
So what about the AI-startups scene in the Nordics? Marko Laiho thinks there are only 3-5 “real“ AI-companies in Finland and a few in the rest of the Nordics. In Headai’s business there is no Nordic competition and the most competitors are in the US, with some in the UK. And most are early stage.
It looks like Helsinki will further establish their presence in Nordic region as a exciting tech hub, especially in intersection of AI and healthcare. We met with Juha Teljo (Business Intelligence and Predictive Analytics Sales Executive) and Mirva Antila (Director Enterprise Sales Unit, Healthcare, Public and Financial Services) at IBM Finland, who talked about IBM’s new Helsinki development center.
Tekes, the Finnish Funding Agency for Innovation (Tekes) and IBM has announced a partnership that will enable Finland to utilize Watson (IBM’s main AI initiative) cognitive computing to help doctors improve healthcare and strengthen and develop the Finnish innovation and business ecosystem in the fields of health and well-being. IBM will also establish a Watson Health Center of Excellence in Finland, the first Nordic Healthcare Competence Center, and the first of its kind outside the United States in Finland. These centers are expected to employ 150 people over the next few years
Finland is an innovator in healthcare, and Mirva Antila points out several reasons why IBM is moving its new European center here: The government’s healthcare reform, Finland’s rather small population that can serve as a testbed, several initiatives to digitalise the healthcare sector, and a number of companies active in AI, and finally – the unique use of the social security number provides access to big data. It’s always about the data.
It’s obvious that the cooperation between large sectors with clear needs, government reforms, innovation and funding agencies, startups, research and global tech companies can start to accelerate regional development in AI, if there is a will.
The Intelligent Conclusions
To sum up, what are our conclusions about the Artificial Intelligence in the Nordics? There’s of course a lot happening as The Nordics is a leading tech region, as we would have expected, but we need to do more here to not fall behind the leading regions, hubs and nations. We are a bit worried that the world might be divided into traditional, data-poor industry-nations, and on the other hand data-rich software nations. It’s unlikely that the Nordics will take a leading role in AI short term, because of a few essential reasons. In short, we lack:
- Massive global data (compared to the US and China)
- Large pools of venture and corporate capital (compared to the US and China)
- The political will to give AI a strong push (compared to other sectors)
- Software giants with incentives to invest in AI (our economies are dominated by traditional industry)
However, we think there are some important things we can do now, to be winners in AI longer term:
- Allocate much more funding for research in specific AI-related areas
- Outspoken political focus on AI as it’s one of the defining technologies of our time
- More dedicated, risk-willing capital for AI-startups, which are usually early stage and research-driven
- Attention to all the AI-activities, startups and research that is actually going on in the Nordics, to make it more visible. Why not create Sweden’s first professorship in AI to put a focus on AI research and development?
We see one very specific and exciting opportunity in Nordic AI. As we have shown in an earlier blog, the Nordics has a strong base of innovative enterprise software companies, competing successfully in global niche market. They usually have solid sector expertise, deep technology competence and access to unique data in industry verticals. There is a clear opportunity to develop these established companies and technologies further, by injecting more Artificial Intelligence and cooperating with both researchers and AI-startups. We believe we have a critical mass of Artificial Intelligence in the Nordics, and now we just have to start the snowball rolling.
Please let us know what you think!
Image: Aalto University’s stand at Slush 2016