The U.S. is leading investment in AI programs today, but China has declared its intention to become "the world's primary AI innovation center" by 2030. The European Commission, too, has announced a three-pronged approach to increase public and private investment in AI. Countries such as the U.K., France, Germany and Finland are building nationwide programs to increase AI innovation. Why are we seeing all these investments now, what are the most recent AI advances in academia and industry — and is it possible for Sweden to keep up?
Can Sweden keep up with AI investments around the world?
Last week, I upgraded my phone to Android 9 Pie, and it’s difficult not to become overwhelmed by all the predictive functionality such as App Actions, Adaptive Battery, Adaptive Brightness and Google Assistant (now also available in Swedish). Since Google announced the transformation from being a mobile-first to AI-first company back in 2016, the company has seriously started the journey to make use of AI in all corners of its products and services. One interesting insight regarding AI implementations is that the user experience will improve more and more as AI becomes increasingly invisible. Another way to phrase this is the more natural, i.e., human-like, the product becomes, the better the experience will be. In short, AI can help products to become more human-like. This also relates to what Google says about Android 9 Pie: great technology should help, not distract.
The technology giants of the world such as Google, Facebook and Amazon are not just making large long-term investments in AI research, they are also already making huge profits thanks to previous investments. They have large research groups and huge amounts of data at their disposal and are thus able to leverage AI to an extent that few other companies and organizations today can match. What will happen five to ten years down the road? Will the “AI divide” (cf. digital divide) continue to increase, or is it possible for the rest of the world, including Sweden, to catch up?
During 2018, European countries and the European Union have announced a surprisingly large number of AI reports and investments. Most of them were published around April of this year, in a surprisingly coordinated fashion.
In April 2018, the European Commission published a press release and a longer associated report about how Europe can start to catch up with the rest of the world and boost competitiveness in AI. The report clearly recognizes the profound impact AI is making and will continue to make on our society, stating “just as the steam engine and electricity did in the past, AI is transforming our world.”
The report contains numerous insightful statements about the current state of AI in the world and also recognizes that the U.S. and China are in the lead today. Some statements portray Europe’s position in certain fields quite optimistically, though. For example, the report claims that Europe is the world leader in robotics and manufacturing sectors, a rather bold statement. The figure below presents an overview of the major proposals in the report.
In addition to the EU report, several European countries also published their national strategies for becoming leading AI nations. Some examples:
- The United Kingdom government published an AI Sector Deal in April 2018, committing close to £1 billion, but it should be noted that the majority of that funding is coming from private companies and other organizations. The government’s goal is to promote the adoption and use of AI by building a skilled workforce, as well as stimulate access to data — two very important challenges, and the building blocks of any successful AI project. The U.K. has produced some noteworthy companies in AI, most notably DeepMind, which was acquired by Google in 2014. The amazing success stories provided by DeepMind have probably not been missed by anyone working in the field of AI. In particular, their contributions to deep reinforcement learning and projects such as AlphaGo, AlphaGo Zero and WaveNet are truly groundbreaking efforts that also resulted in a cover article in the prestigious scientific journal Nature.
- France made a similar announcement on March 27, 2018, committing to spending $1.5 billion toward AI development in the next four years. Here is a quote from the announcement by the French President Emmanuel Macron: “[Artificial intelligence] is a technological, economic, social and obviously ethical revolution. This revolution won't happen in 50 or 60 years, it's happening right now. There are new opportunities and we can choose to follow some innovations or not.”
- Germany's Chancellor, Angela Merkel, announced her plans for a national AI strategy one week prior to Macron’s announcement. The exact size of the investment was not clear but was expected to be around $1 billion as well. Germany has already made some noteworthy advances in AI, especially around self-driving cars and industrial robots, where German companies own significant amounts of intellectual property and numbers of patents. Germany also has the Deutsches Forschungszentrum für Künstliche Intelligenz, which is claimed to be the world’s largest nonprofit contract research institute for software technology based on AI.
- Finland is another country that has built a significant national AI strategy is Finland. Mika Lintilä, Finland’s economic affairs minister, announced in April 2018 that “We want Finland to become a leader in applying artificial intelligence and robotics to the benefit of societies and enterprises.”Finland is in the process of setting up a national steering group (NSG) to develop a strategy and advise the government on matters related to AI. The exact amount of funding earmarked is not clearly stated, but €200 million are available as grants for companies interested in developing their AI usage.
We are still waiting for a strong national strategy around AI in Sweden, but a number of initiatives have already been established.
In late 2017, the Swedish government commissioned Vinnova, Sweden’s innovation agency, to pursue the following tasks:
- Analyze the potential of AI for Sweden, particularly for industry but also for society in general
- Report what Sweden’s position is today in the field of AI, with respect to education, research and industry
- Identify potential bottlenecks and challenges for making use of AI in Sweden
The 170-page report was published — no surprise — in April 2018 along with many other European initiatives. I had the pleasure of being one of the reviewers, and it contains a large number of objective and insightful details about the current state of Sweden’s AI trajectory. See figure 2 below, visualizing the number of scientific AI publications for different countries.
The figure speaks for itself. We have a long way to go in Sweden and Europe to catch up with the U.S. and China. The reports contains a number of interesting conclusions, such as that AI has the potential to double economic growth in Sweden if we make massive use of AI, and that we need to concentrate our efforts to provide incentives for industries to start adopting the latest AI technologies.
This conclusion is based on reports by McKinsey & Company 2017 and Accenture 2016, where scenarios with different levels of AI usage were analyzed in the timeframe of 2016 to 2035. According to the Accenture report:
“With the recent convergence of a transformative set of technologies, economies are entering a new era in which artificial intelligence (Al) has the potential to overcome the physical limitations of capital and labor and open up new sources of value and growth.”
The Vinnova report contains several other interesting insights and suggestions. See figure 3 below for a brief overview of the content.
Several agencies and research programs have also recently started to regroup and focus more on AI in Sweden. Here is a non-exhaustive list of the top initiatives:
- The Research Institute of Sweden (RISE) restructured after summer 2018 and now has around 2,200 employees working to accelerate innovation in Sweden. The institute has more than 75 active AI projects and 60 scientists dedicated to accelerate the adoption of applied AI in Sweden.
- Wallenberg AI, Autonomous Systems and Software Program (WASP), a multi-billion SEK research program that recently received one additional billion SEK specifically dedicated to AI research.
- Zenuity, a billion SEK research institute focusing on self-driving car research, recently also proposed a “data factory” initiative to prepare and help the industry with processing data. Data collection and processing is known to be a fundamental step for successfully building an AI system.
At Peltarion, we are working hard to make the latest AI techniques available not just for the technology giants, but to all companies around the world, including in Sweden. With better tooling, we can hopefully reduce the AI divide that otherwise will continue to increase. It should not be necessary to have large research groups or expensive infrastructure systems to take advantage of the latest AI techniques. Many people do not yet realize the implications of the revolution happening right now within the world of AI, and there is a need to inform not only companies but also politicians about what is happening. That is why we took the initiative to form an AI Council, similar to what Finland did with their NSG.
The AI Council consists of a number of representatives from academia and industry, and also from the Swedish government. It has backing from Marcus Wallenberg, a well-known Swedish banker and industrialist, as an advisor to the AI Council. The goal for the Council is to guide the way forward for a Swedish model of AI, where we can accelerate adoptions of the latest AI techniques in a sustainable way.
It will be impossible to become the leader in all aspects of AI in small countries such as Sweden, especially given the immense AI investments in China and the U.S. However, we are seeing an increasing demand for a sustainable approach to the use of data and AI, especially after recent events such as the Cambridge Analytica scandal. We need to preserve the trust of the users, but we need to do that without hindering innovations in AI. We have an opportunity in Sweden, together with other European countries, to take the lead in a “Sustainable AI” initiative where user trust is prioritized.
The Swedish government recently announced a new AI investment as well, and that decision was taken shortly after an AI Council meeting. Even though we probably had little influence over that decision, perhaps we were able to give a small contribution in that direction. Hopefully, the AI Council can increase the understanding of AI, explain the potential for our industry and inform how to maximize the benefits for our society.
ICML and IJCAI in Sweden
Sweden had the honor of hosting some of the largest AI conferences this past summer, including ICML, IJCAI and AAMAS. Many thanks to Fredrik Heintz and the Swedish AI Society for making this happen, and for all of their hard work in organizing these conferences. I attended one IJCAI session entitled “The Future of AI in Europe,” which was also attended by noteworthy people such as Max Tegmark and the Swedish Minister of Digital Development, Peter Eriksson. The panelists shared their views on this topic, with most of them recognizing that Sweden and Europe are clearly lagging behind the U.S. and China today.
The broader question now is about the way forward toward minimizing the AI divide between nations and companies and helping drive AI momentum in Sweden. There were, unfortunately, no clear answers, at this IJCAI session, for how to solve this difficult and very important problem. However, it was great to hear Max Tegmark speak in clear terms about the rather fragmented situation we have in Europe today, and how we need to demonstrate strong leadership to show the way forward.
Towards a token-free future in NLP
Improving multilingual models for low-resource languages
Extending BERT-like Language Models for Multiple Languages and Longer Documents
Predicting Bitcoin prices with AI
You need not have deep pockets or talented engineers to predict Bitcoin prices. Just use Python to download publicly available data and the Peltarion platform to train an AI on it. How hard can it be? As it turns out, it is not hard at all.
Search text by semantic similarity