AI Sweden heading for an AI transformed Sweden

AI Sweden heading for an AI transformed Sweden

Vision: to accelerate AI in Sweden

AI Sweden was founded on the vision of building a national centre for applied artificial intelligence and accelerate the transformation of the post-industrial Sweden.

Like the mythological Bird Phoenix, AI Sweden in 2019 came out of the ashes of the former shipyards of Lindholmen in Gothenburg, now renamed Lindholmen Science Park.

The AI Sweden bird has in three years spread its wings over 120 companies and organizations in Sweden and is located in nine different towns.

-Our goal is to accelerate applied AI in Sweden, as simple like that, to the benefit of our society, our competitiveness, and for everyone living in Sweden, says Martin Svensson, operational co-director at AI Sweden.

Portrait Martin Svensson
Martin Svensson, operational
co-director of AI Sweden.
Photo: AI Sweden

NorwAI met with the architect of this ambitious adventure at his office at Lindholmen Science Park. More than 50 people is currently working on various projects, both public and private, within the framework of AI Sweden. Svensson says it has been a ‘round-the-clock’ hurry from start until now to build the organization, get the first results and justify its continuation

-Vinnova, the governmental tool for innovation, gave us three years of funding back in 2019. I am happy to say that we have earned the trust to go on with our work. AI Sweden has renewed its contracts with both Vinnova and partners and is funded for several years to come, says Martin Svensson.  

To achieve this, AI Sweden runs projects of national interest in areas such as information-driven healthcare, AI solutions for the Swedish language, data-driven journalism, and AI to help tackle climate change. It provides targeted training for our partners and the general public. The Data Factory enables AI Sweden partners to make available and access data and make use of computing power and storage capacity in AI projects.

-We want to contribute to a culture of sharing, cooperation, and action in the Swedish AI-ecosystem. That is why we run all our initiatives in collaboration with our partners, says Martin Svensson.

The partners represent private companies, the public sector, academia and research institutes. AI Sweden is well distributed across the country with offices in Stockholm, Gothenburg, Lund, Örebro, Luleå, Sundsvall, Skellefteå, Linköping and Karlskrona. In the end, the purpose of the efforts is to improve the quality of life for all people living in Sweden.

-We want to move the society forwards. We have come far in the transformation programs with tools to measure maturity for both big and small organizations, thus helping them to improve their businesses.  We want to contribute to a culture of sharing, cooperation, and action in the Swedish AI-ecosystem, says Martin Svensson.

He also want to extend his cooperation to Norway. AI Sweden is keen on extending their relations abroad, and has set up an office in Montreal, Canada to network into the North American AI communities. Norway is also in his plans for the future, and he is open for future cooperation between AI Sweden and Norwegian partners like NorwAI. 

Hunting high and low

In Sweden his organization is hunting both high and low for transformation and assist companies over the AI threshold. Project managers Anders Thoresson and Astrid Sjøgren runs the “AI Change Project.” Based on the ecosystem of a larger EU program, which is in use in several European countries to help organizations to make use of AI, the two project managers have funneled the key factors down to a lightweight maturity assessment tool for small and medium sized organizations. By mapping all AI start-ups in Sweden and organizing matchmaking events, they assist small and medium size as well as non-technical companies to approach artificial intelligence. One of their learnings so far: easy to say, more difficult to get attention from busy people in small scale businesses. 

Picture og Astrid Sjøgren og Anders Thoresson
Most businesses in the Nordics are small ones. Change agents Astrid Sjøgren and Anders Thoresson have adjusted their
approach to these prerequisites in their transformation program.   Photo: NorwAI

Besides, AI Sweden is licensed by EU to use a European program to systematically assess and contribute to matureness for larger organizations. The AI use cases canvas in The AI Tranformation Program is a cornerstone to offer examples for partners to share and learn with a focus on organization, people and a working ecosystem. Their My AI - my.ai.se – a ‘Linkedin for AI’ copycat, to show talent, data sets and usecases, is such a thrill!

Long term projects

AI Sweden has picked three strategic initiatives to serve as platforms for partners from industry, the public sector and academia to collaborate: decentralized AI, Information Driven Healthcare and Natural Language Understanding (NLU); each domain with long-term investments and a number of underlying projects and use cases.

In AI Sweden’ s own words a thoroughly developed base for Natural Language Processing (NLP) is one of the cornerstones of successful AI applications. The Swedish Language Data Lab was the first of NLP projects initiated. 

The Data Factory/Egde Lab is key in accelerating innovation and use of AI in Sweden and enables some of our most exciting opportunities for partners. Partners can engage in the Data Factory by bringing their own challenges, take part in existing projects, or experiment in the Testbed environment. Openness and sharing are done by excluding NDA-agreements. On the other hand, AI Sweden will not claim any IP rights in the projects.

-Partners also value the strong juridical assistance working with projects to avoid law nitty-gritties, says project manager Kin Henriksson at the Data Factory/Edge Lab. 

Kim Henriksson sitting at a desk
Project manager Kim Henriksson, Data Factory/Edge Lab Photo: NorwAI

GPT-SWE

-It is all about cooperation and mobilizing national resources, says Martin Svensson.

 

The NLU initiative is an example of “the Swedish way” of doing things: the program is driven by AI Sweden and RISE, the Swedish equivalent of Norway’s SINTEF, headquartered at KTH University in Stockholm. The current GPT-SWE model is trained on Linköpings University’s supercomputer, Berzelius, using the Megatron framework from NVIDIA. The supercomputer was financed by the Knut and Alice Wallenberg Foundation, to boost collaboration between academia and Swedish industry across research programs such as the Wallenberg Artificial Intelligence, Autonomous Systems and Software Program and initiatives in the life sciences and quantum technology.

 

Out of this cooperation comes GPT-SWE is the first truly large-scale generative language model for the Swedish language. Based on the same technical principles as the much-discussed GPT-3, GPT-SWE will help Swedish organizations build language applications not previously possible. 

 

Talent work

 

Sign reading "Accelerating applied AI inSweden"

-Talent is important. We will start an “AI Academia” where we learn youngsters, 18-19 years of age, about AI on a half time basis in the autumn. The next half year, the youngsters will be engaged in companies on an 80% basis and also enjoy the last day of the week with us at AI Sweden, says Martin Svensson.

They also try to recruit top European talents to Sweden by a PhD-program called “Eye for AI”. Funded by three partners, foreign PhD’s are paid, work and live in Sweden for 18 months,  staying six months at each partner to learn more about the challenges Sweden can offer the next generation AI programmers. The program is experimental, and the results are still to be seen.

 

-It gives them the possibility for a career in Sweden by getting insights from inside the businesses themselves, says project manager Anders Thoresson.

 

Anders Thoresson and Jon Espen Ingvaldsen in discussions
Connecting AI Sweden and NorwAI – Anders Thoresson of AI Sweden and Jon Espen​​​Ingvaldsen,
Adjunct Associate Professor at Norwegian Research Center for AI Innovation (NorwAI). Photo: NorwAI

Published: 2022-06-30

By: Rolf Dyrnes Svendsen