With world-class analysis institutes in synthetic intelligence (AI), Sweden retains up with all the most recent concepts – and typically even steps out forward. However blue-sky analysis doesn’t at all times result in sensible options that can be utilized by business. That’s the place AI Sweden performs a key position.
An necessary first step within the growth of AI Sweden got here when Mikael Ljungblom was working as a political advisor to the Swedish minister for digital growth. Whereas travelling to see what different international locations had been doing, Ljungblom and his colleagues noticed that international locations resembling Japan and China had been investing in AI. Given the significance of the know-how for competitiveness and societal growth, they sensed a have to develop an AI centre in Sweden. That was in 2017.
“I began to talk to completely different actors within the ecosystem and I met Mats Nordlund, who’s now our colleague,” Ljungblom tells Laptop Weekly. “He was working within the automotive business in Gothenburg – and he had the identical mission as I did. We locked arms and tried to go about this collectively. We had nice assist from Vinnova, which is Sweden’s innovation company.”
Vinnova offered the bottom funding for what was to turn into AI Sweden. Johanna Bergman joined the hassle in 2018. She had beforehand been engaged on pre-hospital e-health on the Lindholme Science Park in Gothenburg.
“We acquired this task to arrange an AI centre in Sweden,” says Bergman. “We began by doing a feasibility research – by soliciting concepts from all actors we may consider. We talked to Swedish business, with authorities authorities, municipalities, areas, academia and a few worldwide actors, making an attempt to know what this needs to be. It was clear we should always work on utilized AI, not basic analysis. Our task was to speed up utilized AI in Sweden.
“In February 2019, we inaugurated the centre, which we known as AI Sweden. Companions of all kinds had been concerned – from startups to firms, academia to public sector. It was necessary to have this unfold from the start.”
Ljungblom is now head of communications and public affairs at AI Sweden. Johanna Bergman is now head of strategic initiatives. She was not too long ago recognised for her achievements by being listed within the Forbes 30 Underneath 30 for Science & Healthcare in Europe.
“After the inauguration, we instantly began up collaborative AI tasks – cross-sector and involving completely different sorts of organisations – to work on actual AI issues,” Bergman tells Laptop Weekly. “That has developed into what we do at this time. Lindholme Science Park helped us quite a bit to start with and is at this time considered one of our host organisations.”
Benefits of a horizontal method
The truth that AI Sweden operates throughout sectors makes it completely different from most AI analysis centres in different international locations. As an alternative of taking a vertical method by specializing in one or two industries, AI Sweden has recognized a couple of areas which might be necessary throughout a number of sectors.
One such space is edge studying, which is a brand new paradigm in AI the place studying happens nearer to the supply of the coaching knowledge. The fashions which might be constructed from the training course of on the sting are then aggregated in a central location the place they’re shared. The information itself is rarely transferred to the central location, which has the fascinating facet impact of serving to to make sure knowledge privateness.
“After the inauguration, we instantly began up collaborative AI tasks to work on actual AI issues. That has developed into what we do at this time”
Johanna Bergman, AI Sweden
That is completely different from conventional machine studying (ML), the place the coaching is centralised and makes use of a centralised dataset. AI Sweden has arrange an Edge Studying Lab, realizing that companions from completely different sectors – resembling automotive, healthcare and the general public sector – have to work with the brand new paradigm.
The Edge Studying Lab features a take a look at platform, together with shared sources for builders, knowledge scientists, college students and researchers to study extra about numerous elements of edge studying. AI Sweden and companions Zenseact, HPE, Volvo Automobiles and Ericsson launched the lab in Gothenburg on 27 April 2022. The lab will assist completely different individuals and organisations shortly construct a piece atmosphere for experiments.
One other horizontal space of focus is the Swedish language. Not one of the world tech actors care about growing fashions for a language like Swedish, so AI Sweden determined to take that up itself.
“What we see now are main developments within the area of language know-how and what we name foundational fashions – enormous fashions that perceive pure language and that may generate language,” says Bergman. “Virtually each week, some organisation someplace on this planet releases a brand new language mannequin – these are huge language fashions that may generate textual content from photographs and pictures from textual content, for instance. Whereas we see main developments with English fashions, not a lot work is being achieved for smaller languages like Swedish.”
To assist fill this hole, AI Sweden arrange a number of tasks addressing language fashions for Swedish. Not solely is the language lab serving to Swedish corporations, but it surely serves for example for different cultures whose languages aren’t but represented in AI fashions.
The purpose of the completely different tasks is to develop a nationwide information hub on pure language understanding (NLU) to assist stimulate NLU work on Swedish. AI Sweden addresses language mannequin coaching, analysis and deployment. It additionally works on moral and safety elements in a number of ongoing tasks round NLU. Swedish authorities and public sector organisations are the primary stakeholders and can be capable to use the fashions to summarise texts, and for translation and writing assist, amongst different issues.
Along with edge studying and Swedish pure language processing, AI Sweden additionally works on a couple of different horizontal subjects, however has not but arrange programmes for them. These embrace privacy-preserving AI, advanced programs, and administration and logistics planning. The thought is to seek out the areas which might be related for a lot of completely different sectors.
Networking and collaboration to synergies
“Networking and collaboration are an necessary a part of what we do at AI Sweden,” Ljungblom tells Laptop Weekly. “We attempt to join individuals who might need widespread pursuits with out realizing it. Our horizontal method permits us to give you utterly new – and typically shocking – collaborations.”
One of many extra shocking synergies is that some hospitals use fashions developed by the automotive business for self-driving vehicles. They use these similar fashions and methods to foretell how the hospital will operate.
One other sudden collaboration is between the automotive business and college researchers learning hen behaviour. The automotive business is utilizing a dataset developed by researchers to classify behaviour of Baltic Sea birds. This dataset is helpful for the automotive business as a proxy dataset for collaborative tasks on self-driving.
“Networking and collaboration are an necessary a part of what we do at AI Sweden. We attempt to join individuals who might need widespread pursuits with out realizing it. Our horizontal method permits us to give you utterly new – and typically shocking – collaborations”
Mikael Ljungblom, AI Sweden
The method of AI Sweden is to both take what’s already taking place in pure analysis and apply it to business or take what already works in a single area and apply it to a different. Not solely is that this useful for society, but it surely additionally helps Swedish organisations stay aggressive on a world stage. By serving to Swedish organisations overcome a few of the widespread challenges, AI Sweden accelerates the usage of synthetic intelligence and helps corporations get forward of the curve on AI adoption.
Expertise growth and networking
When AI Sweden was based, it wasn’t apparent that it may play a task in serving to develop the fitting AI expertise in Sweden. However because it started networking with actors in numerous industries, it grew to become clear that AI Sweden was uniquely positioned to identify what skills had been wanted and to develop consolidated programmes that will assist develop these abilities.
Simply final yr, AI Sweden began a set of expertise programmes. These purpose not solely to develop particular skills, but additionally to assist develop the fitting firm cultures for AI.
AI Sweden additionally tries to attach the dots and get the completely different components of the ecosystem to work collectively. One in every of its goals is to ensure authorities funding is being spent in the absolute best approach. It does this by staying abreast of who’s engaged on what and discovering synergies to profit from the funds.
One other approach AI Sweden helps speed up adoption of AI is by addressing societal challenges. AI modifications the best way individuals work and can most likely remodel industries.
To assist put together organisations for the large modifications, AI Sweden works with companions to ensure they take steps in direction of turning into extra AI mature. As particular person organisations turn into extra mature in making use of AI, your complete ecosystem advantages.
“We’ve discovered a structured approach of doing this,” says Ljungblom. “We use an AI maturity evaluation device that we tailored from the Initiative for Utilized Synthetic Intelligence in Germany. It’s an excellent device that applies a structured method for benchmarking enhancements and figuring out what areas to give attention to when working with AI. This enables organisations to take the work they’re doing to the subsequent stage.”
Nationwide benefits of Sweden
“Finland was fairly early in making AI a strategic precedence,” says Bergman. “We are able to acquire quite a bit from collaborating within the Nordics. For instance, a number of Nordic international locations have very related languages, so with regards to language purposes, there’s nice potential for working collectively. That’s beginning to occur.”
Ljungblom factors out that an attention-grabbing factor about Sweden is its decentralised authorities construction, with autonomous areas and municipalities. “Therefore, we’re compelled to seek out efficient methods of working with quite a lot of completely different public entities,” he says. “If we evaluate the EU to the US and China, we’re way more various and decentralised, with 27 completely different member states. In a approach, this resembles the autonomous construction of Sweden, so we would have one thing to supply there.”
Two elements of Swedish tradition give it a aggressive benefit in AI. One is that it has at all times been a technologically pleasant nation, with a inhabitants that’s fast to turn into digital and to make use of digital instruments. It additionally has an excellent entrepreneurial system with an excellent ecosystem of startups. On the similar time, there are quite a lot of very massive Swedish corporations, resembling Ericsson, Saab and Ikea, and these corporations are likely to sustain with technological advances.
The opposite benefit is that Swedish individuals are used to collaborating. It’s a part of their tradition. In know-how, they collaborate throughout sectors – and between business and academia. This mindset helps with AI, the place an ecosystem is required and the place lots of the constructing blocks could be shared throughout a number of software domains.
AI Sweden depends on this pervasive perspective of cooperation to assist join the dots.