Notes from Nordic.AI at the Reykjavik AI Festival
The last 5 years have seen interest in artificial intelligence and related technologies grow beyond specialist groups, entering the consciousness and daily lives of the general public via the media, as well as its application in products offered by Facebook, Apple, Tesla, and Google. With so many interested in the potential of AI, can these new technologies deliver the benefits many have predicted?
This year, Nordic.AI joined up with the Reykjavik AI Festival, where these questions were addressed with authority, and worked to create new opportunities for collaboration and forge new ties between industry and academia — inside and outside the Nordics.
How do we bridge the industry-academia knowledge gap?
The festival began with an introduction by Hannes Högni Vilhjálmsson of CADIA, an interdisciplinary research center in artificial intelligence spanning School of Computer Science, School of Science and Engineering and School of Business at Reykjavik University. He discussed the importance of gathering researchers, industry, and the public to discuss issues facing society. While many people are familiar with AI through media hype, the event attempted to address concerns and answer questions to help bridge the gap between new technologies and the people they will impact.
Líf Magneudóttir, president of the City Council of Reykjavik, was bullish on AI’s potential. “The subtitle of this conference, ‘A Revolution Underway’, is an understatement,” she said. “We are stepping into a new age of technology that has forever changed society as we know it.” She echoed a common fear that digitization will increase inequality rather than the opposite.
“This huge chance in technology might only help the few- the 5% of companies and the 1% of the people…This revolution has the potential to be so disruptive so fast that it may leave huge groups of lower and middle class citizens worse off.” The role of city planners and politicians in making this transition ethical and effective is paramount. As more than half of the world’s population lives in cities, city policies must be ahead of the game and take the lead when it comes to best practice in AI development and governance.
Reykjavik University president Ari Kristinn Jónsson discussed how university administrators can help move AI forward, even if they feel disconnected from the cutting-edge of AI research. He emphasized the need to increase dialogue between industry and society, and to increase interdisciplinary networks around AI. “The opportunities for value creation through industry and society to ensure we can use these opportunities for the benefit of all and not to move things in the wrong direction. There is nothing more important in that direction than having a dialogue. Hearing from networks, building networks…That is the role of our university,” he said.
Successful industry-academia collaborations
Aljoscha Buchard of the German Research Centre for Artificial Intelligence (DFKI) has lead calculated efforts to bring together industry and academia. DFKI was in fact created specifically around this mission. Founded in 1988, the center weathered the “AI Winter” and lived to see the second wave of digital transformation. “In the first wave of digital transformation 20 years ago, data became digital. We are now in the midst of a second wave where data is digitally interpreted — meaning machine interpretable and therefore usable,” he explained. AI and ML are important tools for connecting data and finding links and patterns.
DFKI has founded more than 70 startup companies, most of which are still active (some have been acquired). They focus especially on human-centered AI — that is, intelligent tools that surround people to be used for tasks like guiding, learning, interacting, and communicating. You can learn more about DFKI projects here.
Dr. Kristinn Thórisson, Professor at the School of Computer Science at RU and Director of the Icelandic Institute for Intelligent Machines, an organization similar to DFKI, outlined the institute’s origins, described some of the successes of recent industrial partnerships, and discussed IIIM’s future in light of increased interest in AI across all sectors.
“Our main mission is to spur progress in high tech innovation in Iceland via word class research in the areas of AI and simulation, and bridge between academia and industry,” he said. IIIM’s activities find commonalities between the goals of researchers and industries.
“Academics need to think years into the future about basic problems that often are not immediately applicable to products or services. On the other hand, industries are thinking about the here and now, and their purpose is to grow and make money,” he said. IIIM takes an active role in moving ideas from research into an applied form. “That may end up transforming the idea, but we will find out much sooner, and with much more certainty, how it can be used in practice than if it was just sitting in a research paper for 10 years.” Research from IIIM often ends up being used for new tools and prototypes.
Their successes can be seen across many application areas, with projects ranging from software for search and rescue missions, improved control in performance of artificial ankles, to detecting epileptic seizures in zebrafish using computer vision.
IIIM is also the first institute of its kind to have published a policy on peaceful R&D (2015), and remains the only AI lab in the world to have done so.
Helgi Páll Helgason of the Icelandic startup Activity Stream, a close collaborator of IIIM, discussed the opportunities of AI in operational intelligence and how this subset of AI is one of the major opportunities for academia-industry collaboration. Activity Stream has created an operations intelligence platform that can be tailored and implemented for selected domains. The technology unifies data streams from diverse custom IT systems so the barriers between data silos disappear and the data can be holistically analyzed.
“We enrich and analyze the unified stream in real-time and generate actionable observations when we find things that employees should know about,” he explained. “For example, if there is a suspicious purchase made by a customer, we will make a fraud observation….Or in energy management, we might observe an anomaly in power consumption and this might mean that something has been left on and even be more serious like being a fire hazard.” In the future, this kind of unified real time analysis of business data is likely to be a standard in enterprise technology.
Jón Skírnir Ágústsson of Nox Medical, also a collaborator of IIIM, further underlined the importance of bridging the industry-academia gap. Nox Medical makes digital devices to record biological signals during sleep. Academic collaboration is important for the company, since sleep studies still use methods and standards developed in the 60s. Although measurement devices got smaller and results are digitized instead of written on paper, the basic methods have barely changed.
“In medicine you can only use the technology that is available, so we can change the game through cross-collaboration,” said Jón. But interdisciplinary work does not happen in a vacuum, and needs a lot of factors to work. One factor that can have outsized impact is tax refunds: R&D tax refunds can be a huge boost to small companies that have difficulty supporting internal research.
Data protection woes
Lisa Mallner of Nordic.AI summarized the findings of an earlier roundtable session geared towards creating an actionable plan for creating a standard process and format for data sharing among industry, government, and academia. The roundtable brought together decision makers across government, industry, and academia representing organizations such as Statistics Iceland, the Icelandic Data Protection Authority, the University Hospital of Iceland, researchers in AI and ML, and AI startups.
The outcome focused on confusion over constitutional privacy rights and new data protection laws such as GDPR, and how the opacity around processes like security audits is overwhelmingly complex even to those whose careers are centered on data collection and analysis. The roundtable looked into some basic resources to better understand best practices in data sharing.
The event ended with a panel session that included Aljoscha Buchard (DFKI), Helgi Páll Helgason (Activity Stream), Jón Skírnir Ágústsson (Nox Medical), Lisa Mallner (Nordic.AI), and Jacqueline Clare Mallett (IIIM), moderated by Kristinn Thórisson (IIIM). The panelists discussed information siloing between AI research and industry, and how this feeds insecurity and fear about the capabilities and risks of AI development. They agreed that more interdisciplinary collaboration is necessary to help solve this problem.
New Nordic research in AI
The event was followed by a poster session that highlighted some of the cutting-edge research in AI done in the region as well as a few interesting AI companies. Some highlights:
- Virtual training to help teachers learn to work with special needs children (2016)
- A neural network for fish fillet quality measurements to help the fishing industry quickly sort cuts of fish for sale (due 2018)
- A modeling of economic monetary theory based on double entry book keeping, which allows the behavior of regulated banking systems to be explored (2015)
- An AI-assisted shopping experience to help customers find the perfect fit (link)