Huawei’s 2nd European Academic Conference explored how to enable rapid growth in Artificial Intelligence in the era of 5G mobile communications and the Internet of Things
BRUSSELS – Leaders from business, research and the EU institutions debated how to drive the development of advanced digital technologies at a 24th November conference organised by Huawei and entitled: The Digital Transformation of Europe: How the Internet of Things, 5G and Artificial Intelligence will change Europe.
Development of faster hardware and software, and the capacity to harness increasing amounts of big data, is driving the “deep learning” behind new forms of artificial intelligence (AI), at every level from mainframes down to smartphones.
For experts at the conference, the growth of deep learning raises a major new policy question: can the latest AI systems work with the EU’s new data protection rules?
After decades of slow progress, AI systems are now developing rapidly by writing their own algorithms, the conference heard, making them like “opaque black boxes”. Figuring out how and why they came to a particular conclusion is getting difficult.
Some of the participants at the conference believe that this fast-growing new branch of computer science could come into conflict with the transparency provisions of Europe’s tough new General Data Protection Regulation (GDPR), which comes into force in May 2018. The directive aims to protect individual privacy against the rush of commercial data applications originating with social media and Internet companies.
Transparency will be crucial
One speaker, Prof. Mark Bishop of Goldsmiths, University of London, highlighted how dynamic recurrent neural networks – a technology behind some AI systems - continually adapt and refine their algorithms in response to internal and external signals. That makes it very difficult to audit or reconstruct why a particular AI system came to a particular decision.
Pearse O’Donohue, acting director of the Future Networks Team at the European Commission, acknowledged: “We can’t audit every system, all the time, in real time, but we need to set the rules and then have random spot checks, so that any system that is doing real-time learning and adaptation is applying” the principles set out in the GDPR.
Paul Rübig, Member of the European Parliament, echoed the need for accountability: “Where I see the danger is that we are creating a new small elite that has the knowledge… and the rest of the people have no idea what is going on. So, the most important issue is to create transparency.”
Policymakers and business leaders will need to find ways of tackling these issues if Europe is to realise the potential of AI. But, as Huawei has pointed out, AI in smartphones may actually help service providers comply with GDPR, by enabling users to do a lot of the AI computing and intelligence in the device. That means really sensitive and personal data won’t need to go to the cloud anymore.
A surge in AI applications
Daniel Cremers, professor and managing director of the Technical University of Munich’s Department of Computer Science, demonstrated how computer vision and AI are both now evolving rapidly. His team has developed algorithms that use the images captured by cameras to build a 3D map of the local environment in real time. This technology may be able to support autonomous robots, self-driving vehicles and drones.
Fuelled by advances in processing power and the availability of training data, AI is turning the heads of scientists and engineers. “Artificial intelligence is driving the talent like crazy,” Cremers added. “In classes where we used to have 30 students, now we have 300, 400 to 500 registered students. All want to learn the basis of artificial intelligence.” He said his research is attracting funding from the German Science Foundation, the European Research Council and big corporations such as Google, Microsoft, Intel and Huawei.
More broadly, the European Commission has allocated about €1 billion to support AI and robotics research in the Horizon 2020 programme, matched by €2.1 billion of private funding, according to Patrick Child, Deputy Director -General at DG Research and Innovation.
Child said that IBM’s recent decision to base its 1,000-employee AI and IoT global research centre in Munich indicates Europe is “very much on the map” in the development of these advanced technologies. He added that the biggest challenge is ”having the funding to reduce the risks for operators to get their good ideas to the marketplace… this will be one of the key themes in the proposal we are working on now for the next Framework Programme.”
Policymakers also need to consider the extent to which startups or enterprises will drive the development of advanced technologies. Some speakers suggested that big companies with big data sets have an inherent advantage. “There are voices suggesting …the best day for startups has already passed,” said Prof. Martti Mäntylä of Aalto University, Finland. “The logic behind that is the capability of gathering lots of data, making sense of data, is simply very difficult for startups to do.”
But start-ups could overcome their data deficit by being more nimble. Laetitia Cailleteau, Managing Director of the Liquid Studio Director at Accenture, described how fintech startups have been quicker than established banks to develop automated mortgage advisors. Among the big banks in the UK “nobody managed to put it at the top of their agendas,” she said, while start-ups who “don’t want to have 3,000 advisers” have been quicker to deploy robots with human support.
5G could be with us next year - Huawei
The Vice President of Huawei’s European Research Institute, Walter Weigel, pointed to the success of London-based DeepMind, which has been acquired by Google, as evidence that European startups are pushing the boundaries of AI. But he said more public support is needed, as the relatively small size of the EU’s research budget, compared with structural funds and farming subsidies, indicates there is “something wrong.”
Another high priority for the European Commission is an early roll-out of 5G mobile technologies, which could vastly expand the IoT and data available to AI developers.
Huawei’s Western Europe Regional Sales Vice President, Gaston Khoury, told the conference that 5G may appear in Europe as early as the end of 2018, based on his recent discussions with European mobile operators.
As 5G will enable computing power to be distributed to the edge of the networks, connectivity should become faster and more responsive, enabling autonomous systems to quickly access real-time information. Prof. Maziar Nekovee, University of Sussex, contended that the combination of 5G and AI “may really accelerate the timelines for self-driving cars.”