Europe at the vanguard of AI

AI being a discipline that has come of age emphatically in the last few years, significant European acquisitions like DeepMind, Swiftkey and, last week, Magic Pony, make a strong case that the region is at the vanguard of the current revolution.

To discuss how we can continue to build upon this momentum and support the European AI startup ecosystem, earlier last week we gathered a small group of visionary European entrepreneurs, academics and investors for a discussion.

Our friend Demis Hassabis, co-founder of DeepMind, was kind enough to share some of his fundamental views about the artificial general intelligence (AGI) his team are working on at Google.

DeepMind’s work opens up a new world of decision-driven possibilities — a perfect example of this being AlphaGo’s move 37 in Game 2 against Lee Sedol, a brilliant and totally unorthodox approach that humanity in its 3,000 years of playing Go had perhaps never truly contemplated.

Not to let us technologists purely focus on the warm and fuzzy benefits of AI, Owain Evans (Future of Humanity Institute) led a thought provoking discussion on the societal, ethical and deeper practical issues of general artificial intelligence. Owain pointed out that it was important to discuss how we teach a general AI machine the values that humans take as “right”, so that if, and perhaps when, general AI truly arrives, its values will be aligned with ours. Practically, this means codifying what humans know intuitively to be in the spirit of rules and laws in a way that computers can understand, and make the same judgments on. Also, when a generalized AI powered machine has learnt how to keep itself going against even unexpected adverse conditions, how easy would it be to turn off?

Among other topics discussed, with lively views from Saul Klein (LocalGlobe), Tom Hulme (GV), Ben Evans (A16Z), Eamon Jubbawy (Onfido), Alex Housley (Seldon), George Danezis (UCL) we covered technical, operational and societal issues including:

• Whether modern AI driven curated content, such as Facebook’s AI curated news feed, is amplifying the echo chamber effect to a degree that swathes of population never see alternative viewpoints, and whether there’s a way to address this, given it’s not part of Facebook’s underlying value function.
• The big benefits to DeepMind of the partnership and support of Google, and whether this means other startups can compete (short answer: data is less of a differentiator than is popularly believed, and many key techniques have been shared by Google).
• How to build an organisation than can effectively channel academics into building successful startups, and the importance of open sourcing and publishing the underlying technology and research.
• To what degree some of the common worry around machines taking human jobs is justified, and how to avoid sensationalism that often accompanies this emotive topic.

We wanted to post this to share with you a view of the broader discussions we hope to be engaging in in the future. A big thank you to all who joined us, and though we had a healthy diversity of viewpoints one thing that seemed to be universally accepted is that this isn’t one of the false dawns for AI that we’ve seen in the past. It’s here to stay, it’s fundamental, and it’s not going out on a limb to say that, as revolutions go, this is bigger than the big tech cycles that have immediately preceded it, be that web 2.0, cloud computing or even mobile.

If you’re a startup doing something interesting in machine learning, computer vision or general AI, we encourage you to get in touch so we can continue to bring together high quality groups in the future to support and guide the ambitious new startups coming out of Europe.