If you don’t see anyone in front of you, you’re either waaay behind, or you’re actually winning. For once in my life (please sit down), I’m seeing a glass half full: Europe’s AI efforts.
For those of you not playing along at home, the EU has determined that they’re losing the race in AI to both the US and China. The current thinking is that a “pivot” is needed. Their move is to establish Europe as a champion of ethics and regulatory safeguards in advance of the other superpowers unleashing sophisticated systems into the real world.
I’m not a fan of regulating things in advance. There’s a lot of ways to do it wrong, and only one way to do it right, so the odds favor a cluster. How will you know you;ve got it right when the thing to be regulated hasn't happened yet? Even if that wasn’t the case, I don’t believe the EU is behind at all.
Let's start with a different framing of the situation. For AI to work, two things need to come together: the software machinery (algorithms, networks, etc.), and appropriate datasets to “educate” the otherwise ignorant machines. One without the other can’t perform anything useful. We engineer neural networks, we feed them data, and they come out with purpose and skills. Similar to how we teach children.
The US and China are both advanced on the AI-machinery side, and they’ve been working hard at building the datasets they need to tackle some pretty complex problems. What the EU fears is resulting systems being ethically challenged because they’re built without constraints in place. In their thinking, developers of these systems need to build-in ethics beforehand.
Children aren’t born ethical and worldly. Like children, neither is the AI machinery we’re currently working with. Kids *learn* to be ethical: ethics is in the datasets that *teach* kids as they grow. Biased and unethical data begets biased and unethical AI. We’ve all heard the stories about a lack of diversity in facial recognition datasets resulting in AI that is unable to recognize ethnic minorities. The US and China are both challenged by the lack of diversity and ad-hoc nature of some of the data they’re working with.
And that’s where Europe steps in. While some might see Europe’s diversity of cultures, languages and political systems as a disadvantage in a globally competitive market, I see opportunity. Where else but Brussels can one find a curated database that translates a huge number of identical documents into a multitude of languages (how better to train language translation AI)? Where else but Europe is there such a depth of discourse and deep history of different philosophical and moral systems, played out over centuries?
Europe is unique in its potential to build curated, ethical, unbiased and diverse datasets that reflect the cultural, philosophical and ethical heritage behind the world’s most storied thinkers. Europe could be home to the world’s pre-eminent AI universities: where machinery goes to learn the fundamentals of our human ethics and values. Europe has spent the last few years putting in place a framework where datasets are increasingly made available and data can flow to a large extent. Now the path towards ethical AI is not through regulation, its through education - and the curated datasets that make that possible. And along the way, Europe could lead in the development of the tools and techniques to build and manage datasets that respect the rights and norms of citizens and the community.
If Europe wants to influence the future of AI, the point of greatest leverage is in the digital classroom. Investments today that focus on building tomorrow’s great AI universities - where machines go to learn - are the most effective route to regain Europe’s role as home to the greatest minds of history.