Pardon the interruption from your regularly scheduled panic about the robot apocalypse. Margrethe Vestager, Executive VP of the European Commission, has informed the BBC that we need some “guardrails” to counter the most pressing risks of artificial intelligence, and no, she’s not talking about the sensational sci-fi spectre of AI extinguishing humanity. It’s all about the rather mundane issues of bias and discrimination – issues that could affect your ability to secure a mortgage or whether you get fair treatment from your local social services.
Vestager insists the real danger lies in the potential for AI to amplify biases hidden within the internet-sourced data used to train these models. So, while we may not need to brace ourselves for a Terminator-style showdown anytime soon, we do need to worry about whether AI is judging us based on our gender, colour, or even postal code.
Meanwhile, across the pond, Ireland’s data protection authority has hit pause on Google’s plans to roll out its AI chatbot Bard in the EU. They’re seeking assurances that Google has done its homework on data protection for prospective users. As one might expect, Google has been asked to cough up this information “as a matter of urgency.” So it’s not just Skynet we need to worry about; it’s data privacy too.
Vestager insists that AI regulation should be a global matter, but admits we shouldn’t be holding our breath for a unanimous international agreement. So, in the meantime, the EU is focusing on creating a voluntary code of conduct with the US, a set of standards for AI usage and development that isn’t legally binding. A nice idea, but we’ll have to wait and see how many decide to play along.
In the face of AI’s rapid development, Vestager stresses the need to be pragmatic when it comes to these regulations. The EU’s proposed AI Act aims to classify AI applications by risk level. High-risk systems like those used for credit evaluations or housing loans will face stricter controls. But the main motto here is, “Let’s get the ball rolling and learn as we go.”
But the AI challenge doesn’t end there. With the potential for AI to meddle in elections and manipulate social feeds, Vestager warns we need to stay “fully on top” of this technology. If not, we could see society crumble under the weight of mistrust and misinformation.
While some tech leaders and researchers have called for a halt in the development of powerful AI systems, Vestager deems this unrealistic. The key, she says, is for developers to agree on the same “guardrails” to prevent any excessive risks.
The proposed AI Act also seeks to restrict the use of biometric identification systems, including facial recognition technology. But don’t worry, the EU isn’t planning on turning into an episode of Black Mirror. Vestager assures that these systems will be strictly controlled and used only in specific circumstances, such as locating a missing child or tracking down a fleeing terrorist.
So, there you have it. The world’s first rulebook on the use and development of AI systems is still under negotiation, and isn’t expected to be finalised before 2025. Hopefully, by then, we’ll have figured out how to keep our AI systems from turning us all into data points or worse, pawns in their game. But hey, no pressure.
