Ketchup effect for AI?
On Thursday, the government finally decided on the AI commission that Prime Minister Ulf Kristerson announced back in August. At the same time, marathon negotiations were underway in Brussels, which produced results this weekend. Hasty in Brussels and too late in Stockholm may be a simplistic summary, but it is not without merit.
What the government has been thinking about since August is something of a mystery. The directives that will guide the AI Commission seem neither innovative nor completely out of place. Perhaps more traditional: identify needs at home and in international work, talk to a few others and submit proposals for constitutional amendments and a few other odds and ends.
Perhaps it was getting the right people that took time. It is positive to have a heavyweight business representative like Carl-Henric Svanberg (chairman of AB Volvo and former CEO of Ericsson). Google's AI company DeepMind has a Swedish head of social policy, Nicklas Lundblad, who will take a seat on the Commission. Together with researchers, trade unionists and a few others, it is perhaps a reasonable mix of backgrounds and skills.
The first point in the Commission's mandate is about competitive, safe and ethical AI development and use in Sweden. The second is about supporting Sweden's international work in AI. Both points are very important tasks. However, the Commission arrived on the scene too late to influence the EU rules agreed in Brussels at 15 minutes to midnight on the day of the AI Commission's presentation.
The commissioner in charge, Thierry Breton, tweeted extensively about the long negotiations and all-night huddles to hammer out the political agreement. Perhaps one would have hoped that bragging about 22-hour shifts and other things would not be appropriate when negotiating rules that are supposed to apply for a long time, in a rapidly developing area with great potential. Even less so when some people talk mostly about the threats and risks of AI. Many in the business community had also urged the Spanish Presidency to slow down the pace of negotiations. But no, no mañana here.
So where do we stand now after the ketchup effect of the last few days in the AI field?
Sweden has been gifted with an AI Commission with an urgent mission to work on the international issues and competitiveness, but with the disadvantage that the most important in that area has just been decided. However, we can only cheer on the work of the Swedish AI Commission. It can probably do a lot of good if it takes the chance - the directives provide plenty of scope.
From Brussels comes the sleepy greetings that Europe has become the first continent to adopt rules on AI, even if they were a bit unclear and perhaps somewhat detailed. There is much to be said about the new AI Act. The aim has been to open up safe and unproblematic AI use in the EU. That is, of course, absolutely right. The question is whether the rules have made many other AI uses unnecessarily unsafe and problematic. The AI Act has unclear rules and imposes administrative burdens on developers. Now comes the real test - some rules in the AI Act start to apply already six months after the Act formally entered into force, and the whole after two years.
Fredrik Sand
Business policy expert