Dual EU 'expert groups' on AI risk duplication

The second is the Global Tech Panel, set up by EU diplomatic service, the European External Action Service (EEAS).

Both groups had their first meetings in June. But while the goal of the first is to produce public reports, the second group operates behind closed doors.

Andrea Renda, senior research fellow at the Centre for European Policy Studies think tank in Brussels, is one of the 52 members of the high-level expert group.

He told EUobserver that the experience has been "mixed" so far.

While he praised the expertise and variety of the members, the size also made things complicated.

The high-level group met three times so far, in June, September, and October.

It decided that AI is such an enormous subject, that it had to divide into two groups – only to realise that even further subdivisions were needed, into a total of 14 working groups.

"This doesn't necessarily guarantee that we reach a very homogeneous result in the end," admitted Renda.

Renda's group has been tasked to produce draft ethical guidelines for developing AI, as well as recommendations for whether to adapt or introduce new legislation, and investment recommendations.

He said that following political pressure from commission president Jean-Claude Juncker, the deadline for the guidelines has been tightened. It is expected by December.

"I wish we had been given more time. But since this is the time we have, however acrobatically, we will try to do the best we can," he said.

Frankenstein problem

The academic warned of a potential "Frankenstein problem" - trying to merge the results of the 14 working groups into a single document when there may be internal contradictions.

"We might end up with the urge to assemble a patchwork of relatively alien contributions that will have to work together," he said.

With so many outspoken experts involved, Renda cited another risk.

"The worst that can happen is that the report comes out and some of the members say: 'I disassociate myself from the report because that's not what I meant and I disagree with some of these things.' It's going to be complicated."

It gets more complicated when you consider that that other group, the Global Tech Panel, set up by EU foreign policy chief Federica Mogherini, also has artificial intelligence on its radar.

The four 'work streams' of the panel may relate to AI in one form or another: lethal autonomous weapons; digital skills; ethics of applied machine learning in surveillance, justice and security; cyber security.

One of the Global Tech Panel members, Tom Fletcher, said he hoped the panel would "come up with some very practical proposals on what we do about artificial intelligence, the sort of ethics and rules and governance of the internet".

That means that both groups are working in some way or another on the ethics of AI.

Fletcher, a former UK ambassador to Lebanon and now visiting professor at New York University Abu Dhabi, said he has not interacted with the high-level expert group on AI.

"I assume within the EU system people are swapping notes and are in touch, but I'm not really aware of that group to be honest," he said.

Renda said that the two groups should meet.

"I think it would be a good idea to establish some ... coherence between the work of the two groups, some interaction," he noted, adding that he was not aware of any such plans.

A spokeswoman for the EEAS told EUobserver in a written comment, however, that the panel "does not aim to produce policy papers or otherwise contribute directly to the preparation or implementation of EU legislation, legislative proposals, policy initiatives or other delegated acts".

However, that contradicts an earlier statement from the EEAS website, which said that the panel was "set to feed into the EU Strategy on Artificial Intelligence the European Commission will put forward in November".

This is the third in a three-part series of articles on how the EU deals with artificial intelligence.