AI experts tell MEPs ban on 'killer robots' is unrealistic
By Peter Teffer
A panel of defence and artificial intelligence (AI) experts on Wednesday (10 October) poured cold water on the feasibility of an international a ban on autonomous weapons systems, requested by a majority of the European Parliament last month.
Wendy Anderson, a former civil servant from the Obama administration and researcher at the Center for a New American Security, was assertive in her views.
Join EUobserver today
Get the EU news that really matters
Instant access to all articles — and 20 years of archives. 14-day free trial.
Choose your plan
... or subscribe as a group
Already a member?
"I don't think bans work. I'm just a realist on this," said the American researcher.
She noted as an analogy that the non-proliferation treaty for nuclear weapons also did not prevent North Korea from developing atomic bombs.
Anderson said that semi-autonomous weapons were already being deployed by the US military, in particular the navy - but also pointed to China.
"I think it's irreversible. I just don't have any evidence or data to suggest that the opposite is true," she said.
"Regulations and controls will not prevent the development of AI. All you have to do is sit and meditate for 15 minutes on what the Chinese are doing to know that," Anderson said.
China is spending billions of investment on AI, with a goal to be world leader in the field by 2030.
Meanwhile, Russia's Military Industrial Committee has set a target for 2025, when it wants 30 percent of military equipment to be robotic.
"The AI train has left the station," said Anderson. "However we feel about it is actually irrelevant."
"The fact is realistically, we are already engaged - I don't really want to say 'in an arms race' because I don't want to use that sort of combative language or infer it - but certainly there is a competitive global arms race on the defence side for AI and who will out-advance the other."
The director of research at The Hague Centre for Strategic Studies, Tim Sweijs, used more indirect language, but also questioned the feasibility of a ban on autonomous weapons - known colloquially as 'killer robots'.
"If you look at military history, I think there has never been an occasion in which states have not taken up the chance to get a military competitive edge over their competitors if they could," said Sweijs.
"If you look at the adoption of machine rifles, or aircraft for military purposes, or unrestricted submarine warfare, or tanks rather than horses - states always adopted that technology when push came to shove. That is an observation that I make as a student of war," he said.
There are of course other experts who disagree - some of them do call for a ban - but the general tone of the panel in the European Parliament was sombre.
"I'm so terribly concerned, really, and I'm really much more concerned after this hearing," said Swedish Green MEP Bodil Valero.
While some suggest that automated weapons systems may decrease the number of human casualties in armed conflict, a large-scale deployment of AI weapons could also raise new problems.
"More precise targeting through AI may make war paradoxically more humane," said Sweijs, adding that AI could also improve medical treatment in the battle field.
But: "The political costs of going to war are likely to be lower if fewer humans are sent to the front, thereby taking away an important constraint of going to war," he added.
Moreover, one scary scenario is that an armed conflict between two AI systems speeds up beyond human understanding and control, leading to a 'flash war' analogous to the financial market phenomenon of 'flash crash'.
If one country decides to give full autonomous decision-making to a weapons system, it will create a dilemma for its opponents.
"Whether to keep the human in the loop and possibly face defeat or follow suit and take the human out of the loop," said Sweijs.
"Then not only the character but also the nature of war will drastically change, leading to a whole new era in human history where machines will fight against one another," he said, adding that this was likely to remain science fiction for now.
A third expert, Allan Dafoe, director of the Governance of AI programme at the University of Oxford, said it would be difficult to have arms control of autonomous weapons.
"It something we should pursue and discuss and explore, but it will be a great challenge, as we learned from history," he noted.
Priority list
Dafoe added that nevertheless autonomous weapons should be high on politicians' agenda.
"I think we should be having this conversation very seriously, more seriously than we are today," said Dafoe.
He encouraged MEPs to "push this up your priority list". There is still some way to go for that.
While a broad majority of MEPs (566 of 751) supported the call for an international ban on lethal autonomous weapons in a resolution last month, Wednesday's hearing was poorly attended.
Of the 30 members of the subcommittee on security and defence, which organised the hearing, only 13 attended parts of the hearing. There were no more than a handful of MEPs present at any time.
Except for centre-left MEP Clare Moody, who chaired the meeting, none were there from beginning to the end.
Site Section
Related stories
- MEPs delay debate about 'killer robots'
- MEPs bar WMD and killer robots from new EU arms fund
- Killer robots debate shortened by unpaid bills
- 'Killer robot' projects eligible for EU defence fund
- Rise of killer robots seems inevitable at EU conference
- On Armistice Day, EU is still best gift we can give our children