Listeners to last week’s TechRadio will know about the current celebrations surrounding the 200th anniversary of George Boole. One of UCC’s impressive series of commemorative events saw Dr Ken Ford of the Florida Institute for Human & Machine Cognition (IHMC) deliver a talk last Friday on the state of the art in robotics and artificial intelligence that offered an important counterpoint to the current debates in defining what AI is and the chances of it surpassing human achievement. It’s a topic as old as the stage play Rossum’s Universal Robots and explored in films like Blade Runner and the excellent Ex Machina.
In one corner we have the harbingers of doom in Bill Gates, Stephen Hawking and Elon Musk. In the other we have researchers like Ford who foresee a less adversarial relationship with AI.
Gates, in an AMA earlier this year on Reddit, said humanity’s decline would start with the automation of ever more complex tasks removing their human element. From there, Gates said AI unchecked would be “strong enough to be a concern” and that he couldn’t “understand why people are not concerned”. An argument based on sound logic.
It’s a view shared by Elon Musk, who has invested $7 million in research grants, challenging researchers to provide ways to make sure AI’s don’t operate without oversight. Having labelled AI “our biggest existential threat”, his interest right now is focusing on practical and legal issues. For example, what happens if your robot car breaks a red light and causes an accident? Can software bugs become the next ‘twinkie defence‘?
Stephen Hawking’s concerns are more long-term in view, perhaps even more so than Gates’. According to the physicist, the ability of AIs to redesign themselves could be especially problematic.
On the other side of the fence are utilitarians like Ford and IHMC who are working with AI not as a peer or eventual replacement for the human intellect – but an enabling tool.
IHMC cemented its position at the forefront of robotics research this year with its Running Man, a robot based on Boston Dynamics’ Atlas, that took second place at the DARPA Robotics Challenge. At the competition teams of researchers used the Atlas to negotiate a mock disaster zone and perform eight tasks including navigating terrain, cutting holes in walls, opening water valves and even drive cars – all performed with limited input from an operator.
In an e-mail exchange Ford said he has been surprised by the ubiquity of AI technology and its spread into devices from smart phones to routers. This tallies with his own use of AI with the Running Man, where AI is used to complete tasks humans can’t – as opposed to don’t want to – do. When it comes to seeing AI as a competitor for humans he doesn’t share the downbeat assessment of Gates et al.
Replicants need not apply
“In my view it is certainly not about mimicking humans,” Ford says. “Most modern AI researchers do not set out to imitate human abilities, but to extend and amplify them.”
Ford sees the apocalyptic view of artificial intelligence as a kind of category error motivated by vanity that the human intellect remains a kind of gold standard for machines to replicate.
“Unfortunately, we have traditionally measured the success of AI systems by comparing them to human performance – which is rather like measuring the performance of aircraft against that of birds and complaining that aircraft do not land in trees or soil our automobiles,” he says.
“Pundits often talk as if our machines are engaged in a competition with the human race. One recalls the folklore story of John Henry and his race against a steam-powered hammer. Some futurist thinkers take this competition idea very seriously and worry that these mechanical rivals for intellectual dominance will soon take over our planet and treat us like domestic pets or worse.
“Even in fictional accounts of superhuman AI run amuck, the source of the hazard was often not that the machine was ‘too intelligent’ but that it was ‘too human’. For example, HAL’s design reflects AI’s old ambition to create an artificial human. However, simpler, more reliable and cost-effective methods exist for creating humans and they are not in short supply and arguably in excess. Rather than intelligent computers becoming our rivals or doing our thinking for us… they will (and have already) become our amplifiers and teammates.”
For the moment I’m more impressed with the utilitarian view which envisions more servile characters R2-D2 than existentially conflicted killers like Hal 9000 or the inquisitive Samantha from the movie Her. Either way, the discussion has to move beyond utopian/dystopian visions to something more mature.
Subscribers 0
Fans 0
Followers 0
Followers