Would you trust a robot with your life?
First, let's put the Terminator to bed, sort of.
Dr Simon Kos is the chief medical officer of Microsoft's $10b plus worldwide health unit. He was in Rotorua recently as the keynote speaker at the Health Informatics NZ Conference, and in the refined environment of the Millenium Hotel Club Room, we're discussing the staggering advances in Artificial Intelligence (AI).
Stephen Hawking recently claimed AI could be the end of mankind. Could the technology go rogue?
"The short answer is yes," said Kos.
"It makes a great plot for a sci-fi movie and the reason I say yes is because in the early experimentation of some of this AI we had some of our own direct learnings."
Kos is talking about Microsoft's AI 'chatbot' Tay, which after being linked to Twitter did indeed go rogue.
He described Tay as a "garbage in garbage out" AI, and when the inputted 'garbage' included sexist and racist content, "soon Tay was starting to respond with this sort of material."
Now the good news.
"Extend that to the Terminator 'Judgement Day' scenario. Would you have an AI agent with the keys to nuclear weapons? (You) probably wouldn't."
Kos said the Tay experience proved the need for AI to have human checks and balances, something he said is quite a simple process.
"(It is) the junction between a future of computers going rogue and how we can shape the future we want."
Kos began his career as a frontline doctor in Australia in the 1990s, working in intensive care and emergency units. His decision to join Microsoft's health unit in 2010 was prompted by the growing digitisation of the medical world.
The second prompt can be traced back as far as the Third Century BC and the Hippocratic Oath instruction 'first do no harm'.
"If I projected out my clinical career I would have caused patient harm by not having the right information to make the right decisions," he said.
"I thought who's fixing that?"
One of Microsoft's current fixes is the AI Project Hanover. Kos said this was essentially a search system on steroids, using natural language processing to enable medical researchers to hone in on all the required research on a subject - he uses the example of a specific cancer gene.
"If you want to stay current in your medical literature, you should be reading about 28 hours a day. This is the conundrum modern clinicians face," he said.
Kos is optimistic about not just about the future, but now.
He said we already have robots that are conducting prostate operations with a far higher level of success than human surgeons, Project Hanover is speeding up research and a US project using AI to speed up retinal exams for diabetics has seen compliance rates for the yearly checks increase from between 33 and 50 per cent to more than 80 per cent.
Virtual reality and hologram technology is coming on in leaps and bounds too, he said. The surgeon in the US operating on the patient in Japan? "We're almost there."
There's a frustration too though.
Kos said the digitisation of medical information was creating an ever-expanding source of untapped knowledge.
"We don't have permission to break open that black box, look at the information and use it, but that's the gold mine sitting there. Your information could be the next breakthrough that leads us to the cure for cancer."
Kos said Microsoft was already "aggressively" lobbying European governments to get access to that gold mine, reassuring that proper privacy checks would be in place.
He's also slightly frustrated by what he refers to as the lag time - the gap between medical best practice being ascertained and going into frontline practice.
"X-rays on pregnant ladies. That persisted long after the evidence was out. My hope is that with AI reducing the latency between when we've got the findings and when it comes into clinical practice, we can start to get better patient outcomes."
He said the recent WannaCry cyber attack on the British National Health Service shouldn't dampen enthusiasm for technology use, but rather act as a wake-up call to keep systems updated.
He's aware of a the darker side of the tech space though, even referring to cases where building management systems have been hijacked and threats issued that hackers would turn off operating theatre lights mid-procedure.
It's big business, he said, with a black market existing for medical data.
"Nation state actors and national espionage, we absolutely see that. We track that through our digital crimes unit. I probably can't say much more than that," he said.
Overall however, Kos sees the benefits of the technology as far outweighing any risks, provided the human element remains.
"I very much see it as human augmentation rather than human replacement," he said.
"We're far from being redundant."
What are your thoughts?
Written by Benn Bathgate. Republished with permission of Stuff.co.nz.