‘We have to be the supervisors of robots’: HPE’s Dr Goh
- 05 June, 2018 11:41
Academics predict that half of our jobs will disappear soon thanks to the automation provided by technologies like artificial intelligence (AI) and robotics.
Which jobs will ultimately be replaced is still up for discussion but humans will reach a stage in the AI world where we will have to “differentiate between a correct decision and a right decision,” according to Dr Eng Lim Goh, vice president and SGI chief technology officer at HPE.
During a Q&A session following his presentation on the impact of AI on society last week at Melbourne’s Luther College, Dr Goh said using historical data, a machine will reach the optimised and correct decision.
“[But] we have to be purveyors of the right decision because some correct decisions are not right socially,” he said. “We still have to go through growing up, playing in the playground with other kids, studying, going to school, going to wine parties and having relationships.
“[During this time] we start to observe and understand the human values system because after all, we are humans [who] have created robots. Like [we did during] the past industrial revolution, we have to be the ‘supervisors of robots’, differentiating between correct and right. Most of the correct decisions are the right decisions but some of them might not be,” he told the audience.
For this reason, Dr Goh stressed that in addition to studying sciences, the humanities will continue to become important in the future.
“Economics and social sciences, history – all of these will continue to be important because that could be our ultimate differentiation; the human side of humanity,” he said.
Making machines ‘self-aware’
To create AI systems that are sentient and self-aware within the next 10 years, we will need to achieve ‘general intelligence,’ said Dr Goh.
“[The question is], how do you get to general intelligence when all you are building today are individual, specific intelligence systems?’ It [the system] plays chess well but doesn’t know anything about driving cars. It classifies images really well but it knows nothing about speech recognition.”
According to Dr Goh, after we have combined 10,000 specific intelligence systems and link them across their neural network, something new may emerge.
“Why do I believe this is possible? It’s because of a Nobel laureate at Princeton University who explained the following: physics does not explain everything in chemistry – it cannot. Everything in chemistry cannot explain biology. Biology cannot explain physiology and physiology cannot explain psychology. That’s the reason why at university you have a degree for every one of them.”
“And he calls this an ‘emergent property’ to try and explain the gap. When you have something more complex above, everything more basic below cannot explain it because something emerges with complexity,” said Dr Goh.
“So if you took 10,000 different intelligence systems, wired them across, will you get an emergent property? I don’t know. That’s a possibility and people are thinking about it.”
Follow Byron Connolly on Twitter: @ByronConnolly