The single biggest challenge to artificial intelligence is that governments, policymakers and regulators are struggling to keep up with new advances, as countries such as Russia and China compete to develop the technology.
‘We are moving from the industrial age to the information age to the age of intelligence,’ said Eric Schmidt, former Executive Chairman of Google’s parent company, Alphabet Inc., now technical advisor.
Some have talked about how technology has driven the rise of fake news, for example. But in a BBC webcast of a conversation in the museum with Professor Brian Cox, Royal Society Professor for Public Engagement in Science, Schmidt said he was ‘strongly in favour of free speech for humans, but not robots’ though, when it comes to putting that into an AI, ‘it turns out to be hard to write that rule.’
The rules for artificial intelligences should be that they must be used for the ‘benefit of the many, not the few,’ said Schmidt, adding ‘they need to be invented in the sunlight,’ rather than a military lab, and they need to be accountable for what they do.
To deal with competition from countries such as Russia and China, Schmidt urged governments to invest more in basic research. He paid tribute to Britain’s remarkable scientific heritage, as epitomised in the collections of the Science Museum. ‘Let’s have more…let’s celebrate it.’
The Tomorrow’s World live event in the museum’s Information Age gallery asked does it help us now and what might it mean for our imminent futures, from studying human intelligence to autonomous cars? In other words, what to think about machines that think?
Expert systems are being considered for potentially life or death situations, notably in autonomous cars, and they should be regulated, said Schmidt. But regulators have to start with the human response, and Schmidt pointed out that the current death toll from car accidents is huge.
He said that a large number of people are working on ethics and fairness in decision-making for AI but the issue they face is the lack of agreement on ethical choices.
With Brian Cox he discussed what lies ahead for AI, prompted by questions from the audience in the Information Age gallery, which also included luminaries such as Sir Venki Ramakrishnan, President of the Royal Society and Dame Mary Archer, Chair of the Museum’s Board of Trustees, along with fellow trustees Anthon Valk, Sharon Flood and Lopa Patel.
Would jobs be destroyed? ‘We are not changing jobs, we are changing tasks’, said Schmidt, explaining how you could see evidence of that from previous technological revolutions, such as the steam age, ‘when you walk around this extraordinary museum’.
As for what people want from the technology, Schmidt said the number one request was a ‘robot to clean up kitchen and put all the dishes away…that certainly makes sense to me.’ But it turns out to be an ‘incredibly hard problem.’
Google, one of the most influential tech companies of the 21st century, is exploring the possibilities of AI with projects such as DeepMind, which is has an ambition to develop more flexible artificial general intelligence.
They discussed DeepMind’s remarkable advances in devising the first computer program to defeat a professional human Go player, the first program to defeat a Go world champion, and arguably the strongest Go player in history, among other developments over the past year. One leading Go player described how the invention of a particular move by the AI ‘was the biggest drama of my life,’ said Schmidt.
Last year, DeepMind set up DeepMind Ethics & Society, a new unit carrying out interdisciplinary research that brings together the technical insights of its company with the people who could be affected by AI.
An explosion in the field of AI came from the rise of so called deep learning, neural networks loosely modelled on the brain that are trained – notably with web data – to do jobs such as image recognition. However, as Schmidt pointed out, ‘the underlying mathematics is not fully understood.’
This was a concern, said Schmidt: because the underlying rationale for making a decision is not understood, ‘you should not use these things for life-critical decisions.’ Equally, so called expert systems have huge potential in advisory and recommendation systems, while real time language translation is now possible. ‘That changes everything.’
Though Moore’s law (the observation that the number of transistors in an integrated circuit doubles approximately every two years) ‘is slowing down’, Schmidt raised the prospect of the next generation of so called quantum computing, expecting that ‘quantum supremacy’ would be demonstrated in the coming year.
He also raised the prospect of the neural networks that underlie machine intelligence being ‘automatically generated’. In other words, computers will generate the neural networks used for decision-making, since that job is becoming too hard for humans.
They pointed out some of the iconic objects around them in the gallery, such as Google’s ‘Cork board’ server (‘you turn it off, and the entire web search would go down’, said Schmidt) and Tim Berners-Lee’s NeXTcube computer, the original home of the world wide web (‘you turn that off, you turn the whole web off,’ joked Cox), which turned out to be a powerful platform to extend the use of information. ‘The web has become that,’ said Schmidt.
This was the fourth episode of ‘Tomorrow’s World Live’, after webcasts about missions to Mars, from the Science Museum, robots from the Museum of Science and Industry, Manchester, and on the future of the mind, at the Glasgow Science Centre.
The partnership was praised after the Tomorrow’s World webcast by Lord Hall, BBC Director General. ‘Absolutely wonderful,’ he told the audience.
Ian Blatchford, Director of the Science Museum Group, said that the Group is delighted with the partnership, honoured in the museum by the Tomorrow’s World gallery, and thanked the partnership for their work together, such as on the Tomorrow’s World hub. ‘We look forward to doing even more this year.’