In the race for AI supremacy, there are few milestones more arresting than the development of the Summit supercomputer housed at the US Department of Energy's Oak Ridge National Laboratory in Tennessee.
In June, it dethroned a Chinese supercomputer as the world's fastest, clocking up speeds of 200,000 trillion “floating point operations” – or petaflops – per second.
The record had been held by China since 2012, when its TH-2 supercomputer unseated Summit's predecessor – Titan – as the world's most powerful.
The relentless drive to eclipse rivals in computing power – which also includes Japan and the European Union as credible contenders – underscores how vital the new frontier of quantum computing will be in developing the AI technologies that will transform our world. AI and supercomputers are already being used in industry to give companies the edge. And the faster the supercomputer the more agile it will be at crunching data to yield intelligent solutions – and at identifying issues that may themselves only be anticipated by the fastest supercomputers.
The benefits of many minds
Yet amid the whirlwind of speculation about a forthcoming AI arms race, it's easy to overlook a critical factor that could shape – and possibly salvage – the ambiguous adventure on which we are hurtling (or being hurtled) headfirst. Even as competition intensifies, it is global co-operation that will unleash the full potential of AI, and also help to ensure that intelligent machines become forces of good that will elevate humanity rather than destroy it.
Innovation comes from the free flow of ideas – and the more diverse the pool of perspectives, the better the solutions that arise in complex endeavours. The global consultancy firm McKinsey says that countries that promote open data exchange reap the greatest rewards in AI innovation.
Similarly, one of the main fears of AI is around ethics – in particular how to prevent biases from creeping into machine-learning systems – and a universal ethical framework for AI requires global coordination. “The only really effective form of AI regulation is global regulation,” historian Yuval Noah Harari recently told the New York Times. “If the world gets into an AI arms race, it will almost certainly guarantee the worst possible outcome.”
Better together
AI insiders are increasingly coming around to the notion that global co-operation in building AI systems will be a liberating force rather one that hinders or sabotages creative processes. “While each country must consider its specific needs, there is early indication – and early promise – that a global framework may actually create more innovation and allow us to solve more complex and pressing global issues,” according to Ganesh Bell, president of US-based AI software firm Uptake. “Many of the questions we are grappling with today will be answered if we take the time to learn from one another.”
Consider Dinglong Huang, a Chinese AI innovator from Shenzhen, and Matt Scott, an American machine-learning developer from New York. Together they founded Malong Technologies, a start-up specialising in computer vision technology backed by Accenture and SoftBank with offices in China, the US and Japan. Writing recently for the World Economic Forum, the partners argued that the most significant advances in AI have been the fruit of cross-pollination of ideas from researchers across different backgrounds.
For example, it was Rick Rashid, a Microsoft Research scientist from Iowa, who demonstrated the first real-time Chinese-English robotic translation in Tianjin in 2012 – with technology developed by a team from the US, China, the UK and Germany. “The AI community is global,” wrote Huang and Scott. “We do our best work when we work together across boundaries.”
Realising AI's potential
Rather than an AI arms race focused on a single prize of global dominance (Russian President Vladimir Putin recently predicted that the nation that leads in AI “will be the ruler of the world”), we are increasingly seeing that societies at the forefront of AI are developing their own unique strengths. Complementarity could drive the AI narrative as much as – if not more than – competition. Singapore has become a pioneer in testing autonomous vehicle technologies. Israel is becoming a leader in AI medical innovation. Japan's robotic expertise and aging demographics give it an edge in AI technologies for elderly care. The European Union is well placed to spearhead AI innovations for protecting privacy.
Complementarity opens significant opportunities for global collaboration in creating optimal AI systems that will enhance life rather than imperil it. And as countries begin to depend on each other's unique areas of specialisation, a system of checks and balances may arise that discourages zero-sum approaches and reinforces the cycle of collaborative innovation.
How fast is the Summit supercomputer? So fast that each person on earth would need to perform a calculation every second for 305 days to crunch what the machine can do in one second. No doubt global players will continue striving to outdo one another in creating ever more dazzling machines – and so they should.
But the power will fall short of its promise – and perhaps lead to tragedy – without the combined vision of humanity to channel these awesome forces into technology that benefits all.
