Ginni Rometty, Chairman, President and Chief Executive Officer, IBM Corporation, leads a century-old company that has successfully reinvented itself and is now at the forefront of the revolution now on everyone’s lips: the advent of artificial intelligence (AI). In a fascinating interview conducted by Fareed Zakaria, host of Fareed Zakaria GPS, she spoke about IBM’s accomplishments in AI and the principles the company adheres to as it sails into these uncharted and challenging waters.
Three simultaneous technological revolutions, Rometty said, have combined to usher in this age of AI, a revolution that will surpass in impact and significance those that have preceded it. The three technological revolutions that laid the ground were the rise of cloud computing, which has made computing ubiquitous and accessible from anywhere; the rise of big data; and the rise of mobility. Underlying all of these, she cautioned, are still serious problems of security and privacy – issues which, left unresolved, might still derail the revolution the three have combined to create.
Watson, the IBM AI system that gained global fame when it beat two longstanding Jeopardy! champions, has quite literally been a game-changer. That it was able to win in Jeopardy! represents a quantum leap from its predecessor, Deep Blue, which famously defeated chess champion Garry Kasparov. Chess, after all, is ultimately deterministic – it’s possible for a computer to calculate every possible move. Jeopardy!, on the other hand, is an open-domain challenge. A player must be able to parse context. “This was not keyword search,” said Rometty.
However, the applications of AI go far beyond the merely trivial. Rometty made a powerful case for the ability of AI to solve some of humanity’s greatest universal problems: healthcare and education. In the world of healthcare today – with medical data doubling every 60 days and with 8,000 new papers being published daily – individual doctors simply cannot keep up to date. The world is awash in unstructured data, with petabytes of data being generated daily in text and tweets, sounds and sensor data, financial transactions and video. None of this can be processed through traditional computation either: we need learning and reasoning systems, capable of doing that which our human brains are cognitively incapable.
Watson, an open platform that IBM has made globally accessible, has been put to uses from the very mundane to the virtually unsolvable: uncanny personality tests based on writing samples; song writing based on analysis of popular songs; even movie-trailer editing (the results of which surpassed in popularity both human-edited trailers and the film itself); weather prediction systems; and expert systems for financial services providers to name but a few. However, it’s in healthcare that IBM has focused much of its own in-house efforts, and particularly on that “emperor of maladies”, cancer.
“We started with breast, lung and colon –the hard-body cancers,” said Rometty. “By the end of the year [Watson] will have figured out what causes 80% of cancers in the world.”
Watson first processed journals and textbooks, “reading” about 18,000 papers. It trained for tens of thousands of hours and was fed X-rays, MRIs and tens of thousands of patient records. Before long it was able to offer diagnosis and treatments. “You could put in limiters,” said Rometty, suggesting that patients or their doctors could exclude certain types of treatments. “People don’t want a black box,” she said, and noted that Watson gives patients a percentage indication of confidence and the sources of data that led to a diagnosis.
Rometty said there had been some resistance from doctors, but only at first and that, in almost all cases, the number of compelling examples doctor saw changed their minds. In a trial at the hospital of the University of North Carolina (UNC), Watson and the UNC tumour board – a panel of oncologists who all examine the same data and render opinions – looked at the same 1,000 cases and made the same decisions in nearly all cases. Watson, however, was able to find more than the tumour board in 30% of cases, a significant improvement on even a panel of trained oncologists.
Watson is also tackling genome sequencing to find gene sequences that may be responsible for cancers and offer the right diagnosis and treatment based on genetic data.
Between IBM and its ecosystem partners, Watson will touch 1 billion people in 2017. Persistent and, in some cases, well-founded misgivings about AI compelled Rometty to lay out a set of principles for IBM employees, which she shared during the session. First, that the purpose of the technologies will be clear; that is, they will be used in the service of humankind, to extend human capabilities and not to create self-aware or conscious systems. Second, IBM will be transparent about when AI is being used, about how the AI systems are trained, about who trained them and on what data sets. Transparency will also be applied in business models: Data being supplied by one company will not be repurposed to help a competitor. And finally, IBM will build skills to ensure the effective and safe use of the technology, and provide new skills training where required.