Do Smart Machines Make Better People?
Humans do two things very well. What happens when we supercharge both?
Most human achievements come from two things: spotting patterns and sharing information.In markets, people share the information of price movements to predict what they should build and sell next. Science proceeds by people figuring out patterns of Nature or Society, and sharing them for peer review and knowledge building.
Both pattern spotting and information sharing are about to be supercharged, respectively by Artificial Intelligence and global cloud computing.
Google, my employer, has done much work in newer types of A.I., like Machine Learning and Deep Learning, where machines figure out patterns, and patterns of patterns, to improve as they go, even predict what pattern to look for next. Other companies, like General Electric with its Predix platform, are pursuing industrial A.I.
Increasingly, there is also the sharing of data, in the application program interfaces, or APIs, that companies expose so software developers will collaborate with their assets to create new things. For example, eBay allows people to access its database to create custom interfaces for selling things on the eBay website. In the so-called “hybrid” environment of traditional and cloud computing, machines are sharing information to produce large-scale computational results.
Both A.I. and the new collaboration are in their earliest days, but not so early that we can’t start thinking about what this will look like, and how to work inside it.
“We’re going to build classes of systems that have this fantastic pattern-finding capability, this adaptability, this ability to feel responsive that will occasionally spook us out,” said Sir Nigel Shaw, an Oxford A.I. researcher with almost 40 years in the field. That doesn’t mean these machines have human-like minds, or minds at all, he added. “They are ‘smart,’ but not smart like us.”
I spoke with Sir Nigel at a World Affairs Council event in San Francisco last month. Besides his work in A.I., he and World Wide Web pioneer Tim Berners-Lee are responsible for The Open Data Initiative, a movement to make accessible to the general public large amounts of digital information. Open Data has put him at collaboration at scale.
The effects of collaboration in a new key — not just among humans, but machines as well — may ultimately be more impactful than A.I. itself. History, at least, suggests that.
Perhaps even more than industrialization, figuring out how to work together at scale has enabled an unprecedented boom in human prosperity. As the chart on world gross domestic product over the past several centuries shows, the dramatic growth happened not with the Industrial Revolution, but with the rise of mass communications and scientific management. (To learn more, Douglas Allen’s “The Institutional Revolution” is very good.)
Information organization and sharing has been accelerating rapidly at least since letters became email, and things like video and chat became a standard feature of daily life. Now the information we are sharing is bigger and moves at a greater velocity.
One example of this are the machine learning contests on run on Google’s Kaggle, where thousands of strangers across the planet compete to answer questions on things like improving passenger security, predicting house prices, or making Wikipedia work better. The collaboration is among teams, with algorithms, and with the community, to learn how to program better.
Then there is Grail, which aims to recruit tens of thousands of volunteers online to share personal health information, for genomic analysis and A.I.-based pattern-spotting at a scale and depth never before done, in search of better understanding and treatment for cancers.
Each new collaborative context also has new rules of behavior (for privacy, security, or even appropriate utility) that must be worked out. Sometimes that happens informally, as when Jimmy Wales declares there should be an online encyclopedia, and people collaborate on Wikipedia. Other types will be worked out among consumers, governments, and companies over many years. The future may not make it easier to be good — it never has before — but it is possible to start planning about the rules we’ll need.
For Sir Nigel, a life inside these two worlds has led to an appreciation of complexity — not just in A.I., but in the way we now work together to build products, systems, and rules of behavior — that borders on the mystic (if mysticism can be this cheerful.)
“Everyone knows something and nobody knows everything,” he said. “We are more deeply dependent on each other than perhaps ever.” In other words, collaboration counts like never before.