Some years back at a science conference in the Italian Alps, I met a man who has become a trusted source of information ever since. Signor Tavani speaks no English, and I speak no Italian, so our only face-to-face meeting was helped along by a multilingual friend. But these days I routinely get emails apprising me of interesting news items, all written in Tavanis native Italian and quickly translated by Google. Indeed, Google Translate ( translate.google.com) has become a daily tool, its sometimes clunky but helpful translations mediating many a conversation.
Google Translate has many languages to work with and plenty of computer horsepower behind it. Thinking about its methods reminds me that Ray Kurzweil has now gone to work for Google. Kurzweil is an Edisonian figure who came up with the first flatbed scanner and the first machine that could read text aloud. He has created music synthesizers and made huge strides in PC speech recognition, the sort of thing that allows us to dictate while the computer types.
In books like The Singularity Is Near, Kurzweil probes what happens when computers become so much smarter than us that they move beyond our powers of prediction.
Meeting your needs
Its fascinating to speculate on what Google might come up with given the aid of someone like Kurzweil, who for all the controversial aspects of his thinking has shown he has a way of producing results. We may get a glimpse of future direction in Google Now, an already functioning product that is in some ways a response to Apples Siri, the virtual assistant built into iPhones. Like Siri, Google Now can talk to you (assuming you have a properly equipped Android smartphone), and it can do what you ask, such as sending emails or looking things up for you.
But because smartphones are our most personal gadgets, Google Now can work all the information we pour into our phones in the way of contacts, searches, emails and calendar items into the database it builds about you. The plan is to create a software tool that anticipates your needs, coming up with whatever it thinks you need to know even before you ask for it. Thus it might know to optimize your travel time because it has learned your route. If it checks your calendar and sees you have an appointment, it might pop up the map you need to the location.
Google is at its best when it brings to bear the aggregated weight of the data it has collected. Couple that with the fact that Ray Kurzweils latest book is called How to Create a Mind (Viking, 2012). At Google, hes already making it clear that he would like to build a truly futuristic virtual assistant, taking the Google Now idea to its next level. Some artificial intelligence researchers scoff at his approach, but Kurzweil believes that basic thought processes can be modeled using a set of principles using pattern recognition that are then exposed to information by the exabyte. Like Google Translate, the idea would tap huge amounts of information and use brute force statistical methods to eventually awaken a far more powerful kind of machine intelligence.
I tend to scoff at this approach as simplistic, but Google Translate does, in its clunky way, continue to pay off for me every day, and both Siri and Google Now are downright useful in their own limited ways. I doubt that any of this will produce a machine with understanding comparable to a human brain, but it might turn out devices that can serve us as if they did. Will the day come when digital mimicry is so realistic that well no longer ask whether machines are self-aware?
Paul A. Gilster is the author of several books on technology. Reach him at firstname.lastname@example.org.