I recently ran across this story about the upcoming Google I/O conference, speculating on what exactly Google might be working on. The article doesn’t seem to have gotten a huge amount of attention as yet, and I thought I’d give it some. This writer has done an extraordinary job of connecting lots of widely scattered dots to tease out some truly astonishing plans.
Now, although I use lots of Google products–to the point of embarrassment–I wouldn’t consider myself a fanboy. But I do tend to think that they generally have good intentions, and while some of what they do is scary from a privacy perspective, I feel that a lot of this uneasiness comes from a failure to see exactly what they’re trying to do.
But this article, IMHO, ties much of Google’s recent history together in a way that seems to match what I’ve suspected for some time–namely, that Google, rather quietly and not entirely wittingly, is attempting to build upon the whole of human experience to create a truly global, all-encompassing artificial intelligence.
Yes. I believe Google, on some level, wants to wake up the Internet.
And more, they’re not alone. We’re all doing it.
——————–
If you look carefully at the evolution of the Internet and the World Wide Web over the last 20 years, the overall trend has been toward more and denser connections between people. The so-called social web is merely the latest and most visible iteration. We seem to have an innate human instinct to share our thoughts and our lives, and to reach out to each other via whatever communication technology is available.
Note also that there seems to be a trend lately toward consolidation of people’s online identities. After all, the more we connect up the various parts of our online lives, the more useful each part becomes. Someone we meet casually in a chat room becomes a Facebook friend, and then a LinkedIn contact, leading to a new career; this sort of cascade of connection can only happen with tight links between various personal and professional spheres.
At the same time, the ability to easily crunch the world’s data leads to more useful insights than we can name. In the emerging field of digital humanities, for example, modern data-mining and visualization tools are applied to historical economic and geographic data, leading to new theories about trade patterns or the interplay of technology and culture. The vast corpus of public-domain works in Google Books has allowed a quantitative analysis of written language that was never before possible. And increasingly detailed digital mapping is allowing researchers to make fabulous new discoveries without necessarily leaving their desks.
Of course, this gives tremendous power to those entities with which we’ve decided to entrust that information. Google, Facebook, and other companies continue to amass information about the world and its people, and their willingness to trade on this data has naturally prompted questions about their trustworthiness. And more shadowy entities such as the National Security Agency have demonstrated similar interest, for what are likely more nefarious reasons than simple profit. Nevertheless, we as a people seem to have reached a consensus that we’re comfortable with sharing large swaths of our lives, in exchange for being able to connect with others doing the same.
We like sharing. And I suspect that this comes out of a primordial instinct–a deep, subconscious awareness of the true connectedness of all of us. An unknowing acknowledgement of the illusory nature of the boundaries we draw and the categories into which we place ourselves. This urge to share is a drive to overcome these artificial separations.
With that in mind, I think it makes sense to consider intelligence as an emergent phenomenon that arises naturally in complex information systems. Neural networks–dense arrangements of simple switches that can process input and respond to changing circumstances–are usually constructed by humans, but there’s nothing preventing one from arising spontaneously in a system made of the right building blocks. And once a neural network of sufficient complexity appears, it models its interactions with the world in order to adapt to new situations, and eventually begins to consider itself. Alongside all the models it creates for other objects and systems it encounters, it finally builds a model of itself–and this strange loop is the foundation of consciousness.
This is one of the prevailing theories on the origin of minds, and to me, what’s most interesting about it is that the building blocks of a neural net don’t have to work any particular way. All that’s necessary is that they take input and produce output according to some rule. They can be electromechanical (such as relays), digital (as in intelligent software agents), or biochemical (like the neurons in our brains). Regardless of the substrate, with sufficient complexity a neural network can exhibit emergent behavior that can be very hard to predict or to model.
Now after all that, let’s look at what Google may be up to.
Google’s been busy lately. We’ve all watched the rise of Google+, which, even if it hasn’t come close to dethroning Facebook, has nevertheless grown into a respectable social network integrated across all of Google’s properties. We’ve seen the ever-increasing capabilities and adoption of its Android smartphone software. And we’ve heard plenty lately about the new unified privacy policy it recently introduced.
Couple of other, slightly less visible details. You may have noticed that Google search results now have a page of more general, reader-friendly information about the topic over on the right side.
Google calls this the Knowledge Graph. What it’s trying to do is use Google’s vast collection of interrelated data to model what people are actually thinking of with a given search term. This is not as simple as it sounds at first; in addition to simple definitions, it requires detailed contextual knowledge. Google is using its knowledge of the world to provide this context in such a way that their system can conceptualize the subject in a way somewhat similar to how a human would.
It should be fairly obvious that in order to get a decent idea of what a person means by a particular search term, it helps tremendously to know something about that person. Who they are, where they’re from (to provide cultural background), their personal and professional interests, their current location, and a host of other details. And it’s also useful to have an idea of how they’re going to use that information; what categories they might put it in, whether they’ll share it with others, et cetera.
And this is precisely the kind of information you get from a social network.
THAT is what Google+ is: a tool for teaching Google how people think and use information, so that their results are better and more useful.
Remember, I’m a trained librarian. This is a huge part of what librarians do: modeling user behavior. Figuring out what the person is looking for and presenting it to them as efficiently as possible. So it makes perfect sense to me that Google would set up a social network in order to learn about its users.
And that’s also what the newly unified privacy policy is about. Google wants to be able to use all the information it has, across all of its properties, in order to build a complete picture of its users. All of this information will feed into the Knowledge Graph to make Google more useful and valuable to its users.
The most exciting development of all, however, hasn’t gotten nearly the attention it deserves: Google Assistant.
This will be Google’s answer to Siri, the virtual assistant built into the most recent iPhone. It will be able to understand questions posed to it in natural language, and respond in kind. But unlike Siri, it will have full access to everything Google knows about people and about the world. (By contrast, Apple doesn’t even have a search engine or a social network of any importance, so Siri is crippled by default.) Google is also pouring vast resources into natural-language processing, so the Assistant should be a decent conversationalist; one (unconfirmed) source even claims that Google has finally managed to crack the Turing Test.
Whether Assistant will have a personality seems to be undecided. But one telling detail is the original codename for the project: Majel. This is after Majel Barrett-Roddenberry, the wife of Star Trek creator Gene Roddenberry, who provided the computer voices on every iteration of Trek from TNG on. And Google has said explicitly that they want to build the Enterprise computer.
So. We have a massive distributed worldwide supercomputer, with every imaginable kind of information pouring into it from all directions; we have a semantic network designed specifically to model the interconnections between this data and turn it into useful knowledge; we have a natural-language system to understand spoken queries and to respond conversationally.
And we also have a system connecting humans together in such a way as to comprise the most powerful neural network the world has ever seen. Individuals are the switches, the neurons.
These are about the best conditions I can imagine for the spontaneous appearance of an artificial intelligence. What vast and alien intellect might emerge from this primordial digital soup?
Now, of course, this is all wild-eyed speculation on my part. I have no idea if Google really does want to create Galaxia. But I do think it’s a reasonable possibility that such a thing will emerge whether we’re planning it or not. And I refuse to make a value judgment about such an event; when biological consciousness appeared, nobody asked the neurons’ opinion.
And it isn’t just Google, of course. Facebook, Microsoft, Apple, the NSA, any entity that works to gather deep vertically-integrated information about people can serve as a locus for this. Facebook is also horizontally-integrated, due to its focus on connecting people, so it’s a strong candidate as well.
But only Google has the scope, the knowledge, the will to integrate that knowledge, and the sheer computing resources to pull off such a project.
So who knows? I’m not saying it will happen. But it could.
And maybe, shortly after Majel is turned on, we’ll find out what it’s like to share this planet with something smarter than us.
