Kevin Kelly is a Silicon Valley mover and shaker, the founding editor of Wired magazine. In 2002 he was at a party with Larry Page, the co-founder of Google. This was long before Google’s IPO, before Google Ads or Youtube or anything like that. In 2002 Google was still small-fry.
Kelly was trying to work out Google’s plan to create a sustainable business. So asked Page, “Larry, I still don’t get it. There are so many search companies. Web search, for free? Where does that get you?”
Page answered: “Oh, we’re really making an AI [artificial intelligence]”.
At this point I like to imagine Kelly glancing over Page’s shoulder, wondering how to get away from this nut job.
Back then Google was a small company offering a search bar on a white screen. But now Page isn’t looking so crazy! In the last few years Google has bought no fewer than 14 AI and robotics companies.
And last week, it announced that it has turned over a “very large fraction” of its searches in the last few months to an artificial intelligence system, which Google calls RankBrain.
A learning machine
How does Rankbrain work? Well in a narrow sense you could say that it’s good at decoding language. Rankbrain has built a kind of giant mathematical model of our language.
But saying “it’s good at decoding language” kind of misses the point. RankBrain is unique because it’s capable of learning. It isn’t just a database – it’s an algorithm that teaches itself how to build a database. And it keeps building, and improving on it.
It learns through trial and error, by connecting unfamiliar words and phrases with familiar ones. In this way it’s able to “learn” new phrases over time and increase its accuracy.
The algorithm understands that words aren’t just independent bits of information. It treats them as fragments of an entire language. As Pedro Domingos puts it in his fascinating new book The Master Algorithm,
“The best way to understand an entity – whether it’s a person, an animal, a web page or a molecule – is to understand how it relates to other entities. This requires a new type of learning that doesn’t treat the data as a random sample of unrelated objects but as a glimpse into a complex network.”
So what does it actually do? RankBrain’s job is to help answer the searches Google has never been asked before (about 15% of queries, if you were wondering). And it appears to be working – out of the hundreds of signals that go into Google’s search algorithm, RankBrain is now the third most important.
Google’s giant AI (by 2024)
Writing in Wired last year, Kelly predicted all this. He sees this as just the beginning for Artificial Intelligence at Google:
“At first glance, you might think that Google is beefing up its AI portfolio to improve its search capabilities, since search contributes 80 percent of its revenue. But I think that’s backward.
Rather than use AI to make its search better, Google is using search to make its AI better. Every time you type a query, click on a search-generated link, or create a link on the web, you are training the Google AI.
When you type “Easter Bunny” into the image search bar and then click on the most Easter Bunny-looking image, you are teaching the AI what an Easter bunny looks like. Each of the 12.1 billion queries that Google’s 1.2 billion searchers conduct each day tutor the deep-learning AI over and over again.
With another 10 years of steady improvements to its AI algorithms, plus a thousand-fold more data and 100 times more computing resources, Google will have an unrivalled AI. My prediction: By 2024, Google’s main product will not be search but AI.”
I’m halfway finished with Pedro Domingos’ excellent book at the moment, and learning all about this stuff. It falls under the umbrella term of “machine learning” – basically, computer programmes that can learn and improve themselves over time.
Machine learning isn’t all about giant companies like Google. Hundreds of small teams in universities and in businesses are working in this area. I wrote last week that Software is eating the world. Well, machine learning is the next big breakthrough for software.
It’s important to the world and its investable. I look forward to telling you more about it.