Perhaps because of the imminent arrival of the Apple Watch, a chorus of commentators have been clamoring — once again — about the pernicious effects of too much information finding its way to us. Notifications vibrating on our wrists while we are in meetings, walking down the stream, or trying to sleep.
That side of the atmospheric big data — the growing pressure from dozens of apps sending us all manner of notifications — might be a good subject for another post, and especially the need for AI to filter the barrage of notifications.
But this post deals with big data on the other side: the information about human activities lying latent in the world around us, that we have to go find. And in both cases, we are going to increasingly be relying on Bruce Sterling’s Engines of Meaning:
Ultimately no human brain, no planet full of human brains, can possibly catalog the dark, expanding ocean of data we spew. In a future of information auto-organized by folksonomy, we may not even have words for the kinds of sorting that will be going on; like mathematical proofs with 30,000 steps, they may be beyond comprehension. But they’ll enable searches that are vast and eerily powerful. We won’t be surfing with search engines any more. We’ll be trawling with engines of meaning.
Human behavior in aggregate has huge impacts on the world, like climate change, deforestation, urbanization, or highway infrastructure. But even at a finer granularity, human activities leave a residue of evidence, just like the game trails in a forest show what the deer and rabbits are interested in.
As one example, consider Orbital Insight, a company that analyzes aerial imagery of the Earth’s surface to winnow out inferences about what we are up to. This isn’t surveillance in the sense of tracking individuals, or attempting to determine if companies are breaking laws (although the EPA is doing that), but instead applying Engines of Meaning to the physical manifestations of human economics.
In the figure above, you see a single oil container from space, and the shadows have been identified by image recognition. The science that Orbital Insight applies allows them to discern — from the size and shape of the shadows — how much oil is contained in the container. And using today’s computing platforms, they can crunch the data latent in such imagery for tens of thousands of oil containers every day. The result? They can predict the price of oil next week, based on the demand for oil last week.
Likewise, they use similar techniques to watch the parking lots of 60 different retail chains, and to assess their results. They measure the density of cars in the lots, and using Google Street View — they guesstimate which stores people are likely to enter. And that’s why Sequoia and Bloomberg Beta have invested $8.7 million into the startup.
Bob Picciano, SVP of IBM Analytics, pointed out that the impact of IoT won’t be the Things, it will be the way the data can be analyzed to shape downstream innovation.
The goal is to be able to predict economic trends based on AI-derived information about human behavior at the macro-economic level that is hard to determine otherwise. Fairly soon, companies like Orbital Insight and Google — who acquired Orbital Insight competitor Skybox, last year — will soon be analyzing the numbers of cars on highways, transport ships in ports, trains carrying coal and oil, and barges shipping grain based on aerial imagery, and deriving a better sense of economic activity — in advance — than other techniques can yield.
Imagine the deluge of data we will be seeing in the very near term from major investments in the Internet of Things. This week, I attended the launch of IBM’s Internet of Things business unit, an ambitious, $3 billion investment over the next few years. Bob Picciano, SVP of IBM Analytics, pointed out that the impact of IoT won’t be the Things, it will be the way the data can be analyzed to shape downstream innovation.
Consider a manufacturer of next generation IoT-enabled washing machines, where information pulled from hundreds of thousands of units can be analyzed, and new firmware can be downloaded to change the way that the machines wash, and the way they interact with their users. This won’t involve a two to three year product design and development cycle, or even buying a new washing machine. IoT will completely change the users’ experience with the device and the company.
But aerial imagery and Iot aren’t the only source of such insights and innovation. Governments are beginning to push data that also can be mined. My friend Paul Kedrosky found out — around the time of the housing boom and bust — that California had started to freely release feeds of various sorts of information. For example, the California highway department would publish data about what stuff was being found on the state’s highways each week. The number of matresses was surprising. But Kedrosky plotted all the information and discovered what he came to call the Ladder Index.
Kedrosky first noticed a trend in the then-current few weeks, around the time he started to track the data: the number of ladders being found was dropping. He thought about it and realized that the number of ladders being found was, obviously, linked to the number of ladders strapped to vehicles traveling the highway. His insight was this: less ladders found meant less ladders traveling on the highways. And who uses ladders most? Construction workers!
Kedrosky pulled up some older data from the highway department, and found a bell-shaped curve — the Ladder Index — that perfectly predicted the housing boom and bust, with an approximately six month lead time. As construction was rising, more ladders found on the highway. And as the boom began to slow, less ladders. That data could have used to hedge the market, or to prepare for the bust.
Kedrosky pulled up some older data from the highway department, and found a bell-shaped curve — the Ladder Index — that perfectly predicted the housing boom and bust, with an approximately six month lead time.
Others are trying to crack the stock market, using similar techniques. Traditionally, most approaches to stock market analysis are based on looking at the historical prices of stocks, or other financial information provided by public companies. But today some funds are applying AI to look at other evidence of human activities outside the stock market — like sentiment analysis of what people are saying about products and brands on social media — to tap into buying behavior, and to therefore make wagers on Chipotle versus McDonalds, or Nike versus Adidas, or how many Apple Watches will be bought this month.
The day is here, already, where we are trawling with engines of meaning, and soon any other approach will be approximately 30,000 steps behind.