All items about prediction

Links

Google introduces AI-as-a-service

Google’s recently launched learning engine tries to predict the future. The prediction engine takes data and tries to guess at outcomes. It’s not quite that simple: you have to supply the engine with a set of training data, and it will then try to predict new data based on what it’s learned, using a supervised learning algorithm.

By offering this as a cloud service, Google has removed an obstacle for many startups. Learning engines can predict everything from future purchases to suspicious behavior, but growing them as the data set expands can be difficult. The prediction engine can be built into applications running in Google’s App Engine, for example, making it easy to experiment with machine learning at scale. While the data is anonymous, Google does benefit from improved algorithms as it learns what works and what doesn’t.

Following on the heels of Google’s investment in Recorded Future, it’s clear the company’s mission goes far beyond putting the world’s information at our disposal. But even if Google can show the world the future, will we change what we do?

Links
Video

In The Shockwave Rider, his 1970s vision of a future that’s arriving faster than we can deal with it, John Brunner talks about Delphi Pools. These public, crowdsourced lotteries let citizens bet on predictions. The government uses this data to decide what’s most important to the population.

Break out your tinfoil hats: now Google and the CIA may be up to the same thing, having invested in a “temporal analytics” startup called Recorded Future last year.

According to Wired:

The CIA and Google are both backing a company that monitors the web in real time — and says it uses that information to predict the future.

The company is called Recorded Future, and it scours tens of thousands of websites, blogs and Twitter accounts to find the relationships between people, organizations, actions and incidents — both present and still-to-come. In a white paper, the company says its temporal analytics engine “goes beyond search” by “looking at the ‘invisible links’ between documents that talk about the same, or related, entities and events.”

Sentiment analysis is nothing new; what’s different here seems to be the visualization and extrapolation of past trends into the future.

Like Brunner’s Delphi, this helps guess what might happen, but rather than soliciting our input directly the way prediction markets do, this uses the trails we leave online — links, comments, retweets, and so on. The predictions can include competitive intelligence, brand monitoring, and personal investigation.

Incidentally, in Brunner’s novel, the government uses the Delphi pools to placate an otherwise implacable citizenry, and often alters the results before publishing them to sway public opinion.

Links

How Apple knows what you like

This MIT technology review explains how the iTunes Genius feature works, parsing millions of iTunes users’ libraries to generate suggestions and recommendations. One of the observations: we’re not unique and special snowflakes. No matter how individual you think you are, you’re part of a large online group.

Discovering the hidden or “latent” factors in your data set is a handy way to reduce the size of the problem that you have to compute, and it works because humans are predictable: people who like Emo music are sad, and sad people also like the soundtracks to movie versions of vampire novels that are about yearning, etc. You might think of it as the mathematical expression of a stereotype–only it works.

Recommendation engines are notoriously difficult to get right, as the Netflix Prize proved.

Powered by WordPress, based on Mina theme.