Technology Review
Providing developers with machine learning on tap could unleash a flood of smarter apps.
From Amazon's product recommendations to Pandora's ability to find us new songs we like, the smartest Web services around rely on machine learning--algorithms that enable software to learn how to respond with a degree of intelligence to new information or events.
Now Google has launched a service that could bring such smarts to many more apps. Google Prediction API provides a simple way for developers to create software that learns how to handle incoming data. For example, the Google-hosted algorithms could be trained to sort e-mails into categories for "complaints" and "praise" using a dataset that provides many examples of both kinds. Future e-mails could then be screened by software using that API, and handled accordingly.
Currently just "hundreds" of developers have access to the service, says Travis Green, Google's product manager for Prediction API, "but already we can see people doing some amazing things." Users range from developers of mobile and Web apps to oil companies, he says. "Many want to do product recommendation, and there are also interesting NGO use cases with ideas such as extracting emergency information from Twitter or other sources online."
Machine learning is not an easy feature to build into software. Different algorithms and mathematical techniques work best for different kinds of data. Specialized knowledge of machine learning is typically needed to consider using it in a product, says Green.
Google's service provides a kind of machine-learning black box--data goes in one end, and predictions come out the other. There are three basic commands: one to upload a collection of data, another telling the service to learn what it can from it, and a third to submit new data for the system to react to based on what it learned.
"Developers can deploy it on their site or app within 20 minutes," says Green. "We're trying to provide a really easy service that doesn't require them to spend month after month trying different algorithms." Google's black box actually contains a whole suite of different algorithms. When data is uploaded, all of the algorithms are automatically applied to find out which works best for a particular job, and the best algorithm is then used to handle any new information submitted.
"Getting machine learning to a Google scale is significant," says Joel Confino, a software developer in Philadelphia who builds large-scale Web apps for banks and pharmaceutical companies, and a member of the preview program. He used Prediction API to quickly develop a simple yet effective spam e-mail filter, and he says the service has clear commercial potential.
Software lets a hybrid car learn your habits and optimize its power usage.
The massive amount of real-time information available is leading to new programming approaches.
The company has many social projects, but may struggle to improve on Facebook.
Does that make any sense, how could they not look at the input data to improve the algorithm, it sounds like a necesary point of reference. Where would I be able to find an example or explanation that says otherwise? Surely engineers will switch algorithms based on qualitative interpretation of the data. Supercomputer data crunching- Eager Beaver says spank you mam, may I please have access to your data.
I believe they want to avoid privacy concerns by saying that they don't look at data, but they probably do look at how it is distributed in order to learn. It is a good move by Google, although the service is quite dumb, trying all the methods in parallel and picking the best one.
Which business model do you think will work? Those that offer free access or those that offer free to premium (freemiums)?
Microsoft should be getting worried about Google’s free stuff:
http://bit.ly/bc6FBW
Has anyone considered the long term implications of these technologies for the automation of knowledge-based jobs? The conventional wisdom is that everyone needs to go to college to be trained for a high skill job. Between machine learning technology and offshoring it seems to me there will be a big impact on employment in the future.
Check out this book: <em>The Lights in the Tunnel: Automation, Accelerating Technology and the Economy of the Future</em> (free PDF at http://www.thelightsinthetunnel.com). It gives a very thoughtful treatment of these questions.
Also checkout the blog at http://econfuture.wordpress.com.
I really think this is an issue that we all need to start thinking about...
But can it pass the 1/P, 2/FM, and 4/C ?
It probably could pass these qualifying exams for actuaries, but you'd have to build a parser that could reliably turn the exam questions into problems it could solve.
Since it's performing an essentially actuarial function, however, you'd have to expect it to at least be able to at least pass the qualifiers.
Artificial Intelligence Eventually
With the pace of googles technology and information aggregation they are not to far off on future linguistics...this just brings it one step closer.
http://mycorporatemedia.com
Manufacturing in the United States is in trouble. That's bad news not just for the country's economy but for the future of innovation.
National Instruments has gathered customer information and data regarding some of the cost differences between building a custom solution versus using NI off-the-shelf tools. Using this data, we built the Graphical System Design ‘Build vs. Buy’ Calculator. The calculator can help show the financial differences between building a custom solution versus buying an off-the-shelf system. This paper discusses the benefits and drawbacks of both a traditional custom design approach and off-the-shelf embedded tools.
View full PDF >
Listen to story > 
Our list of the 50 most innovative companies, including the following:
On Twitter
Become a Fan on Facebook
Subscribe to the Feed
mjaniec
12 Comments
Black boxes for business organizations?
As machine learning and other pattern-recognition technologies become more available and effective, companies will use them more often.
We can learn some lessons from the investment area, where algorithmic trading has already overtaken human discretion at least in terms of number of transactions executed on the US stock exchanges.It's true that quite often trading algorithms are still quite simple, but they are getting better as Moore's law increases the computing power available for them. It seems inevitable that one day algorithmic trading will dominate the investment industry, and human investors will share the fate of chess master Garry Kasparov defeated by IBM's Deep Blue.
Similarly, proliferation of quantitative methods in business may lead to some interesting results. First, managers will be often and often required to know these new technology-enabled methods. Second, the decision making processes, starting with operational aspects, may slowly morph into hardly penetrable black boxes, processing increasing amounts of information generated and gathered by business organizations.
Maciej Janiec
blog.inlevel.com
Reply