Last June, an AI-infused IBM Research computer faced off (and more than held its own) against a seasoned human debater. At CES, IBM is taking this Project Debater concept even further, with a new experimental cloud-based AI platform called “Speech by Crowd,” in which text arguments are collected from large audiences on debatable topics, and automatically constructed into persuasive viewpoints to support one position or another.
CES attendees – as well as those of you not in Vegas – will be able to submit your most “thoughtful and inspiring” pro or con arguments on a given topic each day, the first of which is “Should gambling be banned?” If you’re not at CES, you can try it at home through an IBM website.
According to IBM, the system analyzes arguments for their polarity, strength and relevance to the topic. The system then selects the most representative arguments, edits them to improve fluency and remove repetition, and combines them into a concise and effective paragraph.
Could such a system help settle political arguments in a country that’s deeply divided?
Rometty says Project Debater doesn’t know right from wrong. But it might help a government decide on the merits of raising, say, school taxes? Or assist a consumer who wants to learn about the pros and cons of a new Apple phone. “When it comes to making any kind of decision this can help you do that.”
Predicting the weather
It is a mouthful. But IBM’s new Global High-Resolution Atmospheric Forecasting System, or GRAF for short, is a global weather forecasting system that uses IBM supercomputers and taps into millions of previously untapped crowd-sourced data to more accurately forecast the weather around by the world, by what IBM claims represents a nearly 200 percent improvement. IBM owns The Weather Company; the system launches later this year.
According to IBM, much of the world, outside of the U.S., Japan and some Western European countries, have to settle for weather predictions that cover 12- to 15-kilometer swaths of land, too wide an area to capture many weather phenomena. What’s more, such forecasts are typically updated only every six to 12 hours.
Under the GRAF system, predictions focus on more localized 3-kilometer areas and will be updated hourly.
The new model can draw on sensor readings from aircraft and may in fact help airliners better manage turbulence. “That’s the most important thing for us,” says Ed Bastian, CEO of Delta. “There continues to be a lot of volatility in the air and while we’ve done a lot of work ourselves with our meteorology group and our weather forecasting capabilities, to be able to access additional insight from their weather group would be fabulous. It winds up giving us the best information to have the smoothest ride possible for our customers as well as to minimize injuries. … Weather is a big deal for us.”
Meanwhile, IBM says consumers with the Weather Channel apps can also opt in to share barometric pressure sensor readings – IBM expects around 100 million readings per day – and benefit from these improved forecasts. Hundreds of thousands of weather stations, many from amateur weather enthusiasts, can also contribute data to the model.
“Weather is the number one external factor that impacts all business performance. It costs half a trillion (dollars) in economic value last year,” Rometty says.
The company has recently been weathering a small storm of its own related to its Weather Channel app. The City of Los Angeles filed a suit last week, alleging that IBM is inappropriately mining private data from the Weather Channel app and sharing that information with advertisers. IBM says it has been transparent with its data and will vigorously fight the suit.