Smart machines start to navigate through flood of data for social good

When rising waters threatened the flood-prone region of Patna in northern India last September, residents received early warnings generated in part from artificial intelligence (AI) on their phones.

Flooding in the northeastern Indian state of Assam this month. 

Google, which is developing flood forecasting for India's Central Water Commission, said the results were "an encouraging first step" for the service itself and the so-called machine-learning behind it.

"For our high-risk alerts, we had less than 10 per cent false positives [down to regions measuring 64 by 64 metres]," said Sella Nevo, a software engineering manager with Google AI, and head of the flood forecasting unit. "That's highly accurate."

Minimising false alarms is key. "If we tell people there's a flood and it doesn't happen, they won't trust the alerts," he said, adding that unnecessary evacuations were also to be avoided.

Read the story from The Age by Peter Hannam - “Smart machines start to navigate through flood of data for social good.”