It doesn’t matter if AI is a mystery to you; you can still do your bit.

Roni Rosenfeld is a professor of computer science at Carnegie Mellon University. He leads the University’s machine learning department and Delphi research group. The latter has won kudos for its accurate predictions for the spread of the flu. Rosenfeld, at the behest of the Centers for Disease Control and Prevention (CDC), will now bring his phenomenal AI skills to mapping and predicting the spread of the coronavirus.

One of his favorite forecasting methods is to use the so-called “wisdom of crowds.”

These are inputs received from regular people using just a computer, their intelligence and some minutes of their time.

In an interview with Vox, Rosenberg talks to Sigal Samuel about his new responsibility, the use of machine learning and inputs from common people to map the likely advance of the deadly coronavirus. (Vox)

Concept: Wisdom of crowds

This forecasting method is more useful in tackling the coronavirus. That’s because machine learning needs a lot of historical data. For the flu, Rosenberg has data going back two decades, not so for the coronavirus.

In such a situation, a “non-mechanistic” approach, such as assessing the collective opinion of people on the ground across the country, maybe more successful.

“You gather at least several dozen people and ask them each individually to make a subjective assessment of what the rest of the flu season will look like,” says Rosenberg. “What we’ve learned from experience is that anyone of them on their own is not very accurate, but their aggregate tends to be quite accurate.”

Coronavirus: Data and modeling issues

Rosenberg plans to go for very short-term forecasts and to use “nowcasting,” a method of estimating the current prevalence of the disease.

For forecasting the spread of the flu he would use data sources such as social media mentions, Google (NASDAQ: GOOGL) search trends, access to Wikipedia and CDC, electronic health records and retail purchases of medications. Already tested AI models would use the data to map the extent of the disease, and thereafter, its likely spread.

The problem in extrapolating the procedure to coronavirus is two-fold. One, there are no models in existence. Two, the above data sources suffer from excessive noise and systemic bias. The bias is due to the systematic change in people’s behavior relating to the coronavirus pandemic.

Sigal Samuel: How are you going about making the needed adjustments?

Roni Rosenfeld: To start with, we’ve turned off all the data sources that have to do with social anxiety — Twitter (NYSE: TWTR), Google searches, Wikipedia page access. We’re down to short-term time series forecasting and wisdom of crowds focusing on just the first week. We find those are fairly adaptive.

How you and I can help

We can all volunteer to send in our inputs on the coronavirus at Crowdcast.

Rosenberg’s current view on coronavirus

“My expectation is that sometime in April or May we’re going to see a peak — unless we clamp down really strongly as some other places have done by sheltering in place,” says Rosenberg. “This is based on the contagiousness of this virus and on an assumption of moderate social distancing.”

However, he expects that the country will implement “more severe mitigation measures.”

Related Story:  Artificial Intelligence: The White House Spearheads Global Effort to Analyze Coronavirus Research Database for Clues