CEO DailyCFO DailyBroadsheetData SheetTerm Sheet

Why artificial intelligence is so important in the coronavirus era

April 14, 2020, 3:42 PM UTC

This is the web version of Eye on A.I., Fortune’s weekly newsletter covering artificial intelligence and business. To get it delivered weekly to your in-box, sign up here.

I’m cautiously optimistic about artificial intelligence’s role in helping manage the coronavirus pandemic.

Google’s DeepMind unit, for example, is investigating deep-learning techniques for modeling the structure of the virus’ proteins, which could be useful in developing a vaccine. Meanwhile, the White House has asked researchers to develop machine-learning techniques to quickly analyze nearly 30,000 coronavirus-related studies to better understand the deadly virus.

Despite the number of promising projects, however, none of their A.I. is ready to be widely used today. It will likely take months or years more until the technology is ready to provide tangible results.

“I haven’t seen anything in which A.I. has helped us yet, clinically,” said Eric Topol, the founder and director of the nonprofit Scripps Research Translational Institute. He recently published the book Deep Medicine, which chronicles recent advances in A.I. and healthcare, among other topics. 

Topol is a big believer in A.I.’s increasingly important role in the medical industry. Using A.I. to mine drug databases and discover effective coronavirus (and other illnesses, for that matter) treatments could be helpful, but he has yet to see any specific technique that have advanced from research to a clinical setting.

Another promising area is using A.I. to crunch health-related data gathered from wearable devices like smartwatches or Internet-connected thermostats. Topol, for instance, mentioned the startup Kinsa Health, a seller of smart thermometers that has analyzed data from its products to identify the location of coronavirus hotspots in Florida. 

Topol’s own research team is conducting an A.I.-related coronavirus study based on heart rate data from smartwatches. Although his team doesn’t have enough data yet, he’s hopeful that once his team gets information, they can use A.I. techniques to find regions where people’s heart rates appear to be increasing during resting, a possible sign that they have fevers. That could ultimately show that Covid-19 is growing in a particular community, Topol explained.

“The analytics of that data is purely an A.I. story,” Topol said, explaining that his team will use neural networks—software that learns to identify patterns in huge quantities of data—to crunch the numbers. Because these studies rely on an immense amount of data, neural networks could be crucial to parsing all that information, he explained.

Whether his study will pan out is unclear. But whatever Topol and other researchers learn from their current A.I. studies won’t be for nothing because another pandemic may be around the corner. 

Jonathan Vanian 
@JonathanVanian
jonathan.vanian@fortune.com

A.I. IN THE NEWS

Calling all A.I. researchers interested in energy. The U.S. Department of Energy said it would offer up to $30 million in grants for researchers specializing in using machine learning for predictive modeling and for the task of “decision support” in complex systems, like managing power grids. “This foundational research will help keep the United States in the forefront as applications for ML and AI rapidly expand, and as we utilize this evolving technology to solve the world’s toughest challenges such as COVID-19,” said Under Secretary for Science Paul Dabbar.

Arming against adversarial attacks. Intel and the Georgia Institute of Technology will help lead a DARPA program focused on researching techniques to safeguard A.I. systems from adversarial attacks, which can cause A.I. systems like image-recognition services to misidentify images, among other failures. Intel said it “is the prime contractor in this four-year, multimillion-dollar joint effort to improve cybersecurity defenses against deception attacks on machine learning (ML) models.”

The surveillance economy now includes pets. Several pet devices like streaming cameras are now including machine learning technology to help understand animal behaviors, The New York Times reported. The Times report said that the Furbo company will soon “roll out a new feature that allows it to differentiate among kinds of barking and alert owners if a dog’s behavior appears abnormal.” Furbo used machine learning to study “the video data of thousands of users,” the report said, noting that “Furbo also allows users to opt out of sharing their data.”

Are your kids bored at home? Teach them A.I. Entrepreneur and Dallas Maverick sowner Mark Cuban said his foundation is partnering with the non-profit A.I. For Anyone on a free, hour-long online course intended to introduce students and teachers to the basic fundamentals of A.I. “Parents, want your kids to learn about artificial intelligence while you're stuck in quarantine!” the billionaire said in a LinkedIn post.

SOMETIMES, YOU DON'T NEED A.I.

An interesting post published on Medium’s Towards Data Science vertical examines several ways researchers are exploring A.I. to help manage the coronavirus pandemic. One of the takeaways is that established statistical modeling is more useful than deep learning at forecasting the spread. From the post:

As a result of a lack of data, too much outlier data and noisy social media, big data hubris, and algorithmic dynamics, AI forecasts of the spread of COVID-19 are not yet very accurate or reliable. Hence, so far, most models used for tracking and forecasting do not use AI methods. Instead, most forecasters prefer established epidemiological models, so-called SIR models, the abbreviation standing for the population of an area that is Susceptible, Infected, and Removed.

EYE ON A.I. RESEARCH

All you ever wanted to know about A.I. chips. Georgetown University’s Center for Security and Emerging Technology published a deep dive into the burgeoning market of A.I. chips. The paper analyzes the state of the semiconductor market and breaks down the different types of A.I. chips that are currently in use or are being developed: graphics processing units (GPUs), field-programmable gate arrays (FPGAs), and application-specific integrated circuits (ASICs). One of the reasons companies are interested in A.I. chips is because they want to lower the high-costs of training certain deep-learning projects. From the paper:

First, training costs of AI lab DeepMind’s leading AI experiments, such as AlphaGo, AlphaGo Zero, AlphaZero, and AlphaStar, have been estimated at $5 to $100 million each. One cost model suggests AlphaGo Zero’s training cost was $35 million. AI lab OpenAI reports that of their $28 million total 2017 costs, $8 million went to cloud computing. Multiplying these computing costs by thirty for trailing node AI chips, or even more for leading node CPUs, would make such experiments economically prohibitive. And computing costs for some AI companies have increased so quickly that a cost ceiling may soon be reached, necessitating the most efficient AI chips

FORTUNE ON A.I.

This self-driving tech company’s balance sheet is unaffected by the deep recession—By Adam Lashinksy

Former Google CEO: The coronavirus pandemic will make Big Tech even bigger—By Danielle Abril

Does 5G cause or spread the coronavirus? Here’s what experts say—By Aaron Pressman

Red Hat’s New CEO on IBM, the cloud, and coronavirus—By Jonathan Vanian

BRAIN FOOD

Silicon Valley’s “top foe” in Europe. Margrethe Vestager, the executive vice president of the European Commission (the executive branch of the European Union) is the subject of a short profile by Bloomberg News that highlights Vestager’s role in overseeing the E.U.s proposed A.I. regulations. As the report details, Vestager is sticking to her belief that A.I. needs some form of regulation, “even as companies urgently rush out various AI solutions—from diagnostics to surveillance—in an effort to slow the spread of the novel coronavirus.”

The regulatory guardrails that Vestager sets on AI, illegal content, and more won’t be designed to absorb companies’ trust problems but rather to hold them to account. In her view, it’s a joint responsibility belonging to regulators and platforms, she said in an interview with Bloomberg earlier this year. “I don’t think that one can go without the other.”