May 29, 2019

Axios Future

By Bryan Walsh
Bryan Walsh

Have your friends signed up?

Any stories we should be chasing? Hit reply to this email or message me at steve@axios.com. Kaveh Waddell is at kaveh@axios.com and Erica Pandey at erica@axios.com.

"Axios on HBO” Season 2 will feature exclusive interviews with Sundar Pichai, Jared Kushner, Lisa Monaco, Janet Napolitano, Leon Panetta and more. Tune in Sunday 6 pm ET/PT!

Smart Brevity count: 1,111 words/<5 min. read

Okay, let's start with ...

1 big thing: Medical AI's big data problem

Illustration: Rebecca Zisser/Axios

Facing increasingly overworked doctors and labyrinthine insurance systems, hospitals are searching for a lifeline in AI systems that promise to ease hard diagnoses and treatment decisions, Kaveh reports.

But the data underpinning the very first systems is often spotty, volatile and completely lacking in critical context, leading to a poor early record in the field.

The big picture: Basic clinical decision support (CDS) systems have been around for decades, but a skepticism of technology leads many doctors to ignore or override them. Now, experts say a nascent generation of CDS — infused with AI in academic labs and startups — may reduce the estimated 40,000–80,000 deaths a year that result from medical errors.

  • The grand vision: Researchers hope AI programs can point doctors toward the best medications, lab tests or treatment plans based on minute patterns discovered in huge numbers of patients' past experiences.
  • Last week, we reported on the promise of combining pools of private data to strengthen AI systems, feeding them with ever more examples of past outcomes.
  • This helps solve the quantity issue. But data quality — a constant struggle in health care — remains an enormous threat to medical AI.

The big problem: Record keeping is so bad that doctors laugh when you ask about it.

  • Electronic medical records, central to CDS predictions, are notoriously error-ridden. Doctors fill them with generic diagnosis codes, and 82% of any given record is likely to have been copied or imported from elsewhere.
  • "Filling out documentation right is something that most physicians don't care about," says Jonathan Chen, a doctor and professor of biomedical informatics at Stanford.

Other quirks of health data make more problems for CDS systems:

  • Records reaching back even a year or two become useless for predicting future trends because of how quickly treatments shift. A study led by Chen found that medical data has a usefulness half-life of just four months.
  • Hospital readmission rates are a gold standard for measuring whether a particular treatment worked, but the statistic misses more important outcomes like patient happiness and long-term health.
  • Together, all this means that if doctors and researchers are not careful, "you'll end up learning patterns that are either obvious or you'll completely misinterpret what they mean in potentially dangerous ways," says Chen.

It's the oldest problem in data science: garbage in, garbage out.

  • "We are at a decided disadvantage because our core electronic record is so pitiful," says Eric Topol, director of the Scripps Research Translational Institute.
  • The messy reality of medical records tripped up IBM's much-ballyhooed Watson AI system when it was deployed at a Texas cancer hospital. "[T]he acronyms, human errors, shorthand phrases, and different styles of writing" were too much to handle, Stat News reported in 2017. An IBM executive today blamed data quality for past AI failures.

What's next: The Food and Drug Administration, which currently doesn't review most CDS systems, is considering policy changes that could head off some data issues. Scientists are pushing the agency to impose strict benchmarks and audits to prevent mistakes.

Go deeper: What your hospital knows about you

2. When Pop-Tarts sell best

Shopping in the rain. Photo: Brandi Simons/Getty

The variability of weather costs American firms $600 billion a year — and there's a massive market for AI systems that can help companies bring down those big costs, Erica writes.

Driving the news: A seldom-discussed but lucrative application of AI is to predict how weather will impact businesses, such as retailers, farms and hotels.

  • Betting that insights about weather patterns can slash billions in costs a year, IBM is amassing mountains of sales data from retailers, combining it with weather events and patterns, then selling that back to companies.
  • "Weather’s toll on retail is bigger than it ever has been," says Paul Walsh, IBM's global director of consumer weather strategy. "We’re seeing a 100-year event every year."

The big picture: Insights about the weather can save retailers money by helping them protect their supply chains from storms and floods, and they can also boost sales by clueing the companies in on what to stock the shelves with in different conditions.

Worth noting: Bigger retailers like Walmart and Target are already tracking weather and using it. "They have armies of data scientists," Walsh says, so IBM is attempting to target small and medium-sized stores instead.

For these stores, some decisions — like putting snow shovels at the front of the store the night before a nor'easter or buying up twice as much sunscreen to sell over July 4 weekend — are intuitive and don't need AI's input.

But others are not so obvious and may help drive sales if companies know what products to stock up on ahead of certain weather conditions.

  • Pop-Tarts tend to fly off shelves before hurricanes, according to IBM data.
  • Chocolate sales spike when skies are cloudy.
  • Soup, a perfect winter meal, does particularly well during unseasonably cold springs as well.
3. About that last 10% ...

A Waymo driverless car. Photo: Glenn Chapman/AFP/Getty

If you ask driverless car mavens, they will tell you that developers are 90% of the way to creating a fully autonomous vehicle ready for commercial use.

  • What they don't say is that, when it comes to new car technology, engineers spend 90% of R&D time on the last 10%, reports the WSJ's John Stoll.
  • Almost no one reasonable thinks fully driverless cars, capable of operating with no steering wheel, will be ready before the 2030s, and a lot of folks think it will be longer.
  • GM president Mark Reuss tells the WSJ: "This is quite frankly probably the hardest engineering problem in our time."
4. Worthy of your time
Expand chart
Data: CDC; Chart: Axios Visuals

The story of refuse (Ziya Tong — Wired)

Measles is in half the U.S. now (Eileen Drage O'Reilly — Axios)

Bannon's chippy politics travels to Kazakhstan (Natalia Antelava — .Coda)

The great jobs boom (The Economist)

Workers without degrees are fleeing superstar cities (Eduardo Porter — NYT)

5. 1 name thing: Japanese surname diplomacy

Yesterday, onboard Japan's ship Kaga. Photo: Charly Triballeau/Pool/Getty

A kerfuffle appeared to surface leading up to President Trump's visit to Japan — over how Japanese names are rendered by Westerners, reports the NYT's Motoko Rich.

In international forums and the Western press, Japanese names are routinely presented backward, ,with the family name last rather than first, some senior Japanese officials complained. And they asked that it stop.

  • One of the main complaints was registered by Japanese Foreign Minister Taro Kono, who told foreign media to start calling the prime minister Abe Shinzo, rather than the opposite.
  • Trump left Japan yesterday and it wasn't clear how his retinue addressed Abe.
  • But the request may have had almost no immediate impact in the media —the NYT, for instance, till the end of the visit was still unapologetically reversing Abe's name.

Rich wrote: "The New York Times generally writes Chinese and Korean names with surname first, while using the Western order for Japanese names — although its general policy is to render people's names the way they prefer."

Bryan Walsh