Mar 15, 2019

Axios Future

Steve LeVine

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.

Okay, let's start with ...

1 big thing: AI unready

Photo: BSIP/Contributor/Getty

In two years observing surgeons in teaching hospitals, social scientist Matthew Beane noticed something troubling: doctors were finishing their residencies licensed to use robots in the operating room, but most were barely trained to do so.

At fault, Beane reported, is how hospitals have introduced machines and artificial intelligence to the workplace — a way that has left a large part of the new generation of doctors lacking crucial surgery skills.

  • Driving the news: In new research, Beane found that across high-skill occupations — in law enforcement, banking and more — the early age of applied AI and robots is leaving young professionals unprepared for their new jobs.

The big picture: Beane's results, though perhaps reflecting the type of growing pains experienced by most major new technologies, are a cautionary tale to companies as they tinker with robots and AI that some experts believe will fundamentally disrupt society.

The background: In surgery, you need four hands — the surgeon's own, plus those of a resident to pull and hold once an incision is made. Even six hands may be required — a second resident. But with DaVinci, the standard operating room robot, surgeons can manage procedures alone with hand and foot controls — and Beane found they typically do, mostly with the objective of efficiency and reducing mistakes.

  • "So the resident gets 10 to 20 times less practice," said Beane, a professor at the University of California, Santa Barbara. "Most residents by and large leave without knowing how to use this tool. They are licensed to use it, but not practiced."

Benjamin Shestakofsky, a professor at the University of Pennsylvania, tells Axios that he found a similar dynamic in software development.

  • For his dissertation, Shestakofsky spent 19 months with a San Francisco startup that, while creating a jobs website, outsourced contact with customers. Thus, no one at the company learned the core competency of using personal diplomacy to bring a blockbuster product to market.
  • "When you are creating something, customers using it are likely to have all sorts of reactions that the designers didn't anticipate," Shestakofsky said. "It's a real skill in large part learned on the job — how to interact with people."

The same thing is happening in low-skill occupations: Over the coming years, robots could take the jobs of tens of thousands of warehouse and other workers. But so far at least, few such workers have been trained as specialists to be teamed with the robots.

  • Jeffrey Brown, head of AI and future of work at the Bertelsmann Foundation, cited the example of Riverside County, California, a logistics hub for southern California that has more than a dozen "mega-warehouses" measuring more than 1 million square feet.
  • These warehouses in total employ roughly 15,000 people. But the introduction of robots will reduce that over the next decade or two to about 3,000 robot technicians and maintenance workers, Brown said.
  • "There is concern about skill atrophy. People are being shunted into work not requiring human competencies like working with clients."

Go deeper: In conventional surgery, too, many new surgeons are unready

Bonus: Outlaw apprentices

Illustration: Rebecca Zisser/Axios

In order to overcome their lack of operating room training and become proficient at robotic surgery, residents had to break the rules. They went rogue, learning on their own how to use DaVinci, risking losing everything, Beane wrote in a paper last year.

They reflect a tiny grouping of outlaw apprentices who, faced with leaving educational or training programs without requisite skills, skirt the rules and find a way to get their practice outside the bounds, says Beane. "They find ways to operate on patients without a supervisory surgeon in the room," Beane said. "The only way to learn was going outside the bounds of propriety."

  • Across professions, Beane said, legitimate methods of learning are being sidetracked by the introduction of AI. A few rebels are surfacing who learn the tools of the trade anyway. But most don't.
  • "We've been taking for granted the 5,000-year-old process where we learn by helping the expert," he said. "We are separating experts from learners in the name of efficiency. We are hamstringing future practitioners."
2. A dip in data science wages
Expand chart
Data: Glassdoor; Chart: Andrew Witherspoon/Axios

Companies in every sector are ravenously hiring data scientists, hoping to eke out more sales or improve their efficiency. Pay is good for the average senior data scientist, who makes nearly $140,000 a year and can pick from thousands of openings.

Kaveh writes: But for new entrants, just out of college or a tech bootcamp, the job market is rife with mislabeled job postings and stiff competition.

What's going on: Average wages for data scientists went down 1.4% in February compared to a year earlier, according to a Glassdoor report. There are a few factors behind the slip, according to the report's author, economist Daniel Zhao:

  • New supply of graduates has exploded, but entry-level job openings are few, and their pay is much lower.
  • Companies eager to bring in the best talent are describing jobs as data science when they are really data analysis or statistics roles. Since such work is generally lower paid, it is dragging down data scientists' wages overall.

Data from ZipRecruiter, another job site, confirms the gap between junior and senior data scientist positions.

  • Between 2017 and 2018, senior data scientist positions grew about 3X, according to ZipRecruiter. But the more junior "data analyst" and "associate data scientist" positions only grew 22% and 39%, respectively.
  • Wages for senior job titles grew fast, ZipRecruiter tells Axios, while wages for junior roles grew slower than those for data science jobs overall.
  • But, but, but: ZipRecruiter, unlike Glassdoor, found a big overall jump in data science salaries, which in February 2019 matched the all-time high.

It's lucrative work — if you can get it.

  • Glassdoor pegs data science as the fourth-highest paying job out there, after pharmacists, solutions architects — a technical job that involves setting up new processes inside a company — and attorneys.
  • All of this is attracting top talent from other fields.
"If you’re a physics Ph.D. who has spent years using advanced machine learning techniques, the data science and artificial intelligence fields are more attractive than ever. As these more experienced and more educated workers pour in, students fresh out of bootcamps, undergrad or even master's programs will have a harder time competing."
— Daniel Zhao, Glassdoor economist
3. What you may have missed

Soccer in Vietnam. Photo: In Pictures Ltd./Corbis/Getty

Have you been lost in Spring? No problem. Catch up on the best of Future this week.

1. Women and automation: An overlooked, bleak future of work

2. Zuckerberg cries uncle: Facebook's about-face

3. Living in a lab: Second-tier cities' fate

4. Training unlikely techies: A new model for reskilling

4. Worthy of your time

The future of gas stations (Tracey Lindeman — CityLab)

A speedier ocean carbon sink (Andrew Freedman — Axios)

Walmart's food delivery fight (Sarah Nassauer — WSJ)

Mind-blowing MLB pay (Eben Novy-Williams, Ira Boudway, David Ingold — Bloomberg)

The populist playbook (Simon Kuper - FT)

5. 1 robot thing: A wheelchair that helps you eat
Video: University of Washington

A robot built at the University of Washington can spear a carrot or a tomato, and gingerly lift it up for a person to eat without using their hands.

Kaveh writes: About 1 million American adults need help eating, according to census data. But there's more to feeding somebody than you might think. In experiments, the UW researchers gathered data as volunteers picked up various foods with a fork and fake-fed them to a mannequin.

  • Without thinking, people make all sorts of small adjustments based on a food's size, shape, and texture. Most stab a soft banana at an angle, for example, so it doesn't fall off, and skewer a hard carrot by wiggling the fork in.
  • The researchers taught their robot to identify several fruits and vegetables on a plate, and then pick the best place to pick one up: strawberries in the middle; longer carrots at an end.
  • Then, the arm lifts up the food and holds it near the person's mouth for them to bite it off.

Watch more videos of the robot here.

Steve LeVine