AI Talk: Ghosts, burnout and Pokémon Go to Sleep

June 7, 2019 / By V. “Juggy” Jagannathan, PhD

This week’s AI Talk…

Ghost work – book review

I am frequently on Amazon buying some book or other and they know what I read! Their recommendation system came through with this book published last month: “Ghost work: How to Stop Silicon Valley from Building a New Global Underclass,” by Mary L. Gray and Siddharth Suri. It is ironic that Amazon recommended a book that is critical of its practices! I bought, as usual, an audio and kindle version and sped through it at 2x speed. It is definitely worth reading – not necessarily for style, but for the research content (here is a link to a good review of this book ). First the origins of this work. AI researchers needed a lot of training data to get the recommendation algorithm to work. Someone had to tag thousands and thousands of pictures of a dog or cat or a camel-back sofa. Amazon came with a crowdsourcing platform called MTurk—Mechanical Turk—to facilitate the posting of jobs by requestors which anyone with a minimal background in technology can respond to, complete and get paid for. Now, there are a variety of such platforms which attempt to harness this on-demand workforce. The task may be to tag a picture, translate sentences, clean up reviews posted on various websites, decide if a posting in Facebook is acceptable, moderate content or code a particular algorithm… you get the picture. Anything anyone wants done quickly for which there is not a full-time employee. Uber can also be viewed as harnessing the power of on-demand workers! The qualifications for these jobs are all over the place.

“If trends continue at current rate, economists estimate that by the early 2030s, tech innovation could dismantle and semi-automate roughly 38 percent of jobs in the U.S. alone.” The book takes a look at labor practices over the past century and underscores the following: “The Paradox of Automation’s Last Mile.” Basically, current automation does not really automate fully and there is always leftover work which needs to be handled with human creativity. The same was true a century ago with the Industrial Age, as is true with all the AI solutions trotted out to solve all kinds of problems now. The authors surveyed and interviewed a range of these on-demand workers, both in the U.S. and India. Apparently, there are roughly 20 million workers in the U.S. alone (2016 estimate) who earned some income by completing work posted on on-demand platforms. These folks by and large had a good educational background and chose to work from home and eke out an existence, usually for good reasons. These workers live at the mercy of the platform and have zero rights or benefits. No one to complain to if they don’t get paid, no one to complain to if they get kicked off the platform for any reason, no one to help them understand a problem they are trying to solve or give them any feedback on what worked and what did not work. They are simply a collection of random, nameless identifiers in the overall system. In short, they form the new underclass. They are frequently victims of what the authors call “algorithmic cruelty.”

Considering 40 percent of the population may face this kind of a future, it behooves us to take a deep look at this labor practice and fix the ugliness that underlies their usage, now. The authors  articulate 10-specific fixes. The odds of the entire set being adopted are pretty close to nil, unfortunately, but we can do incremental fixes. Here are some of the fixes:

• Add collaboration to the platform;
• provide resources to execute jobs;
• have an association which will impartially rank and maintain a roster of such workers;
• institute accountability (aka the practice of fairtrade instituted for farmers—ensuring fair price for goods delivered); and
• provide a safety net that offers universal health care and a Universal Basic Income (UBI)!

 In short, treat these people as humans not as some machine. Treat them with empathy and sympathy.

Doctor burnout impact

NPR reported on an interesting fact that is close to our hearts as healthcare professionals. The cost impact of physician burnout has been estimated at $4.6 billion annually. I was curious how they went about this estimation. It turns out they focused on physicians cutting back on hours or quitting. They also factored in the cost of replacing physicians and lost income from unfilled positions. It has been reported that as much as 54 percent of physicians have shown some sign of burnout. And the primary cause for the burnout? Inefficient and ineffective EHRs. I didn’t see any trendlines discussed here—presumably, with all the speech solutions offered for EHRs, the situation is improving. One can only hope!

Pokémon Go to Sleep

Pokémon Go was a sensation a few years ago. The augmented-reality game managed to get the whole nation moving and tracking  elusive Pokémon. Its gamified exercise was considered the best health app of 2016! The makers of this game are on to the next thing: gamifying sleep! Exactly how this is supposed to work is unclear. The product is slated to be released next year. If it can get our youth (and practically everyone) to get a bit more sleep in order to get more Pokémon points…that’s definitely a good thing!

Acknowledgements

My colleague, Philippe Truche pointed me to the article on doctor burnout costs from NPR.

I am always looking for feedback and if you would like me to cover a story, please let me know. “See something, say something”!  Leave me a comment below or ask a question on my blogger profile page.

V. “Juggy” Jagannathan, PhD, is Vice President of Research for M*Modal, with four decades of experience in AI and Computer Science research.