Sunday, February 26, 2017

Robots, work and the U.S. Presidential Election - Part 1

"Will robots take our jobs?" is one of the first questions I'm asked when I talk about my Centre's work in robotic vision (the application of computer vision to robotics). More recently the question has slightly changed to, "When will robots take our jobs?"

As someone who works in a lab full of robots, and sees the many limitations of robots every day, for example, our drinks-carrying robot - trained to avoid obstacles (including people who want a drink), it is easy to dismiss these concerns. I used to joke that if the job involved opening doors or going up or down stairs, humans were in no danger any time soon. One of my favourite videos is a compilation of robots failing to complete simple tasks while competing in the 2015 DARPA Robotics Challenge.


But while robots are a long way from having the balance and dexterity of humans, people do have cause for concern. The pace of technological change is overwhelming. Only 10 years ago the iPhone did not exist and the first autonomous vehicles bristled with so many sensors, and so much onboard computing hardware, they would have struggled to carry a passenger.

CMU's Tartan Racing Team won the DARPA Urban Challenge in 2007 with an extensively modified Chevrolet Tahoe. The first autonomous vehicle to navigate a 96 km course in less than 6 hours
Today your smart phone is orders of magnitude more powerful than the mainframe computers that put man on the moon, and autonomous vehicles are sleek and ready for passengers.

Google's Waymo in 2016 - no longer bristling with sensors and with more than 3 million km self-driven

While it takes human's 3-4 years to gain enough mastery of a subject to earn a university degree, IBM Watson can process 500Gb of data, the equivalent of reading a million books, per second. And while we humans can gain competence through our years of work experience, deep learning enables intelligent machines to also learn from their experiences, or indeed from ours (see robots learning to cook by watching YouTube videos). The only difference is that once one machine learns, that knowledge can be transferred to all networked machines, in much the same way that the IoT allows electric cars like Tesla to receive updates over the internet.

Imagine being able to share all the information you have gained from your life experiences with every other human on the planet. We can hope we are nearing Ray Kurzweil's Singularity, where we humans will transcend our biological limitations. In the meantime the transition to a time when there is no clear distinction between human and machine is likely to be tumultuous as robots do take some jobs and people struggle to redefine work and their place in the world.

For more on robots, work and what this has to do with the US presidential election see Part 2, coming soon.

3 comments:

  1. Hi Sue,
    John G here
    A point often missed in the discussion about robots and jobs is that robots are just one of the means of improving productivity that can result in job loss or redeployment. Productivity gains are also made by better organisation and analysing a task. When I say "productivity" I mean less materials, less waste, less energy and less human labour/effort to achieve a given outcome. We humans have been making productivity gains for 10,000 years and developing new technology of many sorts to do this. Are PLCs robots? Are control charts and statistical process control? These “soft” systems don’t get noticed or blamed for job loss but they are just as much ‘to blame” as robots whose physical presence makes them a more visible target.
    Part of my job is analysing how companies can improve their casting machine productivity. Automation including use of robots is part of the strategy to do this but not the only element. Much of it is better data collection and analysis to understand the causes of process excursions. Better manning productivity is one of the results i.e. less people needed to get the same job done. Improved safety and working environment are also benefits of automation.
    I love robot limitation stories. My real world example is ingot stacking: a horrible soul destroying, back breaking job that was manually done throughout the smelting industry for many years even into the 1980s (ingot skimming falls into the same category and unfortunately there are some smelters still using manual skimming). The first robot stacker installed at the Bell Bay smelter in Tasmania commenced its career by throwing ingots at the CEO who came to see it in action. Now robot stackers are standard kit.
    There is also a problem with how the economic gains coming from productivity improvements achieved by using robots is distributed throughout society. At the moment, it is a small proportion who receive the economic gain. Bill Gates is on to something with his tax the robots proposal. The principle of a levy by the state to help manage the discontinuity caused by job obsolescence sounds good. However, this should not be a tax on productivity gains.
    My grandfather had a farm of his own growing vegetables but he also had a good business in ploughing other people’s fields with his team of horses. Unfortunately, he did not make the transition to tractors; probably lacked the capital to buy them or just didn’t see them coming. Others with tractors took up that business. Eventually, the farm was sold. After an unsettled period he final retrained as a post master.
    The problem is not that all this technology is bad – it makes are lives easier and better. However, the pace of development goes much faster than our ability to retrain. What we need is new training technology to speed up re-training.

    ReplyDelete
    Replies
    1. Thanks John. I agree with your perspective and cover off on a lot of your points in Part 2! Love the ingot stacking story, although am hopeful no one was hurt in the incident!

      Delete
  2. No one was hurt. The safety guarding did its job

    ReplyDelete