AI Technology and Employment

Adding context to the current beliefs about the AI pros and cons which emerged in a survey.

Few topics divide opinion as significantly as artificial intelligence (AI). While organizations such as Amazon, Baidu, Facebook, Google, and many others are investing significant resources in acquiring technologies, as well as developing and enhancing AI solutions, others have a different perspective. In fact, serial entrepreneur Elon Musk described AI technology as ‘summoning the demon,’ believing that the biggest threat the world faces is creating something that will eventually rival human intelligence.

Ever since the changes introduced by the first machines, humans have worried that they will be replaced by automation and their jobs will disappear. For at least a century before John Maynard Keynes coined the term ‘technological unemployment’ in the 1930s, the issue of machines replacing human labor was debated.

Arm wanted to get a sense for prevailing consumer sentiment around the promise of artificial intelligence, and so it commissioned a global opinion survey in early 2017. When it comes to employment, there is no question that AI technology will have an impact on jobs and, while history is no guarantee of the future, until now there has been no large-scale reduction in employment due to technological advancement. However, in our survey, 30 percent of respondents cite ‘fewer or different jobs for humans’ as being the biggest drawback of a future in which AI technology impacts human life.

Figure 1: Respondents’ views on the greatest potential drawbacks of AI.

Transition or Reduction?
This may seem to be an unnecessary concern based on history, and there is some recognition that artificial intelligence may drive a shift in employment as opposed to a reduction. This reflects the experience from previous times when a technological leap forward has increased automation, leading to improved efficiency and productivity. The result of this has almost always been a significant reduction in the cost base in employment sectors, bringing lower prices in key markets. This has, in turn, led to greater disposable income within the general population and has ultimately given birth to new market sectors, thereby creating new employment opportunities.

We can find a good example of how a technological leap drove a shift in employment patterns across several sectors by looking to the agricultural revolution. In the early 1800s, agriculture was the major employment sector. However, at this time the introduction of machinery meant that some farm workers lost their jobs. Some of these workers migrated to the newly created agricultural machine industry and began building the very machines that had cost them their jobs.

A secondary effect was greater efficiency, which meant that food became cheaper to produce, therefore prices began to fall significantly. People were spending significantly less of their disposable income on food, freeing up money to spend on other goods and services. The increased demand for these goods and services boosted employment in these sectors as well.

Until now it has been true that many tasks that are difficult for humans are easy for computers, and vice-versa. In general, computers were seen to be good with routine tasks, such as complex calculations, whereas humans retained an edge in recognizing faces or objects. AI technology will inevitably change the balance here and, due to machine learning techniques, is already making significant headway into facial and pattern recognition.

We were interested to learn our respondents’ opinions in this area. In our poll we asked them whether humans or AI machines would be better in certain industries, particularly with regard to safety and efficiency.

Figure 2: The survey indicated that in all cases, respondents believe that humans will do the job better.

The results showed that respondents tend to believe that humans will be better placed to do all the tasks we asked about, safely and efficiently. In general, the more repetitive the work, such as heavy construction or package delivery, the higher the AI machine scored. Obviously, the responses were based on the respondents’ knowledge of AI technology today, so answers may well change in the near-to-mid future.

Clearly, the rise of artificial intelligence will impact some sectors more than others. In time, some jobs may all but disappear, while others will see almost no impact and, as we have already discussed, new employment types will almost certainly be created. In our poll, we asked respondents to indicate the three types of jobs that were, in their opinion, the most likely to be affected.

Most Vulnerable
While there was a broad spread of opinion, the manufacturing and banking sectors were both high in the minds of our respondents as being the most under threat from AI technology. Even with these results, the responses are very much set in the ‘here and now.’ As an example, as vehicles continue their AI-driven march towards full autonomy, we would expect the number of respondents foreseeing the demise of taxi drivers to rise significantly.

However, there are already some positive applications for AI, especially in the medical field. Google’s own AI company, Deepmind, is working with the UK’s National Health Service (NHS) to use machine learning to combat blindness. By training a deep-learning algorithm with a million eye scans, experts predict this initiative could prevent up to 98 percent of the most severe causes of blindness.

IBM is now in the business of supporting the fight against cancer. Its Watson for Oncology platform can scan clinical trial data, medical journals, textbooks, and other sources and present oncology professionals with reports that suggest the most effective treatment options. In early testing, Watson gave the same recommendation as professional oncologists 99 percent of the time.

For individuals, Your.MD offers basic healthcare via a mobile app. Based on a chatbot, users speak about their symptoms, and Your.MD suggests conditions based on the user’s personal profile and the current symptoms. Processing natural language ensures a seamless user experience while a sophisticated map of the user’s condition is generated through various machine-learning algorithms.

Figure 3: Currently, manufacturing and banking are the sectors believed to be most under threat from AI.

New Roles
In addition to transforming some areas of employment, AI itself will become a whole new industry, spawning new job roles that do not exist today. A recent global study of more than 1000 large companies that are already using or testing AI by Accenture PLC identified some new categories of jobs that can only be performed by humans. The first is ‘trainers,’ people that assist translation algorithms and natural language processors to reduce errors and understand the subtleties of human communication—for example detecting sarcasm.

Another new job category, according to the report, will be ‘explainers.’ These people will bridge the understanding gap between ‘black box’ AI systems and people. This becomes all the more necessary due to the EU’s General Data Protection Regulation, planned for next year; this creates a ‘right to explanation’ allowing consumers to question any algorithmic decision that impacts them.

The final category identified in the report is ‘sustainers.’ These humans will ensure that AI systems operate as expected and intended. Many companies have low levels of confidence in their AI systems at present and these ‘human sustainers’ will monitor and address any unplanned consequences—effectively providing a quality assurance function for AI.

Clearly, change is inevitable, and as people start to come to terms with the implications of AI on society in general and employment in particular, it is only natural that there will be some concerns raised. Overall, we see that the more people know about AI, the more positive their view on an AI-centric future.

Alongside the concerns about employment, our research also identified concerns about privacy and security. At Arm security is always a key consideration as we design our AI-enabling technologies.

Despite some regional differences, the report indicated that across the globe people are generally in agreement that artificial intelligence will have a positive impact on healthcare, transport, and the workplace. All of which will lead to a better quality of life for everyone. In fact, 36 percent of respondents believe that AI has already impacted their lives and 61 percent think that society will be better as a result of increased automation and artificial intelligence.

As I mentioned, the global AI survey was conducted in conjunction with Northstar Research Partners. A free copy is available here.

Jem Davies is a vice president, fellow and general manager of Arm’s machine learning group. Previously, he set the technology roadmaps for graphics, video, display and imaging and was responsible for technological investigations of a number of acquisitions, including most recently that of Apical, leading to the formation of Arm’s Imaging and Vision group.

Davies has previously been a member of Arm’s CPU Architecture Review Board and he holds four patents in the fields of CPU and GPU design. He has worked in Cambridge since he graduated from the University of Cambridge and ran his own business there for eighteen years. Most recently has been at Arm for 13 years.