Erik Brynjolfsson, director of the MIT Initiative on the Digital Economy said that an important threshold has been crossed by AI technologies. Indeed during his talk at this year’s MIT Sloan CIO Symposium, he addressed the fact that AI can now match or even surpass humans in certain skills (for example: image recognition). However, according to Brynjolfsson, AGI (or Artificial General Intelligence) —the point when machines will match human’s intellectual abilities—is still far from being reached. They surpass though, human capabilities in certain areas, offering numerous benefits and opportunities for businesses.
The workforce of the future will include humans
In two recent papers published in the American Economics Association and Science, Brynjolfsson and his colleagues developed a set of questions to identify what tasks AI is now fit at, then they applied this set to the O*Net Database (a database that has detailed descriptions of the world of work) of 964 occupations in the United States.
What the research found is, for each job there is 20 to 30 distinct tasks involved. In most cases, machine learning is able to perform some of the tasks better than human in a specific occupation. But, it is not able to perform all the tasks required for the job better than a human. For Brynjolfsson, most jobs will be partly handled by machines in the future and partly by humans. Thus, leading to a “partnership” between humans and machines (also known as co-bot or collaborative robots) to get work done in a more efficient manner.
Elisabeth Reynolds, executive director of MIT’s Work of the Future task force and panel participant said that only 5% of the workforce will be displaced by AI, citing a research done by McKinsey. She said that the introduction of the co-bot will allow routine worker to do something else, this echoes a Gartner research, in which they established that AI will eliminate 1.8 million jobs by 2020 but will create about 2.3 million jobs in counterpart in that same timeframe, Reynolds suggested: “As a CIO, you need to look at the skills sets that allow for flexibility”.
During her intervention Reynolds cited the example of FedEX: the company decided to introduce robots to move freight in their North Carolina warehouse. They predicted that 25 jobs would be replaced out of 1,300 people that worked there, and ended creating 100 jobs every year. However, you also have examples such as Amazon Fulfillment Centers, in which adding robots to the equation made human tasks less diverse and mobile. Reynolds said “we need to think about how humans are advantaged and what are the skills they bring to a job when designing technology”.
AI Workforce Challenges
The US workforce holds 6 million unemployed people for about 6 millions jobs vacancies. According to Iyad Rahwan the AT&T career development professor, this might be explained by a skills gap. But according to Rahwan’s research, in order to get a high-pay job, you need more education and analytical skills, which may be harder to achieve. Discussions lead to observe that there is a skills mismatch in the US, there is a high growth in highly skilled jobs but a lack of people in regional markets filling them. This can be partly explained by border restrictions, as less than 2% of americans moves across states borders each year. The future ways AI and machine learning are built into work is controlled by humans, so it is the panel opinion that we can think about how machine learning can complement and make more efficient and effective existing work.
Healthcare today, is a sector that holds a strong number of applications for robotic and AI. For instance, physical assistance robots can help humans in certain activities such as lifting up patients in which they may struggle, but also the ability of compiling data can help doctors make better diagnostics for patients.
Reynolds says that in preparing for the future of work, CIOs should look to hire for adaptability. The panel also noted that since jobs will change, enterprises should provide their workforce with lifelong learning opportunities, as automation may change the nature of their job.
One of the greatest challenge when implementing an AI, is to ensure that your data is up-to-date and to make sure that it reflects some underlying process.
For example, if you look to optimize something related with logistics or transportation, and there is a new regulation, this might indirectly greatly impact your business. So in order to have efficient predictive models, you shouldn’t only train your machine learning models on historical data and let it go wild and deploy it, because if you don’t take into account changes that happened in the real world, you might miss further opportunities to optimize your business. “Algorithms need to continuously learn” said Rahwan.
One thing is clear: digital technology will continue to move forward at a high pace, however current skills, organizations, and institutions are still lagging according to Brynjolfsson. He added “business as usual won’t solve the problem, there is an urgent need to reinvent the skills, organizations and institutions in order to keep up”.