William Halal, Jonathan Kolber, and Owen Davies
After decades of failed promises, artificial intelligence (AI) is now taking off. Yesterday’s doubters have been silenced, and the only current debate is about how deep and how fast intelligent machines will automate jobs, and whether the same technological forces will generate enough new work.
Several forecasts suggest AI is likely to eliminate almost half of present jobs by 2025, resulting in massive unemployment. (Rutkin, 2013) Ray Kurzweil, now at Google, extrapolates the growth of computer power to estimate that a US$1000 PC will match the human brain about 2020, and powerful AI systems will soon follow. (Frey, 2016) Ben Goertzel, leader of the OpenCog project said “I am confident that we will have human-level AI by 2025. Maybe sooner.” (Olson, 2013)
Advanced AI systems are being introduced even now. IBM’s Watson Division is partnering with hundreds of companies to automate entire fields of work. The partnership with Cleveland Clinic is developing a medical diagnosis system that promises accuracy surpassing all but the best doctors. Another partnership promises to automate legal work. Google’s Deep Mind is a deep learning system that needs no training as it learns by itself. It recently beat a human master at Go much faster than anticipated by AI researchers, and it taught itself to recognize speech.
This study addresses the looming issue of unemployment by forecasting the future distribution of jobs in categories across the occupational spectrum. We first summarize background data from the literature and present two alternative perspectives for consideration. Then results of a TechCast survey of experts concludes that a Muddling Through period of turmoil but relatively few net job losses is most likely. We also present two alternative scenarios.
Read the full article on AI and Future Jobs …