Experts Ponder the Economic Costs of AI
IDE conference speakers consider the impact on jobs as AI and machine learning move ahead
By Paula Klein
AI technology and the economy may be at a crossroad. Fueled by advances in machine learning, AI is knocking down old computational parameters and challenging humans at all kinds of tasks. At the same time, the impact on displaced and disadvantaged workers is becoming a central part of the AI/automation conversation.
Certainly, technology is on the fast track. The abilities of computers to see, speak, and interface with humans have leapfrogged in the past decade. “Robotics is becoming ubiquitous,” and deep neural networks with better algorithms, rules, and “learning” capacities are challenging humans in everything from poker to legal research, said Erik Brynjolfsson, Director of the MIT IDE, at the March 8 AI and Machine Learning Disruption Timeline Conference.
A robot is a “peripheral brain,” Gill Pratt, CEO of Toyota Research Initiative, told attendees, and data analysis, sensors, and the Internet have significantly accelerated AI developments. Besides autonomous cars, Toyota is developing “smart” home companions and assistants for the elderly.
And AI fervor is clearly evident in the $4 billion in venture capital invested in AI as of last year. These investments include autonomous vehicles, medical, educational, and back-office applications, said Albert Wenger, Partner at Union Square Ventures.
Toyota has committed $50 million over five years to support collaborative research into AI and robotics research at Stanford and MIT.
Even so, conversations about the next steps were measured. MIT Sloan economist David Autor, among others, described huge pressure on medium-skill jobs and diverging wages. His research shows highly unequal income and intergenerational wage disparities. With Americans digging in their heels at the prospect of job displacement, the political climate could be impacting the tone, as well as the substance, of AI dialogues. The economic costs of automation are top-of-mind and many experts at the conference addressed the key question of the day: Can disruptive technology charge ahead and also create good jobs?
Professor Manuela Veloso, of Carnegie Melon University, spoke of “collaborative robots, “or co-bots, that can serve as interim solutions between human-only work and totally autonomous machines. Complementary human/machine interaction is now commonplace and AI is being phased in incrementally.
At Toyota, a new test car can handle two self-driving platforms: Chauffeur and Guardian. The first is a full autonomous system while the latter is a “high-level” driver-assistance system. While Chauffeur probably won’t be in cars until 2020, Guardian could come much earlier, according to Pratt.
It’s necessary to temper over-inflated expectations, but we also need to do a better job explaining AI to businesses, and how it will be phased into the enterprise, said Accenture Managing Director, Nicola Morini Bianzino. We “took repetitive tasks away from workers, but didn’t replace them.” What’s needed is nothing short of a redefinition of work as we apply AI, as well as better education and training.
According to one survey, corporate executives are worried about AI disruption, but it’s not holding back initiatives. Fully 88.5% expect to invest in AI and machine-learning in the coming decade, and 68.9% report that investments in AI are already underway, according to The 2017 Big Data Executive Survey from NewVantage Partners.
And that will likely mean job loss. The challenge, as MIT’s Autor sees it, is how to “create good jobs along with new technology; and to prepare workers for the gig economy when social safety nets are lacking.”
VC Wenger was more direct: We “can’t re-frame everything in terms of full employment; we should be more honest about what work will look like… jobs will go away.” Going forward, we need to ask how to restructure the economy when jobs change.
And solving that dilemma may be where the most brainpower is needed.
Paula Klein is an editor and content producer for the MIT IDE
Read the full conference report by Tim Aeppel here.
Session videos now available on YouTube here.
