How to prepare for the AI productivity boom – MIT Sloan News

open share links close share links

The last 15 years have brought what Stanford University professor Erik Brynjolfsson calls the productivity paradox. While theres been continuing advances in technology, such as artificial intelligence, automation, and teleconferencing tools, the U.S. and other countries have seen flagging productivity.

But a productivity boom is coming soon, Brynjolfsson said at the recent EmTech Next conference hosted by MIT Technology Review. He pointed to advances in technology, particularly artificial intelligence programs that are as good as or better than humans at some things. Businesses should now focus on incorporating the technology into work processes and preparing employees, he said, and policymakers should make sure its adoption doesnt contribute to inequality.

Brynjolfsson has been tracking the lag between introduction of artificial intelligence and corresponding productivity gains. United States productivity grew by about 1.3% in the past decade, he said, compared to more than 2.8% in the late 1990s and early 2000s. This productivity slowdown extends to other countries as well, according to research from the Organization for Economic Cooperation and Development. Brynjolfsson predicted a productivity J-curve, in which productivity declines after a technology is introduced and then rises when businesses have been able to integrate technologies into their workflow, a trajectory over time that has a J-shape.

I think were near the bottom of that J-curve right now and were about to see the takeoff, Brynjolfsson said.

Lagging productivity can be explained two main ways, Brynjolfsson said.

Mismeasurement. Productivity is traditionally measured using a countrys gross domestic product, which is based on things that are bought and sold. But many digital goods teleconferencing, smartphone apps, Wikipedia are available for free. Even though people get some benefit from these goods, they dont show up in productivity statistics. The information sectors share of the economy has barely budged since the 1980s, Brynjolfsson noted. I think most of us realize thats just not a real representation of whats going on, he said.

Happiness surveys also fail to capture the complete picture. Brynjolfsson suggested a new metric called GDP-B that would measure the benefit people gain from items. I think its far from perfect, but its a lot more precise than happiness, and I think its a lot more meaningful than GDP, he said.

Implementation and restructuring in businesses. It isnt enough to just add new technology to an organization. Companies need a complete paradigm shift. To get the full benefit, leaders need to rethink business processes, management practices, and employee skills, Brynjolfsson said.

This intangible organizational capital is essential for companies to see benefit from technological advances, but many companies put misplaced focus on technology itself.

The complete reconceptualization of a business process takes a lot. More creativity, effort, and frankly, time, than simply plugging in new technologies into old business processes, he said. We just havent been doing that in most industries.

About a decade ago, machine learning programs had about 70% accuracy, Brynjolfsson said. They have improved rapidly, to the point that they are now better than humans at identifying some things. This makes it more likely that organizations will move to integrate this technology into their business practices as entrepreneurs and managers gravitate toward these often cheaper and more efficient approaches.

We dont need any additional advances in technology to be able to have enormous effects on productivity and wages, he said.What we do need is some significant changes in business processes. We need to rethink the way work gets done.

There are signs more businesses are taking advantage of artificial intelligence programs. The 2021 AI Index report, which Brynjolfsson co-authored, found increases in not just the quality of artificial intelligence, but also business investment in the technology. The biggest increase was in the field of drug discovery and other biological uses of AI, with a 4.5% increase in investment in drug discovery in the last year.

Powerful technology is available, and every organization has an opportunity to benefit from it, he said. Successful firms will be prepared with the skills needed in the future, and leaders should focus on reskilling their workforce.

Replacing labor with capital and human work with technology brings concerns about decreased wages and increased inequality. Brynjolfssons research has documented how machine learning affects different skills and occupations, and found that there isnt one occupation where machine learning could do all the different tasks. While machine learning will likely reorganize work, it wont mean the end of work or entire occupations, he said.

But the effects will likely be uneven. The economic pie could get bigger, but that doesnt mean everyones going to benefit, Brynjolfsson said. Theres been some evidence of this happening, he said, with his research also indicating machine learning is more likely to affect low-wage occupations.

Inequality isnt inevitable, though. Brynjolfsson argued that to a large extent, it is the result of tax and education policies. He suggested three measures that companies, institutions, and policymakers can take to make sure all workers benefit from the productivity boom:

Reskilling the workforce. Taking advantage of AI and other technologies require different sets of skills. Im not just talking about more machine learning experts. Im talking about people who do more creative work, Brynjolfsson said. And while machines are able to do rote, repetitive work, companies will need people who are skilled at interpersonal, emotional connections.

Adjusting tax policy. Capital is taxed at a lower rate than labor, which might push companies to favor technology over workers. Brynjolfsson suggested leveling the playing field, or introducing measures such as earned income tax credits that help subsidize work.

Focusing on technologies that augment workers instead of replace them. Brynjolfsson said he is working on research that shows how technologists are focused on creating programs that replicate human skills. While that may be a fun goal, it actually isnt a particularly good one in terms of helping reduce inequality. It tends to drive down wages, he said. Id rather have them focused on augmenting human labor.

Read the original post:

How to prepare for the AI productivity boom - MIT Sloan News