J-curve of AI and the coming productivity boom – TechTalks
This article is part of our series that explores the business of artificial intelligence
Digital technologies, and at their forefront artificial intelligence, are triggering fundamental changes in society, politics, education, economics and other fundamental aspects of life. These changes offer unprecedented opportunities for growth in different sectors of the economy. But at the same time, they involve challenges that organizations must overcome before they can fully exploit their potential.
Speaking recently at an online conference hosted by Stanford Human-Centered Artificial Intelligence (HAI), Stanford Professor Erik Brynjolfsson discussed some of these opportunities and challenges.
Brynjolfsson, who directs Stanford’s Digital Economy Lab, believes that over the next decade the use of artificial intelligence will be much more widespread than it is today. But its adoption will also face a period of lull, also known as the J-curve.
“There’s a growing gap between what technology is capable of doing and what it already does versus how we react to it,” says Brynjolfsson. “And therein lies many of our society’s greatest challenges and problems and some of our greatest opportunities.”
Machine learning and increased productivity
According to Brynjolfsson, the next decade will see significantly higher productivity thanks to a wave of powerful technologies, especially machine learning, that find their way into every computing device and application.
Advances in computer vision have been tremendous, especially in areas such as image recognition and medical imaging. Talking to smart phones, watches and speakers has become commonplace thanks to advances in natural language processing and voice recognition. Product recommendation, ad placement, insurance underwriting, loan approval, and many other applications have benefited immensely from advances in machine learning.
In many areas, machine learning is reducing costs and speeding up production. For example, applying great language patterns in programming can help software developers become much more productive and get more done in less time.
In other areas, machine learning can help create applications that didn’t exist before. For example, generative deep learning models are creating new applications for arts, music, and other creative work. In areas such as online shopping, advances in machine learning can create major shifts in business models, such as a shift from “buy then ship” to “ship then buy.” “.
The lockdowns and urgency caused by the covid-19 pandemic has accelerated the adoption of these technologies in different sectors, including remote working tools, robotic process automation, drug research and automation factories.
“The pandemic has been horrific in many ways, but another thing it has done is accelerate the digitalization of the economy, compressing in about 20 weeks what would have taken maybe 20 years of digitalization,” says Brynjolfsson. “We have all invested in technologies that allow us to adapt to a more digital world. We’re not going to stay as far apart as we are now, but we’re not going to go back either. And this increased digitization of business processes and skills is compressing the time frame we have to adopt these new ways of working and ultimately increase productivity. »
The J curve
The productivity potential of machine learning technologies comes with an important caveat.
“Historically, when these new technologies become available, they don’t immediately translate into productivity growth. Often there is a period where productivity declines, where there is a lull,” says Brynjolfsson. “And the reason there’s this lull is that you have to reinvent your organizations, you have to develop new business processes.”
Brynjolfsson calls this the “J-curve of productivity” and documented it in an article published in the American Economic Journal: Macroeconomics. Basically, the great potential caused by new general-purpose technologies like the steam engine, electricity and, more recently, machine learning requires fundamental changes in processes and workflows, the co-invention of new products and business models and investment in human capital.
These investments and changes often take several years, and during this time they do not yield tangible results. During this phase, companies create “intangible assets,” according to Brynjolfsson. For example, they could train and retrain their workforce to use these new technologies. They could redesign their factories or instrument them with new sensor technologies to take advantage of machine learning models. They may need to revamp their data infrastructure and create data lakes on which they can train and run ML models.
These efforts can cost millions of dollars (or billions in the case of large companies) and bring no change to the company’s output in the short term. At first glance, it seems that the costs are increasing without any return on investment. When these changes reach their turning point, they result in a sudden increase in productivity.
“We’re in this period right now where we’re doing a lot of this painful transition, restructuring work, and there are a lot of companies struggling with that,” says Brynjolfsson. “But we’re working on it, and these J-curves will lead to higher productivity – according to our research, we’re near the bottom and we’re going up.”
Switch to AI
Unfortunately, adapting to AI and other new digital technologies is not following a predictable path. Most companies don’t make the transition properly or lack the creativity and understanding to make the transition. Various studies show that most applied machine learning projects fail.
“Only the top 10-15% of companies make the bulk of the investment in these intangibles. The other 85-90% of companies lag behind and make almost none of this restructuring necessary,” says Brynjolfsson. “It’s not just about big tech companies. This concerns all industries, manufacturing, retail, finance, resources. In each category, we see the leading companies standing out from the rest. There is a growing performance gap.
But while the adoption of new technologies is going to be difficult, it is happening at a much faster pace compared to previous cycles of technological advancements, as we are better prepared to make the transition.
“I think what’s becoming clear is that it’s going to happen a lot faster, partly because we have a much more professional class of people who are trying to study what works and what doesn’t,” says Brynjolfsson. “Some of them are in business schools and universities. Many of them are in consulting companies. Some of them are journalists. And there are people who describe which practices work and which don’t.
Another element that can help enormously is the availability of machine learning and data science tools to process and study the huge amounts of data available on organizations, people and the economy.
For example, Brynjolfsson and his colleagues are working on a large dataset of 200 million job postings, which include the full text of the job description along with other information. Using different machine learning models and natural language processing techniques, they can transform job postings into digital vectors that can then be used for various tasks.
“We view all jobs as this mathematical space. We can understand how they can relate to each other,” says Brynjolfsson.
For example, they can make simple inferences such as the similarity or difference of two or more job postings based on their text descriptions. They can use other techniques such as clustering and graphical neural networks to draw more important conclusions such as what type of skills are most in demand, or how the characteristics of a job position would change if you changed the description to add AI skills such as Python. or TensorFlow. Companies can use these models to find flaws in their hiring strategies or to analyze the hiring decisions of their competitors and leading organizations.
“These kinds of tools just didn’t exist five years ago, and I think this is a revolution that’s just as important as the microscope or some of the other scientific revolutions,” Brynjolfsson says. “We now have some for social sciences and business to have that kind of visibility. This allows us to transition much faster than before.
However, Brynjolfsson warns that few companies use such tools. This perhaps speaks more to his earlier point that companies have yet to find the right transition strategy and are relying on old ways to restructure and adapt to the age of AI. And at the center of this strategy should be the correct use of human capital.
“You have hundreds of billions of dollars in human capital, skills coming out, and then the company tries to rehire people with the skills they need. What they don’t realize is that the workers they fired often had skills very close to those they were hiring for,” says Brynjolfsson.
With the help of machine learning, they will have better visibility and insight into their “skill adjacencies,” says Brynjolfsson. For example, a company might find that instead of laying off a bunch of people and looking to hire new talent, maybe all it needs is a bit of retraining and retraining of its workforce. ‘work.
“It’s a lot more expensive to hire someone new than to take some of these people who are already in the business and say, if we teach you Python or customer service skills or other skills, you can do this work that we’re looking to hire people for,” says Brynjolfsson. “Hopefully over the next decade, workers will be in a much better position to make full use of their abilities and skills. it will also be good for companies to understand all the assets they have, and machine learning can help a lot in understanding these relationships.