Author : Tronserve | Monday, 7 October 2019
A big pastime of economists in the 1980s and 1990s was striving to guess how much corporate and industrial productivity would benefit from the then-novel phenomena of personal computers, workgroup servers, and computer networking. Primarily it was difficult to see, but with time, economists did indeed find evidence that information technology contributed to boosting economic productivity.
It is quite early to expect to see data showing a similar boom from artificial intelligence, today’s big IT revolution. The technology is just becoming industrialized, and a lot of companies have yet to even try to use things such as machine learning in any significant way.
But it is not too soon to speculate. There’s no question companies will progressively use AI technologies of various sorts. AI is now well on its way to being part of how companies function. Every company has tons of data to analyze, and that analysis can benefit from even simple machine learning techniques. And companies have processes, from HR to accounting to sales, that can benefit from automation that AI can bring.
Will all that show up in the numbers around output per employee and such, the measures of productivity? Although it can’t be ruled out, a couple big difficulties stand in the way of AI having an effect on productivity similar to the PC era.
One hassle is that AI is dominated by the companies that are already among the most productive in the world. As MIT economist David Autor and colleagues have written, wealth is nowadays concentrated in the hands of what they term “superstar firms,” a situation of “winner take most,” where “a small number of firms gain a very large share of the market,” firms that are the “more productive” ones.
Those companies include Google and Facebook, and others that, Autor and colleagues show, are much more efficient in regards to their labor force. “Many of the canonical superstar firms such as Google and Facebook employ relatively few workers compared to their market capitalization” because “their market value is based on intellectual property and a cadre of highly-skilled workers.”
Google, Facebook, Apple, Amazon and Microsoft, the most well known tech companies in the world, the superstar firms, are precisely the ones that already dominate artificial intelligence all over the world, the companies at the forefront of deep learning and other forms of cutting-edge AI. In a sense, AI is being used to improve productivity that is already significantly above normal. At the same time, something unfortunate has befallen all the non-superstar firms in the world. Back in the 1980s and 1990s, PCs and related technology were a broad global trend benefitting any company that bought PCs, servers and networking. Productivity was theoretically available to all.
With the death of Moore’s Law, the decades-long rule of progress in the semiconductor industry, there's much less technology improvement that is widely available in a direct way to every firm. Fundamental research has contracted across the technology industry, and much of what innovation happens is progressively concentrated in the R&D labs of those same superstar firms.
With superstar firms prevailing over AI, and broad tech progress no longer evenly distributed, how will AI contribute to a boom? Probably it will happen indirectly, a process of “trickle-down” productivity, as ordinary firms adopt the AI technologies provided by Google and Microsoft and Amazon in the cloud. Whether or not productivity does not immediately improve at every firm, improvements could still materialize inside of industries, and as a national or global phenomenon.
It’s important to keep in mind that productivity can take a while to materialize. Back in 1987, Nobel Prize-winning economist Robert Solow was the first scholar to suggest the apparent absence of IT-led productivity growth. “You can see the computer age everywhere but in the productivity statistics,” he famously wrote. It took another decade or so, but gradually the numbers did show progress. An AI boom is possible; certainly, it should not be ruled out. But market concentration and a slowdown in tech innovation broadly speaking will make it more challenging to achieve than was the case for technology revolutions of the past.