MENU

Reason.com

Free Minds & Free Markets

Artificial Intelligence Will Be More Economically Consequential than Steam Power Was

AI could boost economic growth by 1.2 percent annually between now and 2030.

AIAlexandersikovDreamstimeAlexandersikov/DreamstimeThe introduction of steam engines boosted productivity growth by 0.3 percent a year from 1850 to 1910, according to a new report on the likely economic impacts of artificial intelligence. The study, produced by the McKinsey Global Institute (MGI), also suggests that introducing robots to manufacturing led to annual productivity increases of 0.4 percent between 1993 and 2007, and that the introduction of information technologies led to annual productivity increases of 0.6 percent during the 2000s.

The researchers acknowledge that "predicting the economic impact of AI or any disruptive technology is a highly speculative exercise." But they give it their best shot, and their estimates suggest that from now to 2030 artificial intelligence (AI) will have even more of an impact than steam did in the 19th century.

Their analysis encompasses five broad categories of artificial intelligence: computer vision, natural language, virtual assistants, robotic process automation, and advanced machine learning. These technologies' effects on employment, consumption, and production will, they argue, spark about 1.2 percent of activity growth between now and 2030. If you assume the global economy, which now stands at about $87 trillion gross world product, continues to grow at the 2017 rate of 3.1 percent annually for the next 12 years, world GDP would rise to about $125 trillion by 2030. But if the MGI researchers are right, AI will boost growth to an annual rate of 4.3 percent, resulting in a global GDP of $144 trillion by 2030.

To get an idea of how powerfully AI could affect economic growth, let's assume that the last quarter's 4.2 percent GDP growth rate in the U.S. was somehow sustained through 2030. Today's $20.5 trillion economy, growing for 12 years at 5.4 percent, would nearly double to $38.5 trillion by 2030. Taking projected population growth into account, U.S. per capita GDP would rise from around $61,000 now to more than $107,000 in 2030.

Some AI critics think these technologies are so disruptive that development and deployment should be regulated on the basis of the precautionary principle. As tech scholar Adam Thierer of the Mercatus Institute explains it, this is "the belief that new innovations should be curtailed or disallowed until their developers can prove that they will not cause any harms to individuals, groups, specific entities, cultural norms, or various existing laws, norms, or traditions." Since all technologies have the potential for some kind of "harm," the principle is regulatory recipe for capricioulsy shutting down any innovation that attracts the attention of interest groups who believe that they will be adversely affected.

The Information Technology and Innovation Foundation offers a better way to proceed. When it comes to AI, the foundation suggests, "governments should follow the 'innovation principle' rather than the 'precautionary principle' and address risks as they arise, or allow market forces to address them, and not hold back progress with restrictive tax and regulatory policies because of speculative fears."

Every new technology comes with dangers and downsides, but humanity has reaped far more benefits from general purpose technologies like steam, electricity, and infotech than we have suffered harms. Given the significant upsides of deploying A, it is vital that would-be regulators adopt the innovation principle and allow inventors and the private sector to pursue the gains that these technologies will afford humanity.

Editor's Note: We invite comments and request that they be civil and on-topic. We do not moderate or assume any responsibility for comments, which are owned by the readers who post them. Comments do not represent the views of Reason.com or Reason Foundation. We reserve the right to delete any comment for any reason at any time. Report abuses.

  • Fist of Etiquette||

    Their analysis encompasses five broad categories of artificial intelligence: computer vision, natural language, virtual assistants, robotic process automation, and advanced machine learning.

    Full disclosure: Bailey is blog posting AI.

  • Bearded Spock||

    He is unusually restrained and reasonable for a Reason writer.

  • lap83||

    why can't I have a steam powered sex robot?

  • Bearded Spock||

    Maybe not steam powered, but I'm sure someday there will be a Steam Punk model.

    Socially awkward nerds need love, too.

  • Mickey Rat||

    You don't want a high pressure steam leak around your area.

  • Fancylad||

    That's my fetish

  • lafe.long||

    Given the significant upsides of deploying A, it is vital that would-be regulators adopt the innovation principle and allow inventors and the private sector to pursue the gains that these technologies will afford humanity.

    lol.

    Never. Going to. Happen.

  • μ Aggressor||

    Eh, tech moves faster then regulators; their incompetence and the lumbering bureaucracy could be thought of as the innovation principle in practice

  • Mickey Rat||

    When the AIs decide to kill all humans, that will have a great economic impact.

  • AlmightyJB||

    If they didn't try to kill us could we really consider them to be intelligent?

  • sharmota4zeb||

    At this point they are only going after sea creatures off the shore of Australian who don't belong there according to some expert environmentalist. Remind me again in which lands the humans are invasive species?

  • Tom Bombadil||

    This article boosted my Not Caring by 0.8% hourly.

  • MOFO.||

    Dont Dont Dont believe the hype.

  • Stilgar||

    Please stop calling it "AI" as what they are talking about is ML (machine learning) and data mining.

  • Shirley Knott||

    Hey, AI is a blank check the computer community has been kiting since the 50s.
    It's been 5 to 20 years out for that entire span of time, and still is.
    If we don't redefine it, we'll never get there.

    Maybe they should have settled on just what counted as intelligence before announcing it wasn't that hard a problem and they'd have it deployed 'real soon now'.

  • Mickey Rat||

    I liked how Michael Flynn called it "artificial stupid" in his Firestar Saga, because true AI is something of a pipe dream

  • I can't even||

    I'm highly doubtful that the technology can overcome that level of complexity without collapsing any time soon.

  • M.L.||

    True AI is a sci-fi fantasy that will never be realized. But mankind may well destroy itself in the attempt.

  • Scarecrow Repair & Chippering||

    Quick! Strangle it with red tape before it knocks out another excuse for socialism!

  • Azathoth!!||

    I always laugh at predictions about 'A.I'.

    You don't even know what 'I' is yet, you have no clue what consciousness or sentience is.

    And yet, you think you can built it from learning programs

    How precious!

    But here's a clue--while being ABLE to learn is a needed component it is one that cannot be programmed in. The limitations inherent in that programming will always keep the learner below the programmer. And thus, non-conscious.

  • sharmota4zeb||

    Let the free market decide. If you go to an AI cardiologist programmed by the guy who made the phone menu for your electric company's customer service hotline, you deserve the outcome of that treatment.

  • NotAnotherSkippy||

    Oh, a Mckinsey report. So 2-4 recently minted millenial MBA's got out their crayons and sketched something really really goodly.

  • macsnafu||

    .03 and .06 seem...rather small improvements in productivity. Sure, it's cumulative, and sure we know how significant the improvements actually were, but those numbers don't really seem to indicate that.
    And the whole purpose of government seems to be controlling other people. You don't really think they'll keep their hands off new technologies, do you?

  • macsnafu||

    Arg! I mean 0.3 and 0.6. A little more significant, but still not that much.

GET REASON MAGAZINE

Get Reason's print or digital edition before it’s posted online