The onslaught of AI development continues at a breathtaking pace. There’s no apparent limit in sight as to the capabilities of massive amounts of computing power are applied to an increasing range of heretofore human activities. There are significant (not to say dire) implications for education, labor markets and social structures, the power grid, financial markets, and other fields, as I have noted in previous entries in this occasional series.
In a piece I called “Grand Turing Test” (122322), I hypothesized the emerging “arms-race” between generative AI capabilities and our ability to determine whether such output was human or machine-generated. It is a particular problem for educators in the grading process and one I’m glad I’ve escaped at this point. The point of a Turing Test is to see if a human can distinguish a machine response to a particular prompt from a human response in order to determine whether the machine is “intelligent.” It was a breakthrough in conceptualizing computing power back in the now seemingly prehistoric days of 1950, and it’s still useful today.
AI capabilities seem to match and exceed human capabilities in an increasing number of applications; so the implications of the Turing Test are less about whether the machine is “intelligent” and more about what are the distinctive aspects of human intelligence. This seems, as I say, to be a shrinking field. It is shrinking for two reasons. First, the application of brute processing power, especially when that power is “intelligently” directed, outclasses what humans can do on their own (even with the augmentation of pencil-and-paper, calculator, Excel spreadsheet, Google search, etc.) in anywhere near the same time.
Twenty-five years ago, if I had wanted to come up with a good estimate of the planetary GDP in the years 1,000bc, 0, 1,000ad, 1500, 1800, 1900, and 2000 (which I did for a recent lecture series), I could have spent a day or so in the library and come up with some plausible estimates. Ten years ago, I could have done a Google search and found sources and produced something better in a couple of hours. Now, I pop a prompt into Gemini/Claude/Chat etc. and get a digested, footnoted analysis with graphs in a matter of a few seconds. The AI may not be smarter than me, but it’s a hell of a lot faster. Ditto for writing papers/memos, or generating music, videos, pictures, etc. So far, so good (leaving aside serious concerns about plagiarism, hallucinations, skewing, etc.).
The second reason, is that this processing power challenges us to reconsider what is left in our definition of human “intelligence” beyond the “normal” degree of brainpower? Is there something distinctive about creativity (whether artistic or drafting a clever merger agreement) that is irreproducible by machine or is it all just a matter of NVIDIA chips crunching away? Naturally, we humans cling to a conception of intelligence that is unique. For a long time, we’ve disparaged the capabilities of whales, chimps, dolphins, etc. as part of our effort to preserve our self-centered view of the universe. Copernicus, Galileo, Darwin and others have shown us how distorted our traditional conceptions have been. Now, Claude and friends have further narrowed the space of the uniquely human. As with these earlier examples, despite the epistemological resistance and blind demands to stop this threatening technology in its tracks, I suspect we will find that there is very little left to our once god-like self-image.
These new capabilities are starting to upset our well-established modern conception of what it is that people do and why they should be compensated for their activities (i.e. “work”). The last quarter-millennium (or semiquincentennial!) of industrialization dramatically reduced that percentage of human compensation that was earned by physical labor as compared with intellectual effort (artistic effort has always been a tiny fraction). I asked Google’s Gemini for an estimate of this ratio; it came back with physical labor constituting 85-90% in 1750, 75-80% in 1850, 45-50% in 1950, and 23-30% today (with commensurate growth on the “intellectual labor” side). This is one way to look at the “industrial revolution” and we know the tremendous societal disruption that has accompanied this shift. Our entire modern global economic system and the great strides that have been made in standards of living over this period have been built on this transition. Capitalism provided one mode of distributing these gains across the population; but even beyond its other challenges (e.g., environmental concerns, inequality) what’s coming portends a profound crisis.
Going forward, and over a likely much shorter period, we will see a drop in the intellectual labor numbers. Whether this produces an upsurge in unemployment (beyond the serious short-term transitional impacts) or a redeployment to new modes of work remains to be seen. In either case, it bodes poorly for those whose intellectual capabilities (whether innate or educated) are at or below average. Everyone—including the majority of people who think they are “above average” in intelligence [sic]—will have to freshly consider what they do and why they do it. Traditional models of job types and career paths aren’t going to work. What mix of intellectual, physical, and creative work (just to use a simple break-out) will provide the best mix of personal fulfillment and economic support? The answers of 2010 are clearly going to be inadequate for 2030 and anyone who tells you that they have much of a handle on how things will look in 2040 is just lying to you.
In a piece I called “Grand Turing Test” (122322), I hypothesized the emerging “arms-race” between generative AI capabilities and our ability to determine whether such output was human or machine-generated. It is a particular problem for educators in the grading process and one I’m glad I’ve escaped at this point. The point of a Turing Test is to see if a human can distinguish a machine response to a particular prompt from a human response in order to determine whether the machine is “intelligent.” It was a breakthrough in conceptualizing computing power back in the now seemingly prehistoric days of 1950, and it’s still useful today.
AI capabilities seem to match and exceed human capabilities in an increasing number of applications; so the implications of the Turing Test are less about whether the machine is “intelligent” and more about what are the distinctive aspects of human intelligence. This seems, as I say, to be a shrinking field. It is shrinking for two reasons. First, the application of brute processing power, especially when that power is “intelligently” directed, outclasses what humans can do on their own (even with the augmentation of pencil-and-paper, calculator, Excel spreadsheet, Google search, etc.) in anywhere near the same time.
Twenty-five years ago, if I had wanted to come up with a good estimate of the planetary GDP in the years 1,000bc, 0, 1,000ad, 1500, 1800, 1900, and 2000 (which I did for a recent lecture series), I could have spent a day or so in the library and come up with some plausible estimates. Ten years ago, I could have done a Google search and found sources and produced something better in a couple of hours. Now, I pop a prompt into Gemini/Claude/Chat etc. and get a digested, footnoted analysis with graphs in a matter of a few seconds. The AI may not be smarter than me, but it’s a hell of a lot faster. Ditto for writing papers/memos, or generating music, videos, pictures, etc. So far, so good (leaving aside serious concerns about plagiarism, hallucinations, skewing, etc.).
The second reason, is that this processing power challenges us to reconsider what is left in our definition of human “intelligence” beyond the “normal” degree of brainpower? Is there something distinctive about creativity (whether artistic or drafting a clever merger agreement) that is irreproducible by machine or is it all just a matter of NVIDIA chips crunching away? Naturally, we humans cling to a conception of intelligence that is unique. For a long time, we’ve disparaged the capabilities of whales, chimps, dolphins, etc. as part of our effort to preserve our self-centered view of the universe. Copernicus, Galileo, Darwin and others have shown us how distorted our traditional conceptions have been. Now, Claude and friends have further narrowed the space of the uniquely human. As with these earlier examples, despite the epistemological resistance and blind demands to stop this threatening technology in its tracks, I suspect we will find that there is very little left to our once god-like self-image.
These new capabilities are starting to upset our well-established modern conception of what it is that people do and why they should be compensated for their activities (i.e. “work”). The last quarter-millennium (or semiquincentennial!) of industrialization dramatically reduced that percentage of human compensation that was earned by physical labor as compared with intellectual effort (artistic effort has always been a tiny fraction). I asked Google’s Gemini for an estimate of this ratio; it came back with physical labor constituting 85-90% in 1750, 75-80% in 1850, 45-50% in 1950, and 23-30% today (with commensurate growth on the “intellectual labor” side). This is one way to look at the “industrial revolution” and we know the tremendous societal disruption that has accompanied this shift. Our entire modern global economic system and the great strides that have been made in standards of living over this period have been built on this transition. Capitalism provided one mode of distributing these gains across the population; but even beyond its other challenges (e.g., environmental concerns, inequality) what’s coming portends a profound crisis.
Going forward, and over a likely much shorter period, we will see a drop in the intellectual labor numbers. Whether this produces an upsurge in unemployment (beyond the serious short-term transitional impacts) or a redeployment to new modes of work remains to be seen. In either case, it bodes poorly for those whose intellectual capabilities (whether innate or educated) are at or below average. Everyone—including the majority of people who think they are “above average” in intelligence [sic]—will have to freshly consider what they do and why they do it. Traditional models of job types and career paths aren’t going to work. What mix of intellectual, physical, and creative work (just to use a simple break-out) will provide the best mix of personal fulfillment and economic support? The answers of 2010 are clearly going to be inadequate for 2030 and anyone who tells you that they have much of a handle on how things will look in 2040 is just lying to you.
RSS Feed