You don’t have to look beyond today’s newspaper, the latest popular or business magazines or just about any news program, much less endless websites. Even a recent issue of AARP, the magazine for elderly and retired people, has scary ideas about artificial intelligence. AI may be more all-present than Trump.
Among the latest attention-getting issues and headlines are the following: how could AI take over elections—and undermine democracy; does regulating artificial intelligence save humanity or just stifle innovation; how can congress regulate AI; and not least—does regulating AI save humanity or just stifle innovation?
But the biggest fears for many surround the issue of jobs.
Two of MIT’s most knowledgeable guys on artificial intelligence, Andrew McAfee and Erik Brynjolfsson, will tell you to cool it about jobs and AI. They argue, for example, that asking how many jobs will be killed by AI is the wrong question. Sure, Frey and Osborne’s analysis from early 2015 found that nearly 50% of current jobs are susceptible to AI. But by no stretch of the imagination is that the whole picture.
Like previous predictions about gains and losses, they’ve almost always been way off. Furthermore, nearly all these predictions have been about job destruction and not job creation. Destruction always makes for better news. (Having taught persuasion theory for nearly a dozen years, I’ve gotten to the place where I reject nearly most political, economic and technological scares—except for climate.) If the issue is important to me, I look for the real experts to see what they have to say. To the degree possible, I try to identify and ignore my own bias on the subject. Here, there is a genuine reason for real experts, in spite of what the media idiocracy would have you believe.
Of course, even the work of experts is going to be challenged by AI. Just this month the Economist had an article indicating that generative AI will be changing the way lawyers work—meaning some will lose their jobs. . . if they don’t change the way they’re working.
According to a recent report from Goldman Sachs, 44% of legal tasks could be performed by AI. That’s more than in any occupation surveyed except for clerical and administrative support. Lawyers spend an awful lot of time scrutinizing tedious documents—the sort of thing that AI has demonstrated, for more than a dozen years, it can do well. Lawyers already use AI for a variety of tasks, including due diligence, research and data analytics. These applications have largely relied on “extractive” AI, which, as the name suggests, extracts information from a text, answering specific questions about its contents.
“Generative” AIs such as Chatgpt are far more powerful. Part of that power can be used to improve legal research and document review. As Pablo Arredondo, creator of a generative-ai “legal assistant” called CoCounsel, explains, using it “removes the tyranny of the keyword…It can tell that ‘We reverse Jenkins’ [a fictional legal case] and ‘We regretfully consign Jenkins to the dustbin of history’ are the same thing.” Allen & Overy, a large firm based in London, has integrated a legal AI tool called Harvey into its practice, using it for contract analysis, due diligence and litigation prep.
So MIT’s astute David Autor will tell you that you don’t want to set aside all your AI concerns. On numerous occasions, he’s pointed out two major labor force challenges caused, at least in part, by technology. The first is that the great middle-class I grew up in was built on routine work. You had to be a real dumbass in Detroit in the ‘forties and fifties’ not to have a good salary and benefits. I made enough in one year (’53-’54) of factory work to pay for nearly three years of my college expenses. During and shortly after WWII, the auto industry built a middle-class like we’ve never seen since. In contrast, two of the young maintenance guys at my apartment say they’re saving money to go to vocation school or maybe, college. On $12 or $13 an hour, that’s plainly unrealistic. You can’t begin to pay for college with one year of $17 an hour factory work today. Not even a single semester.
So, there’s plenty of routine work today, too, but you’re not going to get enough to live on from a routine work salary.
Autor and his colleagues also point out a second challenge. In spite of our very low unemployment rate, there is actually a very serious joblessness problem among some groups. That’s because people who have stopped looking for work are not included in the joblessness statistics. And a large percentage of these folk, mostly prime-aged, poorly educated men, are in this category.
The conclusion? “As automation takes over truck driving tasks and other similar jobs, this mismatch between desired and available is likely to grow, as will the joblessness and attendant problems that come with it.”
How to resolve this problem? Well, a strong set of government policies is one way.
The second way is going to work on our decaying infrastructure. That includes more than just roads. It also includes ports, bridges, airports, etc. It needs to be said that looking at things like coal mining is just looking in the rearview mirror.
A third way to deal with the negative impacts on routine work is by AI’s potential to create new opportunities for even--truck drivers. For example, truck drivers could be trained to work as technicians to maintain and repair autonomous trucks. Additionally, truck drivers could move into new roles in the trucking industry, such as data analysts or logistics managers.
None of these problems is insurmountable. But they’re going to take a national push by government and the voting public.