You know, I begin to think AI, at least in its LLM versions, is not going to take over as predicted. It's useful, for sure, and will help lawyers a lot with documents and research and briefs, and I suppose will displace a few people, but mostly it will do things that have not been being done.
Yes. I think you’re putting your finger on an important distinction that got rather lost in the AI hype: automation of existing work versus creation of new work.
LLMs are extraordinarily good at things like drafting, summarizing, searching a large body of text, translating, explaining, producing first drafts, and turning messy information into something usable. That means a lawyer who used to spend six hours assembling a brief might spend two hours doing it and then spend the other four hours on something more valuable—or simply handle more cases.
But that doesn't necessarily mean five people get fired. Quite often it means the firm can do work that previously wasn't economical to do at all. A lawyer might now investigate ten marginally worthwhile claims instead of three, give much more attention to small clients, or have an AI continuously review a mountain of documents that nobody previously had time to read.
And there is a very human economic phenomenon behind this. When the cost of producing something falls dramatically, demand for it tends to increase. Cheap photographs didn't eliminate photography; cheap long-distance communication didn't eliminate communication. Word processors didn't eliminate writers. They made vastly more writing economically feasible.
I suspect LLMs will be particularly powerful in this respect because so much white-collar work consists of manipulating language and information. The limiting factor was often not knowing how to do something but the sheer number of human hours required to do it.
So I wouldn't be surprised if, twenty years from now, people look back at today's predictions of massive unemployment with the same bemusement with which we look at some of the predictions that computers would leave everybody with nothing to do. There will certainly be displaced occupations and people who lose their particular jobs. But the bigger effect may be that the economy discovers an enormous quantity of things it was previously too expensive to have humans do.
And there is a slightly amusing paradox: the better AI becomes at this, the harder it may be to notice its economic effect. If it simply enables a lawyer, doctor, engineer, researcher, teacher, or bureaucrat to do twice as much, we don't see an AI job. We see somebody doing considerably more work.
No comments:
Post a Comment