A few months ago, I was waiting on the subway with a friend, a professional editor, who had never used a large language model (LLM). Standing on the stage, she told me about an article she was working on. ChatGPT came out six weeks ago, I put her summary into it on my phone and showed her the result. I’ve been following OpenAI modeling based on transformers since 2019 and forgot the impact they can have on first exposure. My friend couldn’t take her eyes off the little gray box as the article was posted, line by line. It took me a minute to register the shock on her face. She half-jokingly said on the train, “I’ll be out of a job by the end of the year.”
As wave after wave of new AI capabilities emerged over the past few months, I’ve been thinking about my girlfriend and her place in the unfolding world. When GPT-4 was released in March, OpenAI’s press release included a chart of its results in several standardized tests. The touted new model scored more than 80 percent on 11 AP and SAT tests, 77 percent on “Advanced Sommelier (Theoretical Knowledge)” and — most controversially on Twitter — 90 percent on the Uniform Bar Test, the national test becoming a lawyer. OpenAI’s previous model, GPT-3.5 (which supported ChatGPT when it first appeared), had already passed the US medical licensing exam, and had a degree that, if human, would qualify it to become a doctor.
Show more
Such results seem to validate A 2019 Paper By then, a doctoral candidate at Stanford University, Michael Webb. Although entirely speculative at the time, it has upended the prevailing wisdom about who wins and who loses as a result of AI-driven automation. Prior to Webb’s report, studies had been done by Oxford And Mackenzie I expected lower-wage, lower-skilled jobs to be hit hardest, as has indeed happened throughout the entire history of automation going back to the steam-powered textile loom.
The LLM era changed all that. Now, conventional wisdom – iterate and expand on recent paper by OpenAI researchers — is that higher-paying jobs and creative jobs (including mathematicians, tax preparers, fiddlers, writers, and web designers, to name a few) are often the most exposed to automation (100 percent exposure for the listed occupations). just yet ). This has an interesting side effect, because, as Webb’s study showed, so is white-collar work in the United States It was done disproportionately By the most privileged: men, white people, Asian Americans, people in their early working years (25-54), and people who live in affluent coastal cities. It’s been easy for many in this demographic for a long time, but it looks like the AI revolution is going to be a bumpy ride for them.
I spoke with four economists for this article, and though they gave good reasons to believe that AI won’t “take all jobs” — as in previous waves of automation the economy is likely to grow — no one denied that some jobs would. lost. They didn’t know exactly how many there were, and neither did I. But what I do know is that we’ve never seen a wave of automation in which white-collar workers are uniquely vulnerable, and so we should expect this to be done differently.
The core of the difference lies in the relationship between blue-collar workers, white-collar workers, and labour. according to one studyWhite-collar employees tend to feel that they are “expressing” their “full potential” at work more than blue-collar workers do; They also experience higher levels of “inner self-development” at work. according to Another studyWhite-collar workers valued “interesting work (nature of work), achievement, recognition of work done (recognition)”—in contrast to blue-collar workers, whose motivations were “receipt of salary, working conditions, peer relationships, and job security.” (Even more so than in other groups, men derive their self-esteem from achievement and feeling useful. A dramatic illustration of this was study Of the language used by suicidal men, which showed that its consideration useful was essential to men’s well-being, and its absence was devastating. Making a robot useless would have disproportionately bad emotional effects on a man.)
“Nature of work” is one way of saying that white-collar workers care about the tasks we do. Being ‘recognised’ and ‘appreciated’ for ‘achieving’ on these tasks is important to us; It is how we “express” our “full potential”. In other words, large parts of our emotional lives and our social selves are connected to the tasks we perform at work. What happens when AI does these tasks better?
At its most extreme, white-collar work is a type of task where competence is so admired that it becomes a sport or art, and society rewards competence with status and respect, as well as monetary compensation. This is the category of logic and art games. Our shock at the new wave of AI models like ChatGPT and Midjourney comes from their mastery of more creative technical tasks like writing and illustration. But tougher logic sports like chess and Go have long since been conquered by earlier waves of AI, and so it might be helpful for left-brained people to look at how right-brained people deal with possession, both emotionally and practically.
Go is generally considered to be humanity’s most complex game. In 2016, DeepMind’s AlphaGo beat out two of its top players. Lee Sedol, a Korean genius and the second best player in the world at the time, took the hardest part. He became depressed, and two years after the game he retired from the game, citing AlphaGo. “Even if you become number one, there is an entity that cannot be defeated,” he said.
Van Hoy, European champion but below the world leaders, took it one better. He was initially shocked and groveled by his defeat, and actually tried to forget about the game completely. V said AlphaGoDeepMind’s 2017 documentary about Lee and Fan. He said the game is like looking in a mirror. “I see Go; I also see myself. For me, Go is real life.” Later, however, he joined DeepMind – the architect of his defeat – and helped improve its model’s capabilities. Basically, he couldn’t beat them, so he joined them.
These differences seem so enlightening, it’s hard not to look for lessons. I can’t help but think that Lee’s higher rating actually made him more vulnerable to an existential crisis, because he had so much more to lose. Fan was disappointed, but Lee publicly lost out to millions of Korean viewers. It’s hard to undo it, and it’s probably even harder than it was for the propeller to spin on its axis.
I met up with my editor friend again recently – three months after she had her first exposure to ChatGPT. She looked more worried than ever. “I think it’s going to be a hard fall,” she said. She felt younger and more technically adept at wearing her heels and worried that she had not been brought up to be flexible enough for this kind of challenge. I tried to give hope in the form of a story Gregory Clarke, professor emeritus at UC Davis, told me about aristocratic landowners during the Industrial Revolution. Tenant farmers left the country to pursue better pay in the factories in the city which caused the aristocrats to drop the value of farmland, causing huge losses to the aristocracy. Despite this, Clark said, the intelligent aristocrats—those who could adapt—simply followed the farmers into the cities and became urban landlords.
My friend is only partially sold. What is the equivalent now for her?
This is when I remembered a third Go champion playing AlphaGo but not being included in the documentary. This is KG. In 2017, months after Li’s match, he was 19 years old and the best player in the world, having beaten Li in three consecutive tournaments. Like Fan and Lee, Ke also lost to AlphaGo, after which AlphaGo had no human left to defeat.
But Ke’s reaction is, I think, the most interesting and also the most hopeful. Pre-AlphaGo, Ke, a teen with world-class abilities, was also a world-class brat, famous for bucking Go’s culture of modesty. When he challenged Ke Lee to a match, for example, he posted a video of himself as a boxer beating Lee, showing off, and feeding his opponents.
In the aftermath of Ke’s defeat at the hands of the artificial intelligence DeepMind, he undergoes a remarkable change. In his television appearances since then, he’s influenced an attitude of irony, fun, and humility, becoming much likable along the way. Again, in searching for lessons, I couldn’t help but notice Ke’s extreme youth—15 years younger than Lee, 16 younger than Fan—and wonder if he had invested less in a certain way of appreciating and understanding himself. Perhaps he was thus better able to change how he relates to the world on a fundamental level.
What’s important about this story also is that, unlike Fan, whose pivot as a temporary AI research advisor can be seen as a downgrade from the European Go champion, Ke’s pivot has allowed him to stay ahead of the game.
The transformation from “the best player in the world at mankind’s most logically complex game” to “comedian” is very dramatic, and I think the scale of this reversal reflects how profound the changes are happening in the pipe. And if Ke Jie had to do it, what would that mean for the rest of us? My hunch is that economic concerns will dominate in the coming years, but assuming this is resolved, where will the case reappear if GPT-7 swallows up the core competencies of Arts, Design, Science, Law, Medicine, and Engineering?
Webb himself believed that human standing would become something akin to governance, “where the point is that human beings make the decision.” For a judge, politician, or newspaper editor, for example, “We know we can get AI to do it for us—we can ask it to tell us what to do—but we’d rather have a human do it.”
Once again, the cutting edge of Go and Chess – which was solved by AI two decades ago – offers us tea leaves for the divine if we choose to read it. In these realms, Ke Jie isn’t the only high-status genius who pinned himself down like he did; Magnus Carlsen, the world’s best chess player, has become famous in recent years for his “interesting” playing style in response to artificial intelligence creating an indisputable hierarchy of opening moves. Even more heretical, players with much lower skill levels are beginning to eclipse the old masters in terms of popularity: the stylish and attractive Botez sisters are The second most flowing Chess players with ELO ratings nowhere near the best in the world. And Zhan Ying, a Chinese Go player with a much lower skill level than Ke Jie’s, recently dethroned him briefly as the most watched Go player in the world.
If this trend is any indication, we should expect to see softer skills — humor, presence, personality — become the game. In light of this, we may already be halfway there without fully realizing it: perhaps the future belongs to the influencer.