American workers whose careers have been transformed by automation in recent decades are largely the least educated, especially men who work in manufacturing.
But a new kind of automation — artificial intelligence systems called big language models, like ChatGPT and Google’s Bard — is changing that. These tools can quickly process and synthesize information and generate new content. The jobs most exposed to automation today are office jobs, those that require more cognitive skills, creativity and a high level of education. Workers are more likely to be affected high feesand slightly more likely to be female, a miscellaneous research was found.
“Most people were surprised, including me,” said Erik Brynjolfsson, a professor at the Stanford Institute for Human-Centered AI, who prophesied that creativity and tech skills will put people off from the effects of automation. “To be brutally honest, we have a hierarchy of things that technology can do, and we’re comfortable saying things like creative work, professional work, emotional intelligence that will be difficult for machines to do. Now upend that’s all.”
A new set of studies examines the affairs of American workers, using the Department of Labor O*Net database, and hypothesize which of these large language models can do. It has been found that these models can significantly assist tasks in one-fifth to one-quarter of jobs. In most jobs, models can do some of the work, reviews found, including from Pew Research Center and Goldman Sachs.
Today, the models still sometimes produce incorrect information, and are more likely to help workers than replace them, said Pamela Mishkin and Tyna Eloundou, researchers at OpenAI, the company and research laboratory behind ChatGPT. They did a similar studyanalyzed 19,265 tasks performed in 923 jobs, and found that large language models could perform some of the tasks performed by 80 percent of American workers.
But they also found reason for some workers to fear that they could be replaced by large language models, according to Sam Altman, chief executive of OpenAI, told The Atlantic last month: “Definitely going to lose jobs, full stop.”
The researchers asked an advanced ChatGPT model to analyze the O*Net data and determine which tasks large language models could perform. It was found that 86 jobs were fully exposed (meaning every job could be assisted by the tool). Human researchers say 15 jobs. The job that both humans and AI agree is most exposed is the mathematician.
Only 4 percent of jobs have no tasks that could be helped by technology, the analysis found. They include athletes, dishwashers and those who help carpenters, roofers or painters. But even tradespeople can use AI for parts of their jobs like scheduling, customer service and route optimization, said Mike Bidwell, chief executive of Neighborly, a services company in house
While OpenAI has a business interest in promoting its technology as a boon to workers, other researchers say there are still unique human abilities that haven’t (yet) been automated — such as social skills, teamwork, care work and tradespeople skills. “We’re not going to run out of things for people to do anytime soon,” Mr. Brynjolfsson said. “But things are different: learning how to ask the right questions, really interacting with people, physical work that requires dexterity.”
Today, large-scale language models are likely to help many workers be more productive in their current jobs, researchers say, much like giving office workers, even entry-level ones, a chief of staff or a research assistant (although this may indicate a problem for the person. assistants).
Get the writing code: A study of Github’s Copilotan AI program that helps programmers by suggesting code and functions, whose users are 56 percent faster than those who do the same task without it.
“There’s a misconception that exposure is necessarily a bad thing,” Ms. Mishkin. After reading each job description for the study, he and his colleagues learned “an important lesson,” he said: “There’s no way one model is going to do it all.”
Large language models can help write legislation, for example, but cannot pass laws. They can act as therapists – people can share their thoughts, and models can respond to ideas based on proven regimens – but they don’t have human empathy or the ability to read nuanced situations.
The version of ChatGPT that’s open to the public has risks for workers — it’s prone to errors, can exhibit human biases, and isn’t secure enough for businesses to trust with confidential information. Companies that use it overcome these obstacles with tools that tap into its technology in a so-called closed domain — meaning they only train the model on certain content and keep any input private.
Morgan Stanley uses a version of the OpenAI model developed for its business that is fed by about 100,000 internal documents, more than a million pages. Financial advisors use it to help them find information to answer client questions quickly, such as whether to invest in a particular company. (Previously, this required searching and reading multiple reports.)
That leaves advisers more time to talk to clients, said Jeff McMillan, who heads data analytics and wealth management at the firm. The tool does not know about individual clients and any human relationships that may be needed, such as whether they are going through a divorce or illness.
Aquent Talent, a staffing firm, uses the business version of Bard. Typically, people read workers’ résumés and portfolios to find a match for a job opening; tool can do it better. Its work still requires a human audit, though, especially in hiring, because human biases are built in, says Rohshann Pilla, president of Aquent Talent.
Harvey, funded by OpenAI, is a start-up that sells a tool like this to law firms. Senior partners use it for strategy, such as developing 10 questions to ask in a deposition or summarizing how the firm has negotiated similar agreements.
“It’s not, ‘Here’s the advice I would give a client,'” says Winston Weinberg, a Harvey co-founder. “It’s, ‘How can I filter this information quickly so I can get to the advisory level?’ You still need the decision maker.”
He says this is particularly helpful for paralegals or associates. They use it to learn — asking questions like: What is this type of contract for, and why is it written this way? — or write first drafts, such as summarizing a financial statement.
“Now all of a sudden they have an assistant,” he said. “People can do higher-level work faster in their career.”
Other people studying how workplaces use large language models have discovered a similar pattern: They help junior employees the most. The studying of customer support agents by Professor Brynjolfsson and colleagues found that using AI increased productivity by 14 percent overall, and 35 percent for the lowest-skilled workers, who climbed the learning curve more quickly with its help.
“It closes the gaps between entry-level workers and superstars,” said Robert Seamans of NYU’s Stern School of Business, who co-write a paper found that the occupations most exposed to large language models were telemarketers and some teachers.
The last round of automation, which affects manufacturing jobs, increased income inequality by displacing workers without a college education from high-paying jobs, research shows.
AI can do this again — for example, if senior managers call on large language models to do the work of junior staff, potentially increasing executives’ earnings while diverting jobs to those with less experience. But some scholars say that big language models can do the opposite – reduce inequality between the highest-paid workers and the rest.
“My hope is to actually allow people with less formal education to do more things,” says David Autor, a labor economist at MIT, “by lowering the barriers to entry for more elite jobs that pay well.”