First of its kind study finds ChatGPT makes us 40% MORE productive and boosts output quality by 18%

The machines may one day rule over humanity, but in the short term, AI-based chatbots like ChatGPT can help the struggling workers hold their own in writing tasks.

At least that’s the finding of a new study that tracked the writing performance of 453 “skilled, college-educated professionals,” divided evenly into two groups: One learned to use ChatGPT and the other was forced to complete their writing assignments alone.

The researchers found that the chatbot made their subjects 40 percent more productive and improved the quality of their work by 18 percent.

A significant number of the study participants, who were encouraged to use ChatGPT, even stuck with the software for an extra boost in the real world.

After two weeks, 34 percent reported using ChatGPT in a professional setting. And after two months, that figure rose, with 42 percent of participants reporting signing up to ChatGPT for some further help at work.

But in what may be the study’s most admirable finding, as published Thursday in the journal Sciencethe researchers found that the participants with weaker writing skills benefited most from consulting ChatGPT.

The authors of the study, economists at the Massachusetts Institute of Technology (MIT), suggest their work illustrates a path to reducing inequality among workers.

Overall, the advent of ChatGPT ushers in an era of tremendous uncertainty about the economic and labor market effects of AI technologies, the authors wrote.

“These results are consistent with other studies showing productivity-boosting and equalizing effects of recent AI technologies,” they noted, citing two studies from the private nonprofit National Bureau of Economic Research.

The survey participants came from the fields of marketing, grant writing, consulting, data analytics, and human resources, among other professions often asked to perform “mid-level professional writing jobs.”

Half of the employees in the test group worked with the ChatGPT version 3.5.

The MIT economists made sure to assign their subjects tasks that resembled real-world writing tasks they might encounter on the job: 20- to 30-minute test assignments that included the work of writing press releases, drafting short reports, replicating planning document creation , and drafting delicate emails.

Further survey questions helped the researchers confirm that their participants did indeed find the tests comparable to tasks they had previously encountered in their own offices.

To ensure that the control group received an equivalent time-consuming tutorial, the subjects in this second group were instructed on how to use a collaborative writing tool called LaTeX editor Overleaf instead of ChatGPT.

Less than five percent of that control group bothered to use LaTex in their test assignments.

But the researchers not only tracked their subjects’ performance and use of ChatGPT, but also their subjects’ personal reactions to the AI ​​tool.

A third of those who reported not using ChatGPT in their post-study surveys said their work-related writing was too “tailored specifically” to their clients and required “real-time” or “unique” information about their company’s products.

The MIT economists found that those using ChatGPT enjoyed their tasks about 47 percent more than not just the average, but the standard spread of deviations from that average level of enjoyment.

The ChatGPT users were also noticeably more concerned, enthusiastic and optimistic about the future omens heralded by the wider adoption of AI in their industries and professions.

Nevertheless, these feelings quickly dissipated after the study’s 2-week and 2-month follow-up surveys as the participants got used to the new normal.

“They are best interpreted,” they wrote, “as short-term phenomena that reflect respondents’ first experiences with the technology.”