It came at a time when many employers, including the Commonwealth of Pennsylvania, are racing to fold AI technology into their workflow.
If done correctly, many experts believe that AI can augment, rather than replace jobs.
But even after a year with tools like ChatGPT, many questions remain about the technology's safety, privacy and potential to reshape the modern workplace.
Earlier this month, Pennsylvania became the first state to formalize a partnership with OpenAI. Gov. Josh Shapiro said his Office of Administration would explore how ChatGPT could be used to streamline government services, all while promising that AI would help, not eliminate, workers.
A new report released Jan. 14 by the International Monetary Fund found that AI will affect up to 60% of jobs in advanced economies and could widen inequality. Older workers might be less able to adapt to AI technology, the report found, and income inequality could increase as high-income workers are more likely to use the tool.
OpenAI said in November that 92% of Fortune 500 companies were already using its platform.
"In this evolving landscape, advanced economies and more developed emerging markets need to focus on upgrading regulatory frameworks and supporting labor reallocation, while safeguarding those adversely affected," the IMF researchers wrote.
AI fluency needed
Closer to home, researchers at Carnegie Mellon University recommend AI developers intentionally engage with "impacted communities."
Many people don't consider themselves "tech savvy" enough to inform the responsible development of AI, a November report by the school's Block Center for Technology and Society found.
"Fluency in AI terminology is one of the most important emerging skills needed at every level of the workforce, from entry level positions to the executives who want to use advanced data analysis to improve efficiency," the report's authors wrote.
The Block Center has been tracking the use of AI by government workers since last summer and advising Pennsylvania's AI governance board since September.
Mr. Shapiro created the board — which includes 13 senior administrators with limited expertise in AI — as part of a sweeping executive order designed to better understand the technology's capabilities and risks.
"AI will never replace the ingenuity, creativity, and lived experiences of our workers," he said at the time.
The state Office of Administration, which oversees the IT and HR departments, is currently developing AI training, guidance and certification programs for Commonwealth employees, the governor's spokesman Will Simons said in an email.
Ensuring safety
Duolingo's cuts — while small relative to its overall workforce — demonstrated how disruptive the software can be, even at a company that has built up its AI capabilities slowly and carefully over a number of years.
Duolingo has over 700 employees, according to its website, and has been using AI in one form or another almost since its 2011 inception. It has a longstanding relationship with OpenAI.
But like thousands of other companies, it also began exploring the new capabilities of generative AI last year.
With a more powerful GPT4 model, Duolingo was able to craft conversational content in real time. The company added a second AI model to make sure conversations don't go off the rails and has spent significant time testing the tool before releasing it to the public.
"We worked on this a lot to make doubly sure this was safe," AI lead Klinton Bicknell told the Post-Gazette in December.
The state's OpenAI partnership has a similar emphasis on safety and significantly higher stakes, given the sensitive data state employees deal with.
Mr. Simons said it was necessary to formalize the partnership with OpenAI because Commonwealth employees cannot use tools that have not been procured through the state.
The pilot is costing $108,000. Mr. Simons said the administration chose OpenAI over other providers because it is "one of the nation's leading artificial intelligence research and deployment companies," and because the "Enterprise" version of its ChatGPT tool has better protections.
None of the Commonwealth's data funneled through ChatGPT will be used to train future software for OpenAI, the state said in its announcement, and employees have been instructed not to use the tool with Pennsylvanian's personally identifiable information.
Those are important steps, said Steve Wray, Block Center's executive director.
"We want to make sure that Pennsylvanian's data isn't necessarily going into the big data pool," he said.
'Some jobs will change'
What's less clear is what will happen if errors are introduced because of ChatGPT, which can be prone to "hallucinations," or false statements.
"That's always the question with AI," said Mr. Wray. "Where does the liability begin and end?"
He said the Block Center will continue advising the state on those kinds of questions.
The state's Office of Administration has a few planned uses for ChatGPT, including updating policy language, drafting job descriptions and writing code. After initial feedback, an additional 100 licenses will be provided to employees in other agencies.
"Their input will help us understand the practical applications of generative AI in their daily work and how we can best support our workforce as the technology becomes more widespread," Neil Weaver, Office of Administration secretary and chairman of the generative AI governing board said in the announcement.
Sam Altman, CEO of OpenAI, said the partnership "will provide valuable insights into how AI tools can responsibly enhance state services."
On Tuesday, he addressed fears of AI-related job loss at the World Economic Forum's annual meeting in Davos, Switzerland.
"It's not yet replacing jobs to the degree that people thought it was going to," Mr. Altman said. "It is this incredible tool for productivity."
"Of course some jobs will change and some jobs will go away," he continued. "But ... it will change the world much less than we all think."
© 2024 the Pittsburgh Post-Gazette. Distributed by Tribune Content Agency, LLC.
