A recently published survey from the Pew Research Center found that more than half of the 11,000 Americans surveyed say they think the impact AI in the workplace will have on the economy will be major. And that was even before Chat GPT really took off.
Subbarao Kambhampati is a professor at ASU’s School of Computing and Artificial Intelligence. He said the survey gives a glimpse into how public opinion might have shifted, post-Chat GPT.
“I wonder whether some of that optimism gap would have been reduced after the [popularization of] Chat GPT because some of the knowledge work — things related to writing and so on — have become easier to at least fake, if not do it right,” Kambhampati said.
Pew found that over half of respondents say they think using artificial intelligence will have an effect on workers, but less than a third think it will affect them personally.
“Everybody thinks that AI is going to have a big effect on other people’s jobs,” Kambhampati said. “But they don’t think their own job is particularly going to be affected.”
Pew found that nearly half of respondents say AI would be better at treating all applicants the same. But 44% also say AI would not measure up in thinking outside the box for applicants that don’t perfectly fit a job description.
Kambhampati said sometimes the whole web is an AI’s “box,” compared to our more relative “boxes” of knowledge as humans.
“Thinking out of the box is easy if your box is small,” he said. “ If your box is earth-sized, then it’s a very different kind of thinking out of the box.”
Kambhampati said routine or knowledge-based white collar jobs may be some of the first to feel AI’s influence.
“The fact that AI systems can do pretty compelling writing would probably have changed some of the people’s minds,” Kambhampati said.
Pew found that nearly half of respondents say AI would be better at treating all applicants the same. But 44% also say AI would not measure up in thinking outside the box for applicants that don’t perfectly fit a job description.
“One of my bigger worries is that people are more easily impressed by [AI] than not,” Kambhampati said. “The respondents of this survey were a little more skeptical. I’m actually happy because I think people tend to assume [AI] can do everything.”
Right now, Kambhampati said AI systems tend to have a broad but shallow understanding of many things. Even if an AI system can consistently answer questions, he said, it may answer some or all of them incorrectly.
“Consistency is, by itself, not a great thing because you can be consistently wrong,” he said. “But machines, at least, will show the patterns that we didn’t notice ourselves in our own flawed decisions making in hiring positions.”
In terms of potential bias in hiring positions, Kambhampati said we should also be skeptical: “Before we jump ahead and say that AI system had bias, where did that bias come from?”
What data an AI learns from, he said, is key. So is who that data comes from.
“In a way, [biased AI systems are] just reflecting us, back to us,” Kambhampati said.
Overall, Kambhampati said AI bringing changes to the workplace isn’t an unexpected effect.
“My sense is that there is going to be a short-term disruption,” he said. “Because all technologies disrupt employment.”
