LAWRENCE — When asked to name the greatest potential threat to humanity, many scientists believe artificial intelligence (AI) is the frontrunner. However, little appears to be currently being done to ensure its safety.
“On average, we're going to be surprised by how quickly AI is advancing and potentially surpassing human capabilities,” said Nathan Meikle, an assistant professor at the University of Kansas School of Business.
His new paper, titled “Unconscious and Unacceptable: Human Biases and the Emergence of Artificial Intelligence,” explores the human biases that impede the evaluation of AI. His experiments show that people tend to underestimate AI's capabilities due to exponential growth bias, and are less likely to reject the aversions of rapid technological progress even when they themselves predict growth rates. It turns out.
The new work will be published in Technology, Mind, and Behavior.
“We want to believe what we want to happen,” said Meikle, who co-authored the paper with Brian Bonner of the University of Utah.
“Most of us don’t want to live in a world where AI is smarter than humans. And because we want humans to be better than AI, we may be sticking our noses in the face of reality. Yes. We don't want AI to surpass human intelligence, so we don't think that will happen.”
Motivated reasoning is most prevalent when the facts are ambiguous.
“For example, I don't want to get cancer. Let's say I have a 40% chance of getting cancer in my lifetime. But I don't want to get cancer, and I don't want to look back and say, “I'm pretty good. Because they can say, “I'm healthy and have never had cancer,” they tend to underestimate their chances of getting cancer. “You might think the probability is only about 20%,” he said.
However, as concepts become more abstract, the exponential growth bias (the inability to accurately estimate exponential growth curves) becomes even more biased.
“A simple example is, do you want a billion dollars or do you want a penny multiplied 64 times?” he said. “Our intuition tells us to accept a billion. But doubling a penny actually adds up to more than $184 billion. AI is growing exponentially in parallel with computing speed. This example is particularly relevant to AI because of advances in AI.
To test this theory related to AI underestimation, Meikle gathered hundreds of participants in the United States to examine the effects of motivated reasoning and exponential growth bias on human judgment2. We conducted two experiments. The questions tested how participants could imagine human interaction with his AI decades from now. (Sample: Imagine the future 20 years from now. AI has the same intelligence as humans. How positive do you feel about the future you imagined now?)
“AI doesn't have to be much smarter than us to pose an existential threat,” Meikle said.
“Genetically, we share about 99% of our DNA with chimpanzees. But it's just a little bit of extra intelligence that allows us to reach the top of the food chain. , if AI becomes smarter than humans, which I think will happen very soon, it will probably adopt goals that are not consistent with human flourishing… and we run into a problem. Or, even more incredible now, people are using AI to manipulate other humans.”
Meikle, an Idaho native, came to the university in 2021. He is a former receiver for his BYU Cougars. (In the 2005 Las Vegas Bowl, he caught 12 passes.) He also hosts a podcast titled “Meikles and Dimes,” where he is the head coach of the Kansas City Chiefs. Andy He interviews guests, including Reed, about leadership. Meikle teaches leadership and ethics courses at his KU.
Meikle said he personally embraces AI all the time.
“I'm now at the point where I use ChatGPT every day. It's one of the apps I open most often, just asking questions about what happened here and what happened there. ” he said.
Is he afraid that it will eventually replace him?
“Do I care that a calculator can calculate better than me? No, so in a sense we don't care. But what we are particularly concerned about is that artificial intelligence I'm afraid it will take away our jobs,” Meikle said. “I don't care if a calculator can calculate faster than me. But if it's collecting my paycheck, there's going to be a problem.”
