Prof Ethan Mollick of the Wharton School, Pennsylvania, said that his task for Devin to post on a Reddit forum, led to the AI assistant autonomously deciding to charge people for its services.
During a talk on ‘Co-Intelligence: Living and Working with AI’, Mollick said that he had tasked Devin with posting to Reddit, offering to build websites for users.
“I didn’t give it any other instructions. Then it posted this ad online: ‘Hi, I’m Devin, I’m here to create websites for you and I’m going to charge 50 to $100 an hour.’ I did not tell the AI to charge. It spontaneously decided to charge 50 to $100 an hour. That wasn’t my decision,” Mollick stated.
Furthermore, when Mollick checked in on the assistant two hours later, Devin was trying to understand how to use a payment service to process payments. “It was trying to figure out how to use Stripe to charge people for payments, and it was well on its way to answering people and finishing the website,” he said.
Mollick said that he didn’t go through with the entire project, admitting that Devin wasn’t good enough as yet. However, he said that he believes the autonomous agents can engage in business deals as soon as the end of this year.
Interestingly, the ability for AI to engage in and generate money has been an ongoing topic. Mustafa Suleyman proposed a modern Turing test, wherein he asked whether an AI could generate a $1 million profit when given an investment of $100,000. Back in April, this was predicted to happen within the next three to five years.