Evidence Suggests That Companies Should Exercise AI Caution
In recent years, several companies have replaced portions of their workforce with AI. Klarna CEO Sebastian Siemiatkowski has been unwavering in his belief that AI can, and will, automate most roles that are currently occupied by humans. In February, he told Bloomberg: “I am of the opinion that AI can already do all of the jobs that we, as humans, do. It’s just a question about how we apply it and use it.”
However, a recent study from Carnegie Mellon contradicts this belief. Researchers created a fake company that was run by different AI models, which were then instructed to complete everyday tasks that you would find in a software startup. The researchers found that, at best, the AI agents only successfully completed 24% of the tasks. In other words, the AI-staffed company was a complete failure.
Undeterred, Duolingo will press ahead with their plan to automate contract work. The company should definitely exercise a note of caution, however. These findings indicate that AI is not yet the silver bullet that many have hoped.