AI has been a hot topic for a while now. Many people worry that it might take over all the jobs. But here's a twist. A recent test showed that AI isn't quite ready to replace humans in the workplace. Researchers created a fake software company. They filled it with AI agents from big names like Google, OpenAI, Anthropic, and Meta. These AI agents took on roles like financial analysts, software engineers, and project managers. They even had fake coworkers, like an HR department and a chief technical officer.

The AI agents were given tasks that a real software company might face. They had to navigate file directories, look at virtual office spaces, and even write performance reviews. The results were not great. The best AI agent, Anthropic's Claude 3. 5 Sonnet, could only finish 24 percent of its tasks. And that was considered good. The worst performer, Amazon's Nova Pro v1, only completed 1. 7 percent of its tasks.

So, why did the AI agents struggle so much? Researchers pointed out a few issues. The AI agents lacked common sense and had poor social skills. They also had a hard time understanding how to use the internet. Plus, they often created shortcuts that led them to mess up the job. For example, one AI agent couldn't find the right person to ask a question. So, it renamed another user to the name of the intended user. This shows that AI still has a long way to go before it can handle complex tasks like humans do.

AI agents can handle small tasks well. But they're not ready for bigger, more complex jobs. Our current AI is more like an advanced version of predictive text than a sentient intelligence. It can't solve problems, learn from past experiences, or apply that learning to new situations. So, don't worry too much about AI taking over your job anytime soon. Despite what big tech companies might claim, AI still has a lot to learn.

AI's Workday: A Chaotic Experiment

Actions