Humans 'Disguised' as Advanced AI: The Surprising Reason Behind the Experiment

Meta has conducted an unusual trial regarding its latest artificial intelligence (AI) technology. The tech giant involved human personnel to handle services that were intended to be performed by their newest digital agent, Muse.

The trial, referred to as a ‘human concierge’ service, involves human contractors handling a portion of the telephone calls made by the AI agent, as reported by Reuters.

The company notified employees about this trial last week. A Vice President of Meta’s Superintelligence Labs unit stated that the company is still working to refine the AI’s calling features.

Human involvement was found to be able to boost success rates to the 95-98% range, significantly higher than if the process were performed entirely by machines. This further proves that advanced AI systems cannot yet fully replace humans, despite being touted as the future of technology capable of further enhancing autonomous systems.

A Meta spokesperson claimed that employee response to the trial has been quite positive. They also stated that the purpose of this testing is to gather feedback so the company can implement security and privacy protections, as well as refine features before a wide-scale release.

Despite the positive outlook, an internal post cited by Reuters mentioned that some employees are concerned about privacy issues. They warned of the risk of sensitive information being accidentally disclosed through the call centre services.

Furthermore, in one conversation transcript, a human contractor was caught making racist remarks during a call.

On the other hand, data from market intelligence firm Sensor Tower reveals that the Muse application has been downloaded over 2.5 million times as of early this month. The agent is designed to perform various tasks autonomously, such as sending emails, shopping, and booking travel on behalf of users. Another key feature is its ability to make phone calls, where the agent contacts specific numbers to manage various user requirements.