Surprise, Meta's latest AI gimmick is just underpaid humans

Meta's Muse, promoted as an AI that can place outbound calls on users' behalf, relies heavily on real humans to do the work. A report from 404 Media revealed an internal setup where Muse handles dialing and call management only until it hands complex tasks off to "trained human agents" who place and complete the calls, then return transcripts and summaries. Internal message-board posts show employees worried about the optics and potential backlash, urging safeguards or opt-in controls to avoid coverage that would portray the product as merely masking human labor because the AI "is not good enough."

This is presented as part of a broader industry pattern where companies sell the illusion of automated intelligence while employing low-paid workers - often in countries like Venezuela and Kenya - to train and moderate systems. Critics argue that calling the system AI while hiding human operators amounts to exploitation and deceptive marketing, and that Muse's utility is limited outside of enabling problematic uses. Employees' concerns and the exposed human layer position Muse not as a breakthrough but as a performative gimmick that risks negative PR and reinforces long-standing labor and ethical issues in AI deployment.