Facebook's David Marcus on how humans work alongside AI automation to power Facebook M and whether such a service can scale
Time To Meet The Wizard: Facebook's Messenger Head Pulls Back The Curtain On “M” — Facebook's David Marcus says M is already a lot more than just people pretending to be robots. Tweets: @vanessaluvtini and @rossdawson . Thanks: @jmbooyah Tweets: Vanessa Martini / @vanessaluvtini : Last night my car drove me home on autopilot & today I read this article by @Kantrowitz via @buzzfeed #futureisnow http://twitter.com/... Ross Dawson / @rossdawson : Very interesting insights into Facebook's M: AI formulates response, vetted and improved by humans, bot learns http://www.buzzfeed.com/... Thanks: @jmbooyah
Context & Ripple Effects
Facebook first put M into testing in August as an assistant inside Messenger that could complete real purchases on a user's behalf — a brief that pure AI assistants weren't touching. By early November, hands-on coverage found M outperforming Siri and Cortana at actual real-world errands, precisely because humans sat behind the curtain.
Marcus's interview now confirms the mechanism: AI drafts a response, humans vet and improve it, and the bot learns from the correction. That makes M both better than its rivals and structurally harder to scale — every task Facebook completes well is one a person helped complete.
First-order effects
- M's quality edge over Siri and Cortana is bought with human labor per request, so Facebook's expansion of the service is capped by hiring, not by server capacity.
Second-order effects
- The Bot Engine built on the Wit.ai acquisition is Facebook's route off that labor treadmill — machine-learning bots that can absorb what human vetting teaches M — and it underpins the plan to bring businesses and ads to Messenger through chatbots.
Third-order effects
- If the hybrid model holds, virtual assistants split into two tiers: cheap pure-AI assistants stuck at shallow tasks, and human-supervised services that transact in the real world but only where unit economics justify the staffing — with the wizard-behind-the-curtain setup doubling as the standard way to train automation.
The trend: Virtual assistants are converging on human-in-the-loop supervision as the bridge between command-and-response AI and agents that complete real transactions.