Hands-on: Facebook M, powered by humans, excels at carrying out real-world tasks, unlike pure AI virtual assistants like Siri or Cortana
Nick Statt / The Verge :
Context & Ripple Effects
Facebook's M assistant entered testing inside Messenger in August as an AI system with human supervision, and Wired's companion explainer on how its AI actually works set up the question of where the line between machine and person falls. An early hands-on flagged the trade-off: promising as a Siri competitor, but slow on complex requests.
This new hands-on sharpens that picture — M excels at carrying out real-world tasks precisely because humans are doing them, unlike pure-AI rivals like Siri or Cortana. That reframes M less as a chatbot benchmark than as a service whose quality is bought with labor, which is why Marcus was soon fielding questions about whether such a service can scale.
First-order effects
- M delivers completed purchases and real-world errands that Siri and Cortana cannot execute, making it the most capable assistant in the hands-on comparisons even though its output depends on human operators.
Second-order effects
- Apple and Microsoft are left defending assistants that answer questions rather than finish tasks, while Facebook faces a cost structure its pure-AI rivals don't carry — hence the immediate scrutiny over how humans working alongside the automation can scale.
Third-order effects
- If the pattern holds, virtual assistance splits into two products: free, scalable, shallow voice assistants and paid-quality task completion backed by human labor — forcing every player to choose which side of that line to compete on.
The trend: Virtual assistants are diverging into pure-AI responders and human-in-the-loop task executors, with Facebook M betting that supervised labor beats autonomous AI on getting things done.