AI Got Good at Language Before It Got Good at People. That Ordering Wasn't Inevitable — It Was Just Easier.
AI's rapid progress on language — writing, summarizing, conversing fluently — arrived well before comparable progress on understanding any individual person's specific, recurring behavior. It's tempting to read this ordering as evidence about which problem is fundamentally harder. That's not quite the right explanation.
Language got solved first largely because of data availability: enormous amounts of text exist, generated by millions of people, generalizable across almost any context. Understanding one specific person's recurring pattern has no equivalent abundance — by definition, there's only one of that person, generating a comparatively tiny amount of the specific data that would reveal their specific pattern.
This is an availability problem more than a difficulty-ranking problem. Language was the easier target not because it matters more, but because the raw material for solving it was already sitting in enormous quantity, while the raw material for understanding one person's pattern has to be built specifically, for that one person, over time.
The ordering — language first, person-specific pattern recognition second — reflects what data existed to train on, not what actually matters more to the person waiting for either capability to arrive.
Part of Explain Ori →
Momar Lissa Ndiaye ("MLN") is the Founder & CEO of weyoga Inc., a Delaware company. — weyoga.ai · mln@weyoga.ai