What exists and what does not.
NARROW AI
Systems that do one kind of task. Everything currently in use.
A language model generates text. An image classifier recognises objects. Neither does the other's job.
Narrow does not mean weak. Narrow systems can substantially exceed human performance at their specific task.
GENERAL AI
A hypothetical system matching human capability across any intellectual task.
Does not exist. Whether it will, and when, is genuinely contested among people who study it.
WHY THE DISTINCTION MATTERS
Marketing frequently implies general capability. Products are narrow.
A language model that writes well does not therefore reason reliably about your accounts.
WHAT THIS MEANS FOR YOU
Judge a tool by what it demonstrably does, not by the category it claims.
Test it on your actual task rather than trusting a demonstration.
THE PRACTICAL POSITION
Current systems are useful tools with specific strengths and predictable failure modes.
Treating them as general intelligence produces disappointment and occasionally harm.
