Nvidia CEO Jensen Huang on Tuesday reiterated his view that he expects to see the emergence of “artificial general intelligence” within five years, but, he added, it all depends on how you define the term.

Most people think of AGI as software that is sentient—that can think independently. Huang’s answer suggests that the arrival of AGI might be hard to pin down.

One approach, he said at a press conference at the GTC developer conference in San Jose, is to specify that AGI has broad knowledge of many fields.

Imagine he says, that software can do very well on a broad set of tests, “math tests, reading tests, reading comprehension tests, logic tests, medical exams, bar exams…GMATs, SATs, you name it, a bunch of tests.”

If you define AGI as doing well on those tests, better than almost any human being, he says, and you ask “do you think a computer could do that in five years, then answer is probably yes.”