alignment problem — Dizionario inglese

Nome proprio

1. The problem of how to create a superintelligent artificial intelligence whose values would align with the interests of humankind.

   Esempi

   Altri esempi (1)

   As [Brian] Christian notes, the alignment problem bears a real resemblance to parenting: “The story of human civilization has always been about how to instill values in strange, alien, human-level intelligences who will inevitably inherit the reins of society from us—namely, our kids.”

   Properly defining and solving the alignment problem won’t be easy; it will require us to develop a broad, scientifically based theory of intelligence.

   Altri esempi (1)

   For years, AI ethicists have worried about this so-called alignment problem. In short: How do we ensure that the AI we build, which might very well be significantly smarter than any person who has ever lived, is aligned with the interests of its creators and of the human race? An unaligned superintelligent AI could be quite a problem.