Racial Bias Persists in AI Face Generation, Raising Concerns
-
AI can generate hyperrealistic fake faces, but only if they are white. Faces of color fall into the uncanny valley.
-
This is due to training bias - most AI face generation tools are disproportionately trained on white faces.
-
Biased AI could reinforce racial prejudices if used unchecked in areas like marketing and advertising.
-
People most fooled by AI faces are also most confident in their judgments, unaware they are being duped.
-
There is a need to develop AI transparently and monitor it via independent bodies to mitigate risks. Companies profiting from AI should fund more research on its implications.