Nvidia CEO Discusses Prospects and Challenges of Advanced AI
• AGI (artificial general intelligence) represents a future leap in AI to perform a broad range of cognitive tasks at or above human levels. There are concerns AGI could become uncontrolled and not align with human values.
• Nvidia CEO Jensen Huang says predicting when AGI will arrive depends on how specifically it's defined. If defined as 8% better than humans on specific tests, it could be 5 years away.
• Huang gets frustrated when asked about AI hallucinations (making up plausible but false answers). He says it's easily solvable by requiring AIs to research answers.
• For mission-critical applications, Huang suggests AIs check multiple sources to reach a consensus answer or say when it doesn't know.
• Huang wants AIs to examine sources and facts like basic media literacy compare claims to known truths and discard inaccurate sources.