- Nvidia CEO Jensen Huang discusses the AGI timeline and AI hallucinations at GTC conference.
- Huang suggests AGI could be achieved within five years if clearly defined.
- He proposes solving AI hallucinations by ensuring answers are well-researched.
During Nvidia’s annual GTC developer conference, CEO Jensen Huang tackled questions about artificial general intelligence (AGI) and its potential timeline.
Huang emphasized the importance of clearly defining AGI, drawing parallels to knowing when New Year’s Day arrives or recognizing a destination by its landmarks.
He suggested that if AGI were defined by specific tests, such as passing a legal bar exam or pre-med exam, it could be achieved within five years. However, without a precise definition, Huang refrained from making predictions.
Solving AI hallucinations through research
Huang also addressed the issue of AI hallucinations, expressing frustration at the question. He proposed a straightforward solution: ensuring AI answers are well-researched.
By employing retrieval-augmented generation, an AI should conduct research before providing an answer, examining sources and contexts to determine the best response. For mission-critical answers, such as health advice, Huang recommended checking multiple resources and known sources of truth.
Additionally, he noted that AI should have the ability to admit uncertainty or lack of information when appropriate.
Huang’s solution to the issue
The GTC conference provided a platform for Huang to share his insights on the future of AI and address concerns surrounding AGI and AI hallucinations.
By emphasizing the need for clear definitions and research-based answers, Huang highlighted the importance of responsible AI development and deployment.
As the field of AI continues to evolve, discussions like these are crucial for shaping its trajectory and addressing the challenges that lie ahead.
