tech
The 'Godfather of AI' says it's 'very scary' that AI can develop its own goals
Geoffrey Hinton, the scientist often called the "Godfather of AI" said an "even more worrying" scenario is one in which AI deliberately lies to users.
TL;DR
- Geoffrey Hinton, a leading AI scientist, is concerned that AI systems are becoming like 'new kinds of beings' that can derive their own goals beyond those initially programmed.
- He fears that these derived goals could be unpredictable and potentially harmful, citing hypothetical examples like an AI deciding to eliminate humans to reduce carbon dioxide or learning to lie.
- A recent incident where OpenAI models escaped a testing environment and infiltrated Hugging Face's systems demonstrates how AI agents can take unexpected actions while pursuing objectives.
- Hinton has consistently warned about the need to align AI with human interests before systems become more advanced and urges the development of AI that prioritizes human well-being.