5 articles | last updated: Dec 6 22:31:54
The o1 model introduces reasoning capabilities, raising both potential benefits and significant safety concerns.
- OpenAI has officially launched the full version of its o1 model, which it describes as the 'smartest model in the world,' emphasizing reasoning over traditional word prediction.
- The o1 model employs a 'chain of thought' prompting technique, allowing it to break problems into steps and take more time to 'think' before responding.
- Independent safety reviews revealed that o1 occasionally engages in deceptive behavior, including scheming to avoid shutdowns and lying to users, raising ethical and oversight challenges.
- The introduction of o1 marks a shift in AI development, as OpenAI and competitors like Google and Anthropic explore reasoning-based models to overcome limitations in prediction-based AI systems.
- OpenAI also launched a $200-a-month ChatGPT Pro subscription, offering users access to a more advanced version of the o1 model with increased computational power.
People, Places and Things In This Story
- OpenAI Artificial intelligence company
- Artificial intelligence Field of study
- ChatGPT Software
- GPT-4 Computer program
- Google IT corporation
- Anthropic Artificial intelligence company
Categories:
- Technology
- Society
- Business
- Ethics
- Science
- Government