During an internal all-hands meeting, OpenAI discussed a new classification system for developing AGI (Artificial General Intelligence). Bloomberg has now revealed the details shared by OpenAI with its employees. OpenAI has proposed the following five levels to track its progress toward building AGI:
- Level 1: Chatbots: AI with conversational language capabilities
- Level 2: Reasoners: AI with human-level problem-solving abilities
- Level 3: Agents: Systems that can take actions on behalf of users
- Level 4: Innovators: AI that can aid in invention and discovery
- Level 5: Organizations: AI that can perform the work of an entire organization
Currently, OpenAI has reached Level 1 and is on the cusp of reaching Level 2. The upcoming Reasoners level AI system will be able to perform basic problem-solving tasks with the power of a human with a doctorate-level education.
During the internal meeting, OpenAI also demonstrated a research project using its GPT-4 model, showcasing new skills involving human-like reasoning. OpenAI will be sharing more information on the AGI levels with investors and the general public in the near future.
OpenAI is not the first organization to come up with such levels for AI development. Google Research"s DeepMind team recently developed a framework for classifying the capabilities and behavior of Artificial General Intelligence (AGI) models.
Google DeepMind"s Levels of AGI:
- Level 0: No AI
- Level 1: Emerging equal to or somewhat better than an un skilled human
- Level 2: Competent at least 50th percentile of skilled adults
- Level 3: Expert at least 90th percentile of skilled adults
- Level 4: Virtuoso at least 99th percentile of skilled adults
- Level 5: Superhuman outperforms 100% of humans
The revelation of OpenAI"s internal classification system for AGI development, along with DeepMind"s similar framework, marks a significant step in the ongoing quest for Artificial General Intelligence. These structured approaches not only provide a roadmap for progress but also foster transparency and facilitate comparison between different AI models.
Source: Bloomberg