OpenAI is using a new five-level framework aimed at mapping out the path towards artificial general intelligence (AGI), according to a report from Bloomberg. The company, known for its cutting-edge AI models such as ChatGPT, seeks to bolster transparency regarding its safety measures and long-term goals.
Framework Breakdown
The new structure categorizes AI progress into five stages, starting with the capabilities that current AI models possess, which are centered around human-like conversational abilities. OpenAI is currently situated in this initial stage and is approaching the second stage, referred to as “Reasoners.” This phase will see AI systems solving basic problems similarly to a human with a Ph.D., but without utilizing external tools.
Sam Altman, OpenAI’s CEO, has previously projected that AGI might be achievable within this decade. The classification system serves as an integral part of OpenAI’s broader strategy for clarity and safety as AI technology advances. This detailed framework will be presented to both investors and stakeholders to provide a clearer understanding of the company’s milestones.
Employee and Stakeholder Communication
During an all-hands meeting, OpenAI’s leadership introduced the new classification system to its employees. Plans are also in place to brief external stakeholders, including investors, to proactively address safety and ethical concerns related to AI advancements.
Beyond “Reasoners,” the system includes “Agents” at Level 3, designed to perform tasks on behalf of users for extended periods. Level 4, known as “Innovators,” will feature AI capable of creating new innovations. The pinnacle, Level 5, termed “Organizations,” envisions AI systems carrying out the functions of an entire organization. This framework remains a work in progress and may evolve with ongoing feedback.
Expert Opinions on AGI Development
In October, both Sam Altman and Chief Technology Officer Mira Murati conveyed optimism to the Wall Street Journal that AGI could be developed within ten years. Nvidia CEO Jensen Huang suggested that AGI could appear within five years, should it be defined as passing any test. Elon Musk, on the other hand, has forecasted that AI might surpass human intelligence by 2025 or 2026, asserting that “sentient compute” could exceed human intelligence entirely within five years.
OpenAI is reportedly already achieving results that are far beyond the AI models that are publicly available. In November, we wrote about the company’s Q* model, a research project within the company. What seems to set Q* apart from GPT and potentially any other AI model is the level of self-awareness it is reportedly exhibiting. While it is unclear if the project is still active, it seems OpenAI is several steps ahead than what it is currently releasing.
Last Updated on November 7, 2024 3:37 pm CET
