OpenAI has announced plans to release its o3-mini model, which is a successor to the o1 model released last September. The company’s CEO, Sam Altman, had previously stated that the new model would be more powerful than the previous one but did not claim it achieved artificial general intelligence (AGI).
However, a recent post by Altman on X contradicts his earlier statement. He wrote that Twitter hype is out of control and that OpenAI is not deploying AGI next month or building it. Instead, he claimed the company has “some very cool stuff” coming but urged people to cut their expectations.
Despite this, OpenAI introduced its o3 model in December as part of its “12 Days of OpenAI” campaign, and a Reuters report stated that the new models are capable of reasoning through complex tasks and solving challenging problems. The o3-mini model is also being safety tested, and Altman said it has passed the ARC-AGI challenge, which is considered a benchmark for AGI.
However, not everyone agrees that OpenAI’s o3 model has achieved AGI. François Chollet, an independent software engineer and AI researcher, stated that passing ARC-AGI does not equate to achieving AGI and that the model still fails on some easy tasks. He also warned that the upcoming ARC-AGI-2 benchmark may pose a challenge to the o3 model.
Brian Jackson, principal research director at Info-Tech Research Group, took a more pragmatic view. He said that while the community expects AGI to be achieved this year or next, it’s not clear what AGI means. Instead of focusing on whether one company or specific model has achieved it, CIOs should focus on AI implementation and execution guided by a responsible AI governance framework.
Jackson emphasized that CIOs are more concerned with automating complex knowledge-based tasks, setting up infrastructure environments they can trust, managing costs, and monitoring performance to ensure they meet service-level agreements.
Source: https://www.cio.com/article/3806039/altman-now-says-openai-has-not-yet-developed-agi.html