On December 20, 2024, OpenAI announced that it had achieved artificial general intelligence (AGI) with its latest model, o3. This was a major milestone in the field of AI, and it has been met with a mix of excitement and apprehension. o3 was able to achieve AGI by exceeding the 85% threshold on the Arc AGI test. This test is designed to measure a model's ability to generalize to new tasks and solve problems that are easy for humans but difficult for AI. o3 was able to achieve this by using a new paradigm of reinforcement learning on Chain of Thought to scale inference compute. This allowed it to use more resources to think about the questions and come up with the correct answers. Some people believe that o3's achievement is a major breakthrough that will lead to a new era of AI innovation. Others are more cautious, arguing that o3 is still not truly intelligent and that its performance on the Arc AGI test is not representative of its overall capabilities. (summary developed by Gemini 1.5 GenAI)
