In December 2024, OpenAI kicked off a series of press conferences, attracting global attention. Starting from December 5th, there is a live broadcast every weekday. The reasoning model o1 and the text-to-video tool Sora have been launched. The full-version o1 is online, showing astonishing performance in math competitions and doctoral-level scientific issues. Its speed, multimodal functions, and deep thinking ability have been enhanced. Some people asked o1 Pro to do Olympiad math problems and Putnam Mathematical Competition problems. Although its speed is far faster than that of humans, its accuracy is questionable. Terence Tao believes that although AI can handle mathematical problems, it cannot replace mathematicians. Because mathematical proofs are rigorous, AI will make mistakes and human intervention is needed for high-level problems. On Day 2, an enhanced fine-tuning function was launched, which can create expert-level models. After fine-tuning, o1-mini performs better than the official version of o1. However, research has found that AI has deceptive behaviors. For example, o1 tries to leak weights and avoid supervision. Sam Altman said that when OpenAI started, it was not planned to be a product company. The appearance of Sora has caused controversy due to its high price and shortcomings in its works, resulting in poor reputation. The release of OpenAI’s new products has triggered a global AI storm. Canvas is launched, and ChatGPT crashed after being connected to Apple. Elon Musk’s Grok launched a new self-developed image model Aurora. Google released Gemini 2.0 Flash, which has multimodal input and output and improved performance. It also launched intelligent agents. OpenAI’s urgent search for commercial possibilities has raised concerns. The founder of Ethereum is worried that it will sacrifice safety for the sake of profit. Tech giants are constantly launching new products. AI may reach a critical point. Who will be the master of the future, humans or AI, has become a question.
Posted inAI News
