After 12 days of technical sharing livestream events, OpenAI released its next-generation inference model o3 on the final day. This is an upgrade following the earlier release of the o1 inference model. The o3 model series includes o3 and o3-mini, with o3-mini being a smaller, streamlined version fine-tuned for specific tasks. OpenAI stated that the o3 model could approach the realization of artificial general intelligence (AGI) under certain conditions, meaning artificial intelligence capable of performing any task that a human can accomplish.

Well-known AI assistant Grammarly recently announced its acquisition of productivity platform Coda, marking an important step for Grammarly towards becoming an AI productivity platform. Image source note: Image generated by AI, image licensed through provider Midjourney. Coda's CEO and co-founder Shishir Mehrotra will take over as the new CEO of Grammarly. This acquisition will combine Coda's flexible and powerful AI.

In a recent series of Threads posts, Instagram head Adam Mosseri emphasized the need for social media users to be vigilant in the face of the growing prevalence of AI-generated content. He pointed out that the images and information users see online may not be real, as AI technology is now capable of creating content that closely resembles reality. To help users better assess the authenticity of information, Mosseri mentioned that users should consider the source of the content, and social platforms also bear responsibility.

Adam Mosseri, the head of Instagram, stated on social media that users should be wary of images they see online, especially those that can easily be mistaken for reality, such as AI-generated content. Mosseri emphasized that AI technology has significantly enhanced its ability to produce realistic content, so users should approach this content with caution and consider its source, while social platforms have a responsibility to assist in this matter. He wrote, 'As an internet platform, it is our duty to label AI-generated content as much as possible.'

DeepSeek AI recently launched DeepSeek-V2.5-1210, an enhanced version of DeepSeek-V2.5, aimed at improving AI performance in math, programming, writing, and reasoning tasks. Earlier versions of the model achieved some success in solving math and reasoning tasks, but the stability of performance across various applications needed enhancement, especially in real-time coding and meticulous writing. These shortcomings underscore the potential for developing a more flexible and reliable AI model.

Tencent Cloud recently announced the launch of its AI coding assistant - Tencent Cloud AI Code Assistant. This tool aims to assist programmers in writing code by predicting and providing code suggestions, significantly enhancing development efficiency. The core advantage of this AI code assistant lies in its ability to deeply understand the context of the code, offering more accurate logical and grammatical code completion suggestions, surpassing traditional keyword-matching completion methods. It can also learn and adapt to a programmer's coding style and historical code patterns, providing customized code completions that align more closely with individual habits.

Perplexity AI recently announced that its publisher program has welcomed over a dozen new international media partners, marking a new stage in its integration with the news industry. New partners include ADWEEK, Los Angeles Times, Mexican Daily News, Stern, NTV, and Encyclopedia of World History, while also expanding the existing collaborations with Time Magazine, Fortune Magazine, and Der Spiegel. As part of the collaboration, Perplexity will work with publishers to enhance the attractiveness of the partnership.

On December 5th, the Byte Bean Bag model team launched the latest code model evaluation benchmark - FullStack Bench, covering over 11 real-world scenarios, supporting 16 programming languages, and including 3374 questions. Compared to previous evaluation standards, this benchmark can more accurately assess the code development capabilities of large models across a broader programming domain, facilitating the optimization of models in real-world programming tasks. Current mainstream code evaluation benchmarks, such as HumanEval and MBPP, typically focus on basic and advanced.

Now, scientists can precisely "design" proteins as if they are programming. This is no longer science fiction but a reality being created by the Dutch company Cradle. In the grand tapestry of biotechnology, Cradle's founder, Stefan van Groningen, likens proteins to an "alien programming language." His team is utilizing artificial intelligence to decipher this mysterious language, bringing revolutionary changes to biotechnology. Traditional protein development has been like a long and costly blind exploration. Researchers must go through dozens of experimental rounds, each of which may...

Recently, OpenAI, in collaboration with the nonprofit organization Common Sense Media, launched a free training course on artificial intelligence (AI) designed to help teachers understand AI technologies and prompt engineering. This new initiative was officially released on Wednesday, aiming to provide educators with the skills and resources to effectively and responsibly integrate AI tools into the classroom. Image source note: Image generated by AI, image licensed from provider Midjourney.