What’s New in AI? Baidu’s Wenxin, GPT‑4 Multimodal, Docker’s Policy Shift & More

On March 16, Baidu officially released its knowledge‑enhanced large language model Wenxin Yiyan, positioning it against ChatGPT and GPT‑4. Baidu demonstrated the model’s abilities in literary creation, business copy, mathematical logic, Chinese comprehension, and multimodal generation, such as continuing the novel “Three‑Body” and naming tech companies.

Just two days earlier, OpenAI announced the multimodal GPT‑4 model, which can understand images, process up to 25,000 tokens of text, and deliver higher answer accuracy. GPT‑4 can also generate lyrics, creative texts, and follow complex instructions. An example image and GPT‑4’s humorous response are shown below.

Docker sent an email to all Docker Hub users with organizations, warning that without upgrading to a paid team plan, their accounts and images will be deleted. Free personal accounts remain for six months, while the cost for hosting public images in an organization jumps from $0 to $420 per year.

Microsoft announced a multi‑billion‑dollar investment in a supercomputer that powers ChatGPT, using thousands of Nvidia H100 and A100 Tensor Core GPUs and Quantum‑2 InfiniBand networking on Azure, enabling larger and more complex AI model training.

Google opened its PaLM API and introduced MakerSuite, a tool that helps developers quickly build generative AI applications, offering models optimized for conversational interaction, content generation, summarization, and classification.

According to a Stack Overflow survey of thousands of developers, the most proven technologies are open source (6.9), cloud computing (6.5), and machine learning (5.9). AI ranks first for future widespread use, followed by open source, machine learning, and cloud computing. Low‑code/zero‑code and blockchain received low adoption scores.

The U.S. government has demanded that ByteDance sell its TikTok shares, threatening a ban if the sale does not occur. TikTok plans to invest $1.6 billion to protect U.S. user data and content.