The diverging path of China's two leading AI players shows where the country's artificial intelligence industry is headed. DeepSeek and ByteDance, the two leaders of China's AI industry, are adopting vastly different strategies. On Monday, DeepSeek released DeepSeek V3.2, another open-weight model that anyone can tinker with. The startup says it performs on par with the latest models from OpenAI and Google, and it even beats them on some key mathematics benchmarks. That same day, ByteDance, whose dominance in AI applications we covered previously, introduced ways for people to use its chatbot, Doubao.

Vision language model (VLM)-based mobile agents show great potential for assisting users in performing instruction-driven tasks. However, these agents typically struggle with personalized instructions -- those containing ambiguous, user-specific context -- a challenge that has been largely overlooked in previous research. In this paper, we define personalized instructions and introduce PerInstruct, a novel human-annotated dataset covering diverse personalized instructions across various mobile scenarios. Furthermore, given the limited personalization capabilities of existing mobile agents, we propose PerPilot, a plug-and-play framework powered by large language models (LLMs) that enables mobile agents to autonomously perceive, understand, and execute personalized user instructions. PerPilot identifies personalized elements and autonomously completes instructions via two complementary approaches: memory-based retrieval and reasoning-based exploration. Experimental results demonstrate that PerPilot effectively handles personalized tasks with minimal user intervention and progressively improves its performance with continued use, underscoring the importance of personalization-aware reasoning for next-generation mobile agents. The dataset and code are available at: https://github.com/xinwang-nwpu/PerPilot

We introduce a new problem KTRL+F, a knowledge-augmented in-document search task that necessitates real-time identification of all semantic targets within a document with the awareness of external sources through a single natural query. This task addresses following unique challenges for in-document search: 1) utilizing knowledge outside the document for extended use of additional information about targets to bridge the semantic gap between the query and the targets, and 2) balancing between real-time applicability with the performance. We analyze various baselines in KTRL+F and find there are limitations of existing models, such as hallucinations, low latency, or difficulties in leveraging external knowledge. Therefore we propose a Knowledge-Augmented Phrase Retrieval model that shows a promising balance between speed and performance by simply augmenting external knowledge embedding in phrase embedding. Additionally, we conduct a user study to verify whether solving KTRL+F can enhance search experience of users. It demonstrates that even with our simple model users can reduce the time for searching with less queries and reduced extra visits to other sources for collecting evidence. We encourage the research community to work on KTRL+F to enhance more efficient in-document information access.

Given that China already bans Google, Facebook, Twitter, and a host of foreign news websites (including time.com) In fact, ChatGPT parent company OpenAI's decision not to launch in China--Chinese and even Hong Kong phone numbers aren't permitted to sign up--appears to preempt that very fact, with the San Francisco-based firm telling Reuters that "conditions in certain countries make it difficult or impossible" to operate. Read More: Why China, Russia's Biggest Backer, Now Says It Wants to Broker Peace in Ukraine Nevertheless, canny Chinese netizens have found numerous workarounds to access the revolutionary service, such as using virtual private networks and an overseas friend's phone number; purchasing logins via online marketplace Taobao; or simply taking advantage of a variety of proxy bots embedded in ubiquitous messaging service WeChat. Chinese social media was so abuzz with ChatGPT content this month that one AI-generated fake government notice rescinding traffic regulations sparked bedlam and a police investigation in the eastern city of Hangzhou. Unsurprisingly, China's government has now stepped in with explicit bans on WeChat hosting proxy ChatGPT services, while a strident frontpage op-ed on the perils of investing in AI-related firms (and cited ChatGPT), which published earlier this month in the state-owed Securities Times newspaper, was linked to a fall in Chinese tech stocks.

Microsoft-backed OpenAI has kept its hit ChatGPT app off-limits to users in China, but the app is attracting huge interest in the country, with firms rushing to integrate the technology into their products and launch rival solutions. While residents in the country are unable to create OpenAI accounts to access the artificial intelligence-powered (AI) chatbot, virtual private networks and foreign phone numbers are helping some bypass those restrictions. At the same time, the OpenAI models behind the ChatGPT programme, which can write essays, recipes and complex computer code, are relatively accessible in China and increasingly being incorporated into Chinese consumer technology applications from social networks to online shopping. The tool's surging popularity is rapidly raising awareness in China about how advanced U.S. AI is and, according to analysts, just how far behind tech firms in the world's second-largest economy are as they scramble to catch up. "There is huge excitement around ChatGPT. Unlike the metaverse which faces huge difficulty in finding real-life application, ChatGPT has suddenly helped us achieve human-computer interaction," said Ding Daoshi, director of Beijing-based internet consultancy Sootoo.

Technological breakthroughs in the U.S. never fail to inspire challengers, followers and opportunists in China. It's ChatGPT's turn to capture the imagination of the world's largest internet population. On WeChat, ChatGPT's "trending index," an indicator of a keyword's popularity on the social network, rose 155 folds within the last 30 days. It's fascinating to watch how OpenAI's powerful language model sparks great interest among the country's tech giants, startups and ordinary people, not least because it offers a lens to understand the state of the AI race between two superpowers. Unlike many other major Western internet platforms, the ChatGPT site isn't blocked in China, yet.

Nowadays, Multi-purpose Messaging Mobile App (MMMA) has become increasingly prevalent. MMMAs attract fraudsters and some cybercriminals provide support for frauds via black market accounts (BMAs). Compared to fraudsters, BMAs are not directly involved in frauds and are more difficult to detect. This paper illustrates our BMA detection system SGRL (Self-supervised Graph Representation Learning) used in WeChat, a representative MMMA with over a billion users. We tailor Graph Neural Network and Graph Self-supervised Learning in SGRL for BMA detection. The workflow of SGRL contains a pretraining phase that utilizes structural information, node attribute information and available human knowledge, and a lightweight detection phase. In offline experiments, SGRL outperforms state-of-the-art methods by 16.06%-58.17% on offline evaluation measures. We deploy SGRL in the online environment to detect BMAs on the billion-scale WeChat graph, and it exceeds the alternative by 7.27% on the online evaluation measure. In conclusion, SGRL can alleviate label reliance, generalize well to unseen data, and effectively detect BMAs in WeChat.

A patient walks under the guidance of AR navigation inside one building of Xinhua Hospital. Leading public hospitals in Shanghai are using artificial intelligence on their WeChat to streamline patient registration and provide better treatment. The Internet hospital service has been widely adopted by hospitals and patients since the COVID-19 pandemic broke out in 2020. More patients are using online reservations, but many do not know which department they need to register for. To help patients, Xinhua Hospital is using AI.

Tencent unveiled its progress in chips for the first time this week, which immediately gave its stock price a modest boost. Silicon seems distant from the giant's main arenas of video games and social networks, so observers suggest that Tencent's move is to signal that it's aligned with China's long-run goals to self-develop semiconductors at a time its gaming unit is under a slate of regulatory assaults. Other major tech powerhouses, from Alibaba, Baidu, to Huawei, have all answered Beijing's silicon push with their in-house chips. On the other hand, for a company with as much data to handle as Tencent, one should wonder why it didn't get in on semiconductors earlier. The three chips Tencent unveiled on Wednesday are all custom built: one for AI inference, one for video transcoding, and another for network interface.

The Pentagon's first-ever chief software officer abruptly quit earlier this month, and now we know exactly why: Nicolas Chaillan, former CSO of the United States Air Force and Space Force, told the Financial Times that the United States has "no competing fighting chance against China in 15 to 20 years" when it comes to cyberwarfare and artificial intelligence. Chaillan, a 37-year-old tech entrepreneur, added that cyber defenses at many government agencies are at "kindergarten level," and that companies like Google were doing the US a disservice by not working with the military more on AI, since Chinese companies were making a "massive investment" in AI without getting all hung up on the ethics of it all. And while quitting your job because America has already lost the AI race is a bit dramatic, Chaillan isn't the only one who's concerned about China's dominance in this arena. A growing number of leaders in Washington and Silicon Valley are worried about the US falling behind in the race to AI supremacy. Congressional hearings on the future of AI have been going on since 2016, and Chaillan said he plans to testify in some upcoming ones.