
Baidu APP held a product upgrade communication meeting in Beijing and launched many major upgrades such as search recursive self-evolution, AI API, AI lens, and GUI visualization.
AI-generated summary
Baidu APP held a product upgrade communication meeting in Beijing to showcase its latest progress in AI search and multi-modal interaction.
On October 30, Baidu APP held a product upgrade communication meeting in Beijing. Baidu Search realizes recursive self-evolution and completely upgrades its search AI API. During the event, the lowest price is only 4.9 yuan/thousand times. At the same time, major upgrades have been launched around the search box, AI lens, GUI visualization, and AI interactive applications to bring users a smarter, smarter, and more fun experience.
Search is moving from “one question, one answer” to “one task, one delivery”. Complex problems can be handed over to search for multi-step search and sorting, and the conclusions after verification and summary can be returned directly. High-frequency transactions can directly deliver results, compressing "search-jump-operation" into the shortest path; matters that require long-term attention will be continuously tracked after the user confirms, and proactive reminders will be given when changes are found.
Currently, this capability covers more than 170 tool scenarios. When parents search for information about further education, the search will combine previously recorded children's grades, home addresses and boarding preferences, compare commutes, tuition fees, and admissions policies to give more specific school selection suggestions. When users plan their trip to Sanya during the Spring Festival, the search not only generates a complete five-day itinerary and budget, but also proactively reminds them when hotel prices are reduced, saving 2,100 yuan on a five-night trip.
In terms of AI lenses, Baidu aims at the three major breakpoints of traditional photography of "not being able to fully understand, not being able to explain clearly, and not being able to ask questions deeply", so that the lens can move from "fixed" to "flowing". The upgraded lens supports simultaneous recognition and dynamic tracking of multiple targets. Various plants and musical instruments in the picture can be identified at once. Jumping monkeys, swimming fish schools, and identification tags can all be followed steadily. Users no longer need to struggle to describe the target. They can accurately ask questions by clicking or circling. After taking the photo, they can seamlessly enter a real-time conversation and follow their curiosity to ask questions step by step, moving from "knowing what it is" to "knowing why it is so." At the same time, the AI lens supports real-time dialogue in 9 dialects including Northeastern, Sichuan, Cantonese, and Changsha dialect, making "talking to the world through the lens" a reality.
In terms of GUI, the GUI scene coverage of the search page has been increased by 5 times, accessing more than 5 billion authoritative resource data such as government affairs, people's livelihood, and central banks, covering more than 20 industries. Relying on AI intelligent visualization and professional chart templates, complex data can be "visible, comparable, and interactive", and the ecosystem is open, and authoritative professional data institutions and partners can join.
In terms of AI interactive applications, Baidu APP has released AI interactive dramas and AI interactive applications. Users have changed from "watching and watching" to "playing". They can choose the plot direction in the interactive drama, or they can operate and get feedback in the interactive application. The content form has changed from one-way consumption to participation in co-creation. The "Baidu AI Future Creation League" launched at the same time offers an olive branch to interactive content creators with millions of cash incentives and billions of traffic support.
It is worth noting that Baidu has also newly upgraded its search AI API this time, with a limited-time lowest price of 4.9 yuan per 1,000 queries.
The high cost performance and good results are due to its AI native four-layer architecture: the content layer brings together more than 80 languages in the world, hundreds of billions of high-quality indexes, and the data credibility rate is increased by more than 50%; the engine layer realizes multi-Agent collaborative self-evolution through an online feedback closed loop. Baidu Search precipitates every real retrieval and user feedback into training signals, and goes through "retrieval-feedback-retraining" "The closed-loop continuous reflow drives the model to recursively self-evolve, and iteratively upgrades continuously without manual intervention, and the overall iteration efficiency increases by more than 80%; the architecture layer supports high concurrency and low latency, and has high-quality enterprise-level service guarantee capabilities; the application layer supports API, MCP, and Skill full solution access, and the effect is increased by 20% after the Wenxin Assistant is accessed.
Baidu APP redefines the competition dimension of AI applications with "accurate search, good answers, and good delivery" to bring users a smarter, smarter, and more fun experience.
AI outlook — possibilities, not facts
Baidu AI Future Creation League will attract interactive content creators to participate
Likely · Within months

The US Federal Communications Commission has proposed a ban preventing Chinese telecommunications firms from acquiring American phone numbers, citing national security concerns and the need to curb malicious activities such as illegal robocalls.

Japan's All Nippon Airways (ANA) announced that due to an illegal intrusion into the system of its outsourcing manufacturer Amadeus, the personal information of approximately 37,800 customers was leaked, of which approximately 1,700 involved sensitive information such as passport numbers. Those affected are passengers who booked tickets between August 2023 and September 2024.

The v9 TPU built by MediaTek for Google uses Intel's EMIB-T process and is expected to be mass-produced in the first quarter of 2028, but it faces the challenge of low yield. Foreign investors are worried about being constrained by Intel in the future. Vice Chairman and CEO Tsai Lixing emphasized that MediaTek does not rely on a single supplier. TSMC has always been the most important partner and has high confidence in yield improvement.

Japan has recently experienced a large-scale cyber attack, with a total of 260 million personal data leaked by companies since April this year. Muji, ANA, JAL and the Ministry of Foreign Affairs have all reported personal data leaks or system intrusions. Japanese Chief Cabinet Secretary Minoru Kihara said that he will propose emergency information security countermeasures next week and introduce high-performance AI technology made in the United States to strengthen defense.

During the Wuzhen Summit of the World Internet Conference from November 2 to 5, Tongxiang, Zhejiang Province will showcase to the world the innovative results of the "Wuzhen Seven" composed of seven local artificial intelligence companies, covering underlying computing power, simulation data, embodied robots and intelligent vehicle driving, marking the local area's acceleration in building a closed-loop artificial intelligence industry.

Cathay Pacific Airways reports that personal data and flight details for 53,000 bookings were compromised after a breach at technology provider Amadeus IT Group.