CCTV News: On the morning of March 27, the 2025 Zhongguancun Forum Annual Meeting opened in Beijing. In 2025, the forum annual meeting used cutting-edge technologies such as embodied intelligent robots, AI models, and intelligent displays to create eight major categories of technology conference hosting scenarios such as intelligent sensorless participation, robot integration and collaboration, and conference smart management.
On the first floor of Zhongguancun International Innovation Center, there is a "intelligence" volunteer service team composed of nearly 100 robots. This "intelligence" is the wisdom of "wisdom". For example, this Kuafu robot undertakes the leading service function of the 2025 Zhongguancun Forum Annual Meeting.
In addition to providing welcome, guidance, inquiries and other services to the guests, relying on big model technology, robot "intelligent" volunteers from different companies can also form team cooperation.
In the robot cafe, the robot barista is busy making coffee. Once the coffee is ready, the universal humanoid robot will deliver the coffee to the guests.
In addition, these robots "wise" wishes can also run, dance, write poems and draw, which is versatile. Introduce a simultaneous model to improve real-time and translation quality
From the cutting-edge technology display outside the venue to the application of innovative technology in the venue, the 2025 Zhongguancun Forum Annual Meeting demonstrates the concept of science and technology hosting everywhere. In 2025, multiple venues introduced simultaneous large models. What kind of experience will this bring to the participants?
The simultaneous large model was cited for the first time at the 2025 Zhongguancun Forum Annual Meeting. It can synchronize the content of Chinese speeches into English in real time and display them on the large screen. At the same time, it can also replicate the tone, making it possible to give speeches to guests on site in different languages like the same person.
The simultaneous model used in the venue adopts an end-to-end architecture, which is very different from the previous traditional simultaneous interpretation software. Traditional software needs to convert the language into text first, then translate the text, and finally output the voice, while the simultaneous large model improves real-time and translation quality.
