Xiaopeng Motors' Chairman He Xiaopeng on Robotaxi and End-to-End Technology

TapTechNews July 11th news, Xiaopeng Motors' chairman He Xiaopeng stated that the current capabilities, regulations, and business closures of Robotaxi cannot meet expectations, but it is currently in an accelerated development stage. He predicted that Robotaxi will begin to enter a turning point in about two years.

In addition, He Xiaopeng also said that currently many companies researching and developing L4 autonomous driving technology still adopt the traditional algorithm + small AI model technical route. He suggested that these enterprises quickly switch to the end-to-end technology.

TapTechNews note: End-to-end learning, also known as end-to-end training, refers to not conducting modular or phased training during the learning process, directly optimizing the overall goal of the task, emphasizing the entire process from the user's input to the result output, and there is no need for human intervention in the middle process.

Referring to TapTechNews' previous reports, the Xiaopeng Infinite XNGP function that can be driven across the country has already started public beta with the Tianji XOS 5.2.0 Pioneer 1 version. The official claimed that this is the first end-to-end AI large model mass-produced and installed in vehicles in the country, and Xiaopeng is one of the only two global automakers to implement end-to-end technology.

In the release conference in May this year, He Xiaopeng said that the goal of its end-to-end large model is to iterate once every two days, and the intelligent driving ability will be enhanced by 30 times in 18 months; by the third quarter of 2024, it will be able to be driven on every street in the country.

Related readings:

The Xiaopeng XNGP that can be driven across the country goes online: XOS 5.2 starts the pioneer public beta and adopts the blacklist mode

The end-to-end AI large model iterates once every two days, and the Xiaopeng XNGP opens 15 new cities in Shandong and Hebei

Xiaopeng Motors releases the end-to-end large model: The goal is to iterate once every two days and the intelligent driving ability is enhanced by 30 times in 18 months