At the "2025 Tech Landscape Summit" annual ceremony held in Beijing, Ping Xiaoli, Vice President of Baidu Group and head of Baidu's Digital Human & E-commerce business, stated that digital humans have undergone several generations of development and are now at the threshold of "surpassing real humans." From the initial "obviously fake" stage of 1.0, to the goal of "surpassing real humans" in the 4.0 phase, this evolution is underpinned by continuous breakthroughs in AI technology, with visual models, large language models, and agent-related technologies forming the cornerstone of digital human advancement.
She pointed out that digital humans are accelerating their evolution. The digital humans of the 1.0 era merely achieved a virtual human effect; while possessing a human image and voice, they typically had stiff expressions and heavily synthesized speech. The 2.0 era introduced hyper-realistic digital humans. With the advent of large models, high-precision cloning of human appearances became possible, supporting large movements and moving beyond a "cardboard cutout" effect. This era also enabled the generation of language scripts and interactive Q&A for digital humans, which is the current mainstream stage in the industry. Last year, Baidu took the lead in launching highly persuasive digital humans, ushering in the 3.0 phase for AI digital humans. These digital humans not only exhibit highly coordinated form, spirit, voice, and appearance but also possess the ability to think, make decisions, and coordinate multiple agents to complete specified tasks.
Digital humans also hold immense potential to become a new form of interaction in the AI era. Ping Xiaoli mentioned that Baidu recently released the industry's first real-time interactive digital human, enabling digital humans to perceive and understand the physical world, interact naturally like real people, and even provide emotional value to humans.
Ping Xiaoli indicated that we are about to witness digital humans entering the 4.0 stage. These next-generation digital humans will not only possess world models and world knowledge, enabling continuous autonomous evolution, but will also support personalized emotional interactions tailored to each individual. Like tireless digital perpetual motion machines, they are expected to deliver greater productivity and surpass real humans in more application scenarios.
Comments