LecturaInvestigaciónRadarMarco de inversión
Iniciar sesión / Registrarse
← Volver a la lectura
NoticiasTema · 图形学与机器人的联系Material3 min de lecturaEditorial de DisconfirmAI

Meta Researcher 남길주: Image or Video Output May Not Require 3D Modeling; Robotics Still Needs Spatial Understanding

According to ZDNet Korea, Meta Reality Labs researcher 남길주 argued at AI Tech Day 2026 at Korea University that producing an image or video does not necessarily require building a complex 3D model. Drawing on roughly five years of his own 3D hair-reconstruction research, he discussed possible directions for 3D graphics expertise in robotics and autonomous driving, where spatial and physical reasoning remain necessary.

2026-10-20 11:58 GMT+8

According to ZDNet Korea, Meta Reality Labs researcher 남길주 sees generative AI as changing the role of 3D modeling in video production. In a talk at AI Tech Day 2026 held at Korea University, he offered a conditional judgment: when the final output is an image or video, building a 3D model is not necessarily required, because generative AI can produce visuals without individually performing the traditional modeling steps. He used this to discuss new directions for 3D graphics in robotics, autonomous driving and other physical-world tasks; the report does not establish this across every film and game production workflow.

Traditional computer graphics works by first constructing the 3D shape of characters and objects, then computing movement and light reflections to render the scene. Generative AI skips these steps — it learns from vast amounts of image and video data and produces new visuals directly from text descriptions. Early AI video mainly generated faces; later it learned to handle complex backgrounds and light reflections. OpenAI's Sora can already realistically reproduce human gait and surrounding environments, and recently techniques have emerged that control a character's facial expressions and body movements.

The report also recounts 남길주's own research experience. He has been researching computer vision and 3D graphics at Meta Reality Labs since 2020, and spent roughly five years working on 3D hair reconstruction. He eventually changed his approach: instead of hand-designing complex graphics pipelines, he applied only basic knowledge and let AI learn the rest from data. He described the result as better than the original method.

Iniciar sesión / Registrarse
LecturaInvestigaciónRadarMarco de inversión

He connected this shift to what AI researcher Richard Sutton calls The Bitter Lesson — in the long term, general methods that leverage large-scale computation to let AI learn on its own have achieved higher results than approaches where humans meticulously design rules. He treated the hair-reconstruction pivot as one concrete instance of this approach.

But 남길주 also drew a clear boundary: robotics and autonomous driving are different from video production. Naturally rendering a person's movement on a screen and having a robot accurately grasp an object are two separate problems. For a robot to complete a grasp, it must understand the object's position, shape and surrounding space; an autonomous vehicle needs to identify obstacles and the positions and movements of other vehicles. These tasks still require technologies that handle 3D space and physical phenomena.

So his conclusion is not that 3D graphics is obsolete, but that its battlefield has moved. Manufacturing 3D design, physics simulation, optics and scientific research also require graphics technology. 남길주 said that using graphics to digitize the world and contribute to the development of physical AI, robotics and autonomous driving is the new vision.

남길주 emphasized that these tasks still require technologies handling 3D space and physical phenomena, which is precisely why graphics expertise retains value in robotics and autonomous driving.

Fuentes

  • zdnet.co.kr↗
Tema

图形学与机器人的联系

图形学与机器人的联系,可以从研究员观点、会议议程和综述讨论的具体用途来观察;这些材料尚不能证明整个行业的重心已从影视游戏迁往机器人。

3 pasos1 piezasActualizado 2026-10-21 11:55

Cómo llegó hasta aquí

  1. 2025-11-28Bianews所载访谈中,Google DeepMind研究员谭捷以“机器人是在真实世界里做图形学”作类比,并分享强化学习与Sim-to-Real研究如何影响机器人发展的个人判断与经历。bianews.com ↗
  2. 2026-07-01SIGGRAPH 2026发布新闻稿,介绍大会中图形学与机器人交叉的课程和论文,包括NVIDIA主导的物理AI系统课程及利用仿真训练机器人的研究方向;这说明议程安排,不代表已证实行业重心迁移。s2026.siggraph.org ↗
  3. 2026-10-20据ZDNet Korea报道,Meta Reality Labs研究员남길주在韩国高丽大学AI Tech Day演讲中表示,最终产物为图像或视频时不一定需要先建复杂3D模型,并讨论机器人与自动驾驶仍需空间和物理理解。zdnet.co.kr ↗

Lo que aún está pendiente

  • 남길주关于图像或视频输出可不先建模的判断,适用于哪些具体制作流程,哪些场景仍需显式空间或物理约束?
  • 谭捷所述精细操作与跨任务泛化之间的差距,能否在明确任务条件和评价方法后进一步核验?
  • 会议议程和综述中的研究方向,会怎样反映到实际项目与技能需求?目前还缺哪些可核验的行业证据?
Leer el tema completo →
Continuar investigando →