AI 热点

通用机器人公司 Genesis AI 获 5 亿美元融资,估值 35 亿美元

SiliconANGLE更新于 11月18日 10:06

General-purpose robot maker Genesis AI reels in $500M at $3.5B valuation - SiliconANGLE

AI 导读

巴黎通用机器人公司 Genesis AI 宣布完成 5 亿美元 A 轮融资,估值 35 亿美元,由 Premji Invest、Khosla Ventures 和 Eclipse 领投,NVIDIA、AMD、Zoom、三星、爱立信的创投部门及 Eric Schmidt 参投。其机器人 Eno 采用轮式底盘而非双腿,手掌和手指按人手设计,有 20 个自由度,可借助人类手部操作演示数据训练,由自研 GENE 模型驱动,演示中能弹钢琴、做饭和还原魔方。资金将用于招聘 AI 研究员和机器人专家,并扩大训练数据收集。

推荐理由

AI 评分 59

融资细节之外,原文还交代了 Eno 用轮式底盘和仿人手设计绕开机器人训练数据难收集的问题。

主题

标签

正文

Genesis AI SAS today announced that it has raised $500 million in funding from a group of prominent investors.

Premji Invest, Khosla Ventures and Eclipse led the Series A round. They were joined by the venture capital arms of Nvidia Corp., Advanced Micro Devices Inc., Zoom Communications Inc., Samsung Electronics Co. and Ericsson. Former Google Chief Executive Eric Schmidt contributed as well along with other tech executives.

Paris-based Genesis develops a general-purpose robot called Eno (pictured). Unlike humanoid robots, it moves using a wheeled base rather than legs. The wheeled base can adjust Eno’s height using a mechanism that consists of two large panels connected by a hinge. The module enables the robot to adapt its configuration to each task.

Eno is powered by a custom artificial intelligence model called GENE. Genesis says that the algorithm is easier to train than traditional robotics-optimized neural networks, which can translate to higher output quality.

The dataset on which an AI model is trained must be tailored to the robot it’s intended to control. If a company is building an AI model for a four-legged robot, the training dataset must include footage of four-legged robots. The more specialized the robot, the more difficult it is to collect the necessary training data.

Genesis has come up with a workaround to that challenge. Eno’s palms and fingers are based on a custom design closely modeled after the human hand. As a result, the robot can be trained on demonstrations in which humans perform tasks with their hands. That kind of training data is much easier to produce than footage of robots performing work.

Eno’s robotic hands have 20 degrees of freedom, which means they can perform that many motions. The human hand has 27 degrees of freedom. Additionally, Genesis covered the hands with a soft material designed to mimic skin.

The GENE foundation model that powers Emo can perform tasks comprising dozens of steps. Genesis has released demo videos that show the system playing the piano, cooking and solving a Rubik’s cube. According the company, that versatility enables Eno to operate in not only industrial environments but also homes.

“We’re working to build the world’s best general-purpose robot,” said Genesis co-founder and CEO Zhou Xian. “By building it all ourselves, we can obsess over every detail and improve the entire robot faster in one closed loop. This funding will accelerate our efforts to deploy our robots in the real world at scale.”

The company will use the capital to hire more AI researchers and roboticists. Additionally, Genesis will step up its training data collection efforts in a bid to expand its AI model’s capabilities.

Photo: Genesis AI

来源

SiliconANGLE