【Generative Physical AI】NVIDIA
https://youtu.be/AYSfcgVv9-U

【Generative Physical AI】NVIDIA


NVIDIA Generative Physical AI needs three types of computers:

  1. Nvidia AI supercomputers: to train AI models.
  2. Nvidia Jetson Orin and Next Generation Jetson Thor robotic supercomputer: specialized computers designed to run the AI models.
  3. Nvidia Omniverse: a virtual world where robots can practice and get better at tasks.


Humanoid Robots with LLM Integration

【USA】technology Figure 、Agility Robotics

【China】scalibility & gov. policy UBTECH Robotics Unitree

【Taiwan】supply chain (key components、robot arm/body、 system integration) Techman Robot


Why Humanoid Robots?

RethinkX

<This time, we are the horses: the disruption of labor by humanoid robots>

"For the purposes of illustration, consider a humanoid robot with a total lifetime cost of $200,000 that works 20,000 hours before decommissioning: its labor would cost $10 per hour. Even at this relatively high cost point, humanoid robots are already competitive with human labor in a substantial fraction of the global economy. In reality, lifetime costs of humanoid robots are likely to be far less than $200,000 right from the start."

"Suitability to existing environs : All existing facilities, equipment, and infrastructure is designed around the human form. That makes the humanoid form a natural choice for mass-produced robots in the near term...."


Robin Hanson

<Population fall won't last long>

"Many papers that make this point. For example, a 2019 Evolution and Human Behavior paper is titled “The heritability of fertility makes world population stabilization unlikely in the foreseeable future.”? It assumes that world fertility soon stops declining on average at a near replacement level, except that it has a heritability of 0.3 in 30yr generations. It finds that world fertility starts to rise again ~2060, and by 2100 rises to 2.21 instead of 1.82 without heritability.?"


【Prompt】You are an economics professor. Please share your prediction based on the above 2 big trends:

- Population fall won't last long

- a humanoid robot cost down estimation: cost $10 per hour

【ChatGPT】(....) "Investment in Technology and Infrastructure: There will likely be increased investment in technology development, particularly in robotics and AI, as well as in infrastructure to support a growing population and the integration of new technologies." (read more)


Notes

  • Price (cost) of NVIDIA Generative Physical AI

  1. Nvidia AI supercomputers

DGX Cloud: $36,999 per instance for a month.

DGX Station: $49,900, which is a discount of over 25% off the list price.

DGX GH200: no specific pricing mentioned.

2. Nvidia Jetson Orin and Next Generation Jetson Thor robotic supercomputer

The prices for these products are not explicitly mentioned in the provided sources (1, 2) . However, NVIDIA Jetson Orin is a single-board AI computer designed for edge AI and robotics, and Jetson Thor is a high-performance AI computer for robotics and autonomous systems. The prices for these products vary depending on the specific models and configurations

3. Nvidia Omniverse

The price for NVIDIA Omniverse Enterprise subscriptions varies (1):

Omniverse Enterprise Nucleus Subscription: $1,000 per named user per year

Omniverse Enterprise Creator Subscription: $2,000 per concurrent user (CCU) per year

Omniverse Enterprise Reviewer Subscription: $100 per concurrent user (CCU) per year

For example, a small design firm with 5 3D designers and 10 reviewers would cost $21,000 per year for an annual Omniverse Enterprise subscription. A larger deployment with 100 designers, 150 reviewers, and 30 concurrent creators would cost $135,000 per year.

Running Omniverse on a GPU instance is also priced at $4,500 per GPU per year (2)

  • 「從近期發展狀況來看,美國和中國是人形機器人發展的兩大主要國家。美國技術研發方面領先,中國則在市場規模政策支持方面具有優勢。」 2024-May-16 DIGITALTIMES * 林欣姿
  • China-工業和信息化部-《人形機器人創新發展指導意見
  • 台灣在全球機器產業供應鏈中潛力雄厚,產業鏈可區分為三大類,分別是「關鍵零組件」、「機器人主體」以及「系統整合」三方面。機器人主體上,又區分為「工業型」與「服務型」機器人。第三類的系統整合部分,本就是台灣電子產業優勢,藉由系統整合提升生產線效率、降低營運成本,更可以達到多元化的產業商機,目前投入機器人系統整合的企業集團包含「和椿科技」、「盟立自動化」、「均豪精密工業」等,在各自擅長的領域中開發新商機。」。2024-Apr-19 卓越電子報 * 李振麟
  • <Semiconductor legends take a stroll in a Taiwanese night market — Nvidia, TSMC, MediaTek, and Quanta heads seen eating dinner> 2024-June-01 Toms Hardware * Jowi Morales

Barry Lam (Quanta), Jensen Huang (Nvidia), Morris Chang (TSMC)


Relevant articles

【Liquid AI】MIT - Daniela Rus

【Spatial AI】Stanford- Fei Fei Li

【AI】 What to build? (Use LLM to filter options)

【Generative Physical AI】NVIDIA humanoid robots

要查看或添加评论,请登录

Katherine Shih的更多文章

社区洞察

其他会员也浏览了