Core Insights - 1X has launched the new 1X World Model, a groundbreaking AI update tailored for its humanoid robot NEO, marking a significant advancement in the field of robotics [2][3] - The 1X World Model enables NEO to instantly convert any instruction into executable AI capabilities based on video models that adhere to real-world physics, paving the way for robots to possess self-learning abilities [2][3] Group 1: Technological Advancements - The 1X World Model allows NEO to learn from vast internet-scale video data and apply this knowledge directly to the physical world, enabling it to execute tasks it has never encountered before [3][4] - Users can issue simple voice or text commands to NEO, which will generate visual predictions of subsequent actions and convert these predictions into precise, executable movements using an internal inverse dynamics model [3][4] - NEO demonstrates the ability to generalize beyond existing training data, successfully completing new tasks such as lifting a toilet seat or ironing clothes, showcasing the transfer of human knowledge to machines [3][5] Group 2: Autonomous Learning and Adaptability - The 1X World Model represents a paradigm shift, allowing robots to autonomously gather data and master new skills, thus accelerating the development of general-purpose humanoid robots [4][5] - Unlike traditional AI models that rely on manually collected data, the 1X World Model enables NEO to self-optimize and benefit from the continuous evolution of video models [6][7] - NEO can maintain stable performance in dynamic and unpredictable environments, effectively handling changes in lighting, clutter, or disorder, which is a significant breakthrough in humanoid robotics [7] Group 3: Product Availability and Pricing - NEO is available for purchase on the 1X online store, offered in three colors: brown, gray, and dark brown, with an early bird price of $20,000 for priority shipping in 2026 [9] - Customers also have the option to subscribe to NEO for a monthly fee of $499 [9] Group 4: Company Overview - 1X is a leading AI and robotics company in the United States, focused on developing the home robot NEO, with a mission to enhance people's lives through safe and intelligent humanoid robots [10]

RT Theoretically Media (@TheoMediaAI)The Official Release has finally dropped! Veo-3.1 is here!After a week of biting my tongue at the "leaked specs" we can finally start putting this update through it's paces!3.1's video model enhancements are one thing, but the real sauce here is in the new features!(more) https://t.co/Yo8Ke3fKA4 ...

闻乐 发自 凹非寺 量子位 | 公众号 QbitAI 为什么语言模型能从预测下一个词中学到很多，而视频模型却从预测下一帧中学到很少？ 这是UC伯克利大学计算机副教授 Sergey Levine 最新提出的灵魂一问。 他同时是Google Brain的研究员，参与了Google知名机器人大模型PALM-E、RT1和RT2等项目。 Sergey Levine在谷歌学术的被引用次数高达18万次。 "柏拉图洞穴"是一个很古老的哲学比喻，通常被用来说明人们对世界认知的局限性。 Sergey Levine的这篇文章以《柏拉图洞穴中的语言模型》为题，又想要揭示AI的哪些缺陷呢？ 在文章的开头，作者提到人工智能就是在研究能够反映人类智能的灵活性和适应性的假想智能。 一些研究者推测，人类心智的复杂性和灵活性源自于大脑中应用的一个 单一算法 ，通过这个算法可以实现所有多样化的能力。 也就是说，AI如果能复现这个终极算法，人工智能就能通过经验自主获取多元能力，达到人类智能的高度。 在这个探索过程中，语言模型取得了非常成功的突破。 甚至，LLMs实现能力跃升背后的算法（ 下一词预测+强化学习微调 ），也非常简单。 单一终极算法 假设 ...