𝗙𝘂𝘁𝘂𝗿𝗲𝗫 𝗣𝗵𝘆𝘀𝗶𝗰𝗮𝗹 𝗔𝗜 𝗗𝗮𝗶𝗹𝘆: 𝗜𝘀𝘀𝘂𝗲 𝟮𝟵
Physical AI is moving fast. Capital is flowing into new areas, and researchers are solving hard problems.
Here are the top updates from today:
𝗠𝗮𝗿𝗶𝗻𝗲 𝗥𝗼𝗯𝗼𝘁𝗶𝗰𝘀 𝗚𝗲𝘁 𝗕𝗶𝗴 𝗖𝗮𝗽𝗶𝘁𝗮𝗹 Shihang Intelligent raised over 1 billion RMB in an A-round. This is a global record for marine robotics funding. Investors like Temasek are backing this move into underwater intelligence.
𝗪𝗼𝗿𝗹𝗱 𝗠𝗼𝗱𝗲𝗹𝘀 𝗔𝘁𝘁𝗿𝗮𝗰𝘁 𝗠𝗶𝗹𝗹𝗶𝗼𝗻𝘀 Companies like Daxiao Robotics and GigaAI are securing massive funding. The focus is shifting from simple video models to world models that understand physical interactions.
𝗛𝘂𝗺𝗮𝗻𝗼𝗶𝗱 𝗥𝗼𝗯𝗼𝘁𝘀 𝗘𝗻𝘁𝗲𝗿 𝗙𝗮𝗰𝘁𝗼𝗿𝗶𝗲𝘀 Humanoid robots are no longer just lab projects. • SERES debuted "Xiao Sai" at a super factory. • Songyan Dynamics released an open-source humanoid using HarmonyOS. • Huawei is integrating humanoids into its ecosystem.
𝗞𝗲𝘆 𝗥𝗲𝘀𝗲𝗮𝗿𝗰𝗵 𝗕𝗿𝗲𝗮𝗸𝘁𝗵𝗿𝗼𝘂𝗴𝗵𝘀 Researchers are finding better ways to make robots move and think: • μ₀: A model that predicts 3D trajectories instead of dense pixels. This makes training faster. • Tencent Robotics X: They open-sourced a full-stack system from data collection to real-robot deployment. • EQRL: A method that lets robots spend more compute on hard tasks and less on easy ones. This cuts inference costs by 32%. • DiPOD: A fix to prevent instability during robot learning.
𝗜𝗻𝗱𝘂𝘀𝘁𝗿𝘆 𝗧𝗿𝗲𝗻𝗱𝘀 • Hardware scaling: Companies are racing to mass-produce dexterous hands. • Leasing models: JD.com is expanding robot leasing to lower entry costs for businesses. • Data is the new gold: Companies like Noitom are focusing on data infrastructure rather than building hardware.
The gap between AI code and physical action is closing.
Source: https://dev.to/future_x/futurex-physical-ai-daily-issue-29-0616-2d23
Optional learning community: https://t.me/GyaanSetuAi