X @Balaji

AI Market Cycle & Investment - AI is entering the trough of the hype cycle, presenting an opportune time for investment and building as the space thins out [1] - The industry should focus on the unexpected plateau of AI, which has currently topped out at "far better web search and summary" [3] Limitations of Digital AI - LLM-style digital AI may primarily be repeating rather than truly thinking, limiting its ability to generate novel thoughts [4] - Digital AI operates middle-to-middle due to bottlenecks in prompting and verifying [5] - Online data used for AI training lacks digital signatures, leading to internal inconsistencies and fake data, especially with widespread AI models [9][13] Potential of Physical Specific Intelligence - Physical AI, particularly in robotic guidance, can achieve end-to-end solutions for well-specified physical world problems [6] - Significant AI progress is expected in economically valuable physical world problems like self-driving cars, termed "physical specific intelligence" [7] - Chinese robots are poised to excel in various tasks and form factors, with sidewalk robots and delivery drones already in use [5] - N robots sensing the physical world converge on a consensus reality through SLAM or equivalent, learning from the same physical world [10][12] Future Applications - The industrial application of physically specific intelligence lies in factory robots [15] - The consumer version envisions a garden of intelligent things like robot cars, dogs, and drones that can communicate and understand users [15]