A Turing Test for Physical Intelligence

As I was trying to shove my protesting cat into his cat carrier, I thought of a great Turing Test for physical intelligence. The man himself. He actually loves sitting in his cat carrier… unless you show any sign of wanting to put him in by force. Then, not so muchRecent progress in artificial intelligence has been nothing short of astounding. At least, in silico. Real-world impact lags the perceived intelligence of modern LLMs by quite a bit, and the explanation is fairly obvious: optimizers are very well suited to deterministic, parallelizable, and verifiable environments, less so to path-dependent, single-threaded, fuzzy environments. The AI cannot, yet, do a job end-to-end. Reality, unfortunately, has a lot of detail, and much more closely resembles the latter.Thanks for reading Maximum Effort, Minimum Reward! Subscribe for free to receive new posts and support my work.I should point out that this isn’t a particularly special property of AIs, humans are also likely to fare much,…