RT @yudapearl@twitter.com

For this reason, GPT's understanding should be tested on questions whose answer is "NO, it is not possible under any circumstances". Such questions are the core of causal logic. e.g., can a drug be good for men, good for women and bad for people? https://twitter.com/yudapearl/status/1640075242470277120

🐦🔗: https://twitter.com/yudapearl/status/1640079735282864131