我孩子訪校時看有理科教授拿machine learning text book，

It will take perhaps 1-2 years to build up a model to explain certain experiment in traditional way, with machine learning could shorten it to 1-2 month. My concern: could this be like curve fitting with infinite # of parameters. 

The reality is if the machinine-learning-built model can predict other things which could be verified later or explain other experiements too, why should anyone care too much how it is labeled. 

 

所有跟帖： 

• 很可能ML代表了人類裏大多人的認知過程。但能不能表示他也擁有最具創造力的人群的智能還有待觀察。 -whaled- ♂ (0 bytes) () 05/26/2023 postreply 08:36:32

• 解決一些大分子之類的問題應該比較適合吧。 -borisg- ♂ (0 bytes) () 05/26/2023 postreply 08:38:56