tag:blogger.com,1999:blog-9633767.post7380142875315396581..comments2024-03-25T09:11:17.877-07:00Comments on The Curious Wavefunction: Lab automation using machine learning? Hold on to your pipettes for now.Wavefunctionhttp://www.blogger.com/profile/14993805391653267639noreply@blogger.comBlogger2125tag:blogger.com,1999:blog-9633767.post-49184883975143182312017-07-10T20:18:23.235-07:002017-07-10T20:18:23.235-07:00As researchers, we also acquire a tremendous amoun...As researchers, we also acquire a tremendous amount of "prior knowledge" (all those years of learning!) that helps us to propose the best hypotheses and design the best experiments. ML/AI does not have that prior knowledge, unless we encode and input it. <br /><br />Perhaps, if we do supply sufficient prior knowledge to AI/ML, it can be used to distill what we know into all best-possible hypotheses and corresponding experiments. The number of experiments will likely be staggering, even for "high throughout" robotic assays (how many atoms in the Universe, you ask?), and it's up to us to further filter and select the most promising ones. <br /><br />With that philosophy, AI/ML is another valuable tool to formalize the differences between what we know and what is unknown/possible.Anonymoushttps://www.blogger.com/profile/00424300940467817307noreply@blogger.comtag:blogger.com,1999:blog-9633767.post-72196914549408886812017-07-07T07:29:42.684-07:002017-07-07T07:29:42.684-07:00Sometimes answers arise out of the errors we commi...Sometimes answers arise out of the errors we commit while conducting an experiment or in the design. Something goes wrong and when trying to figure that out, we get an answer. AI may not be able to that kind of thinking since every mistake is unique and one cannot program for it.Lhttps://www.blogger.com/profile/03186014514110985045noreply@blogger.com