Upgrade your account to a Community Supporter account and remove most of the site ads.

Reply to thread

Message: <blockquote data-quote="jian" data-source="post: 9698864" data-attributes="member: 78087">Hmmm… the use case for disabled students is interesting, I’ll have to think about that. There are of course all sorts of entirely valid use cases for machine learning, such as distribution logistics or analysing vast amounts of scientific data (such as radio telescope pictures of distant galaxies, according to a talk I went to last month).There are lots of problems with generative AI, one of which is that it’s very much a solution looking for a problem, thanks to private capital and corporations having invested so much money into developing it, which is why there’s so much genAI in everything we’re offered now from Duolingo to Google. Does any of it help? Not really so far, that I can tell.In my specific sector (medicine, specifically primary care, but also epidemiology) genAI is of very little use so far. The limited use case is genAI scribing for doctors who have to write notes not in their first language (a common issue for international medical graduates) but those functions still hallucinate a lot and a doctor who is not writing in their first language is, not surprisingly, less likely to catch the mistakes. So not really a viable solution, sadly.GenAI has so far not been any use for decision support (“do you think this might be cancer?” and so on) which is what people keep hoping for. So it’s not going to replace clinicians in diagnosis (let alone treatment) any time soon.The NHS is talking about rolling out a machine learning triage tool for the latest version of its app - basically Dr Google - and if Dr Google (a much better developed system) is anything to go by it will just increase unnecessary consultations and anxiety. Which is why health secretaries should not take policy advice from investment-based think tanks.</blockquote>

Verification