Can AI in Healthcare Be Trusted? | WSJ Tech News Briefing

foreign [Music] this is your Tech news briefing for Monday February 6th I'm Zoe Thomas for The Wall Street Journal as artificial intelligence Technologies become a bigger part of our everyday lives a lot of people are asking how much can we trust it that question is one that wsj reporter Eric neeler and researcher and inventor ramachalapa try to answer in their book titled appropriately enough can we trust AI in a recent live q a Eric and Rama looked at one particular area where trust in AI plays a big role Healthcare they spoke with wsj live journalism news editor Sarah Castellanos about it here are highlights from their conversation so welcome to both of you it's so great to have you here so AI is being experimented with across so many sectors to you know varying degrees of success you guys mentioned a number of examples in your book what are some of the most Cutting Edge uses of AI and Healthcare specifically today Eric do you want to start sure um AI really feeds on large amounts of data and um Healthcare and Medicine generate a lot of data uh in the sense of screens of the body tumor biopsies photographs of skin lesions these sorts of things data from patient records in the hospital vital signs a real endless list of information that is collected about a patient's health so people who are able to combine Ai and and Medicine are right on this Cutting Edge of being able to take a computer algorithm with a large amount of data to look for patterns that otherwise just would be impossible or just take too long to figure out so that's what we're looking at right now and it really cuts across a lot of fields in medicine so tell us a little bit about the sort of the AI and um covid example here I mean Eric it seems like it was used but potentially AI was used but potentially not to Great Lengths or to great effect right right and so what was happening is at the time um uh doctors and computer scientists and data Engineers were working together is there a way to predict hospitalization for covid when a patient comes in if they show these sorts of Vital Signs these sorts of symptoms can we say that this person um you know either has kobit how they're gonna um how long it's gonna take for them to recover what kind of hospitalization rates should we prepare for a big surge of patients you know uh if we're seeing these sorts of things in our in the population and so there was a lot of attempts to to do this and um some of the later review studies that six months or a year later look back on that initial 2020 surge um found out that um it it really didn't do so well there was um there was a problem with um people coming in with what's called comorbidities or different different sort of um symptoms and health effects may maybe that it led to to a covet infection that kind of threw things off there were also different hospitals having different sorts of patient populations um you know large Urban medical centers rural clinics uh Suburban dock in the Box you know all these different places were getting uh coveted patients and so trying to pull all this data together and to have a big enough data set to train the algorithm it was really difficult and it was a very fast-moving epidemic as well a fast-moving virus and evolving virus so I think folks have learned a lot from that um and I think we'll see what happens you know in the future and how they're able to take these algorithms and maybe do a planning for planning purposes but but maybe not for uh this fast diagnosis what do you see are you know the potential challenges or you know know even ethical challenges here about GPT being potentially used in healthcare one day sure I mean aside from the issue of plagiarism in a medical license exam um I think there's a there is an issue of accuracy I think there's a lot of discussion and a lot of interest in commercial chat Bots for example Telehealth really gained a foothold during the pandemic and that seems to be a place where um the creators believe that this is going and also money will be flowing there will be products coming out and I think um the thing is we you know we want guard rails and and we don't quite have them yet right I think we need trustworthy chat GPD that's that's what the AI is the uh you know so what we are looking at because misinformation you know can can arise or you know incorrect information so we need to constantly watch it and make sure uh you know it doesn't do any any uh harm I think in terms of the bias we have to be careful uh certain communities being left out um a certain uh um patient populations minority populations that um either lack access to health care or are somehow have a history of discrimination for one reason or another and that um uh may not be included in medical studies we have one question about what are some other limits that exist for AI well I like to say AI doesn't like surprises especially the way it is designed Now using data so it can only handle what has been seen But there is a way out of that we are now using more and more synthetic data you know so what is possible and so that AI can be prepared or trained uh you know to handle things that have not happened but could you know for example you know sensors can only capture what has happened but imagination right you know it is much more varied and so forth so I think introducing more and more synthetic environments and so forth for training AI systems would be a good thing to do so yeah that's what I think is an exciting thing that might happen the other limit is autonomy um you know creativity is one side autonomy is another side where their fears of autonomy of something called General Ai and you know right now ai really just exists in in the data set where it lives or in the algorithm it lives it's not going to jump from you know one data set to another or you know infiltrate one set of computers to another it's it's not really going to happen right now and and I well we'll see what the future holds but autonomy is one thing that people are really concerned about and it just is not part of the discussion right now um we have a question about um whether you can speak to any specific Hospital systems or health care companies or even private practices who are leading the pack in terms of implementing AI effectively and ethically I think there are many examples right now so um I think there are small companies for example in the uh initiative we have from National Institute of Aging where I identified about eight companies you know from our first year Grant to promote you know AI research and and move products from labs to hospitals the AI and medicine is going to be slow because it need you know we need to go through and make sure it works and everything right so it's going to take time but I think as we see more and more FDA approvals and so forth you will see it's integrated it's not like you are going to have you know AI is working here in in a hospital sometimes it may just be in the background you know with a better way of triaging you know people who come into emergency and so forth we have time for one more audience question here there is a question about cyber security so are there cyber security or data security concerns around a using AI Tools in hospitals I'm sure there are many um Eric why don't you take that one why don't you take that one sure I think there's been some concerns uh in in recent years about internet connected devices um uh you know heart monitors Health monitors patient monitor sensors that that have an internet connection and and that can be hacked um I think you're probably going to see similar things uh as AI sort of like rhombuses you know becomes part of the background um I think uh anytime that you have large amounts of data that can be accessed um that you that the threats are there and and you just need to have the same security precautions that we have had as we've gone to digital Health in the past 15 20 years thank you both again really great conversation thanks for having me on the show great to be here thank you and that's it for today's Tech news briefing for more Tech stories check out our website wsj.com I'm Zoe Thomas for The Wall Street Journal thanks for listening [Music] foreign

Channel: WSJ News

Views: 72,718

Keywords: AI, healthcare, can ai be trusted, can we trust ai, ai healthcare, artificial intelligence healthcare, artificial intelligence help physicians to, ai technology, healthcare tech, healthcare technology, chat gpt, chat gpt healthcare, telehealth, telehealth technology, ai security, heart monitors, sensors, ai detector, wsj, tech news briefing, tech news, tech news wsj, ai, ai in health, ai in medicine, ai in health and medicine, artificial intelligence, medical diagnosis, techy

Id: -hTmsyPOJhw

Length: 10min 5sec (605 seconds)

Published: Tue Feb 07 2023