LogFAQs > #884491199

LurkerFAQs, Active DB, Database 1 ( 03.09.2017-09.16.2017 ), DB2, DB3, DB4, DB5, DB6, DB7, DB8, DB9, DB10, DB11, DB12, Clear

Topic List	Page List: 1
Topic	neat graph of jobs threatened by automation
Soviet_Poland 08/09/17 8:11:46 PM #50:	ChromaticAngel posted... Once again, disagree. AI software has lately been more accurate at diagnosis than real doctors. http://www.sciencemag.org/news/2017/04/self-taught-artificial-intelligence-beats-doctors-predicting-heart-attacks (second source) http://spectrum.ieee.org/the-human-os/biomedical/diagnostics/ai-predicts-heart-attacks-more-accurately-than-standard-doctor-method It's happening, whether you like it or not. You will be much better off if you fight for things now that will make life better when it finally happens then you would if you be 1) trying to fight against it, or 2) saying that it won't happen really loudly because it's happening. 1) The articles mentioned the role of AI machines as an adjunct (i.e. doctor's double checking the work), which is what I mentioned in my post. I didn't say the machines won't be used. I'm saying fully replacing doctors is a looong way away. 2) These studies are done in not very realistic practice scenarios. How do the studies account for patient noncompliance? Poor historianship? Like I said, these machines are sensitive in their stratification for things like CVD risk, identifying individuals who would otherwise not have been considered high risk, but now that decreases the specificity of your findings. Where is the longitudinal study or systematic review of a patient population placed on statins not previously identified as not high risk by AHA criteria from these machines compared to traditional management? Do these individuals have a higher risk for statin-induced myopathy? Rhabdomyelitis? Sure, you prevented the heart attack, but a larger proportion of people suffer kidney injury, but since the study in question was measuring cardiovascular related incidents, its no problem right? Diagnosis is one side of the coin. Does AI help improve outcomes though? Where is the study of machines seeing patients and patients actually taking the machine's recommendations? Compare them to human interaction. You can't force a patient to take a statin, so the machine might recognize who is at risk that otherwise wouldn't have been identified, but the human element of patient care is pretty difficult to overcome. Unless you're planning on shoving those pills down a patient's throat without permission lol. Sure, these machines can take things like patient noncompliance into account and through churning "Big data" find patterns a human mind wouldn't have picked up on. I don't doubt that would improve outcomes. But from the sounds of it, the machines will be used as an adjunct to a physician, not a complete replacement. --- "He has two neurons held together by a spirochete." <cite>Soviet_Poland posted [p:884491199] in neat graph... [t:75654644]...</cite> <quote><cite>ChromaticAngel posted...</cite> <quote>Once again, disagree. AI software has lately been more accurate at diagnosis than real doctors. http://www.sciencemag.org/news/2017/04/self-taught-artificial-intelligence-beats-doctors-predicting-heart-attacks (second source) http://spectrum.ieee.org/the-human-os/biomedical/diagnostics/ai-predicts-heart-attacks-more-accurately-than-standard-doctor-method It's happening, whether you like it or not. You will be much better off if you fight for things now that will make life better when it finally happens then you would if you be 1) trying to fight against it, or 2) saying that it won't happen really loudly because it's happening.</quote> 1) The articles mentioned the role of AI machines as an adjunct (i.e. doctor's double checking the work), which is what I mentioned in my post. I didn't say the machines won't be used. I'm saying fully replacing doctors is a looong way away. 2) These studies are done in not very realistic practice scenarios. How do the studies account for patient noncompliance? Poor historianship? Like I said, these machines are sensitive in their stratification for things like CVD risk, identifying individuals who would otherwise not have been considered high risk, but now that decreases the specificity of your findings. Where is the longitudinal study or systematic review of a patient population placed on statins not previously identified as not high risk by AHA criteria from these machines compared to traditional management? Do these individuals have a higher risk for statin-induced myopathy? Rhabdomyelitis? Sure, you prevented the heart attack, but a larger proportion of people suffer kidney injury, but since the study in question was measuring cardiovascular related incidents, its no problem right? Diagnosis is one side of the coin. Does AI help improve outcomes though? Where is the study of machines seeing patients and patients actually taking the machine's recommendations? Compare them to human interaction. You can't force a patient to take a statin, so the machine might recognize who is at risk that otherwise wouldn't have been identified, but the human element of patient care is pretty difficult to overcome. Unless you're planning on shoving those pills down a patient's throat without permission lol. Sure, these machines can take things like patient noncompliance into account and through churning "Big data" find patterns a human mind wouldn't have picked up on. I don't doubt that would improve outcomes. But from the sounds of it, the machines will be used as an adjunct to a physician, not a complete replacement. --- "He has two neurons held together by a spirochete."</quote> ... Copied to Clipboard!
Topic List	Page List: 1