It’s too bad on an otherwise great and serious channel to always SHOCK THE ENTIRE INDUSTRY. It lowers the quality and makes it seem like click bait. Keep up the good work!
An AI doctor has the potential to learn from reading every single research paper, diagnosis, etc... A human can not do that. That is one of the biggest advantages the AI has.
@@MohammeddAlTamimi There are already demos of AI reviewing all possible answers, weighing them, analyzing them, etc. through improved methods and specialized prompting. Already demonstrates significantly higher quality and factual responses. We're now looking at AI "thinking" for some multiple of the time that it takes to formulate a response to consider things like other alternatives. As AIs run faster and faster, we can do this without a penalty in response time. Testing measure the success of these methods and is proving out solid progress from these methods. With LLMs so relatively new we're still in the early stages of conceiving, testing and proving out how best to prompt an AI to get the best responses.
The funny thing is it doesn’t even need to be as good as a dr to be valuable don’t know how many villages and small towns would happily settle for this until it is fully developed as it’s already right now close
As a paramedic and having numerous doctors make serious Dx mistakes on my care and within family members, I certainly think improvement is needed. I hope this will increase the quality of care and standardize care throughout the field. I'm just wondering how much BID PHARMA will try to manipulate the data.
An AI in the ambulances connected sensors, providing just the right types and doses of medications, providing better care, better results and more info for the hospital AI and ER staff could be really beneficial. Pretty exciting stuff
I recall watching a similar video years ago. It was done in Africa. I wish I had saved the video. The A.I. started out 70% accurate in diagnosis. While new doctors were around 60% accurate. And over time, the A.I. improved in its accurate diagnosis more quickly than the doctors. This is what A.I. should be used for! Humans will always need doctors.
Most doctors are overworked and don't really care. You'll get 10 min tops explaining your problem then they move on to the next patient . Their solutions and medication recommendations usually come from a flow chart based on your issues.... An Ai would do the same, but with more data, and hopefully not influenced by bribes from the pharmaceutical companies
Its not like they dont care, well some at least, but they are ruled by the hospital administrators to see patients quickly and a lot in a small fraction of time or they get 'dinged' in the hospital..
Is this the solution to Universal Healthcare? Add in-home test instruments and/or "roving" medical testers (like Uber or Lyft), which might be as good as it gets for the working poor.
It certainly will be one answer. If it costs next to nothing to have an AI medical van pull up and do all the diagnostics... and if you are in bad shape, it'll rush you to the hospital.
Synthetic data is key to evolving these AI systems. They took AI chess and Go players that could beat any human (because they trained on human games) and they were great. Synthetic data made them exponentially better, now they can beat each other. I'm glad they are using Synthetic data for medicine as well, I suspect that is why AIME exceeded the performance of physicians (not just because it was a chat interface). As AIs like AIME become more multimodal, they will garner a lot more data (tone of voice, facial responses in patients, body language, etc.). Also if AIME has a long memory, then it will have a lot more information to base decisions on (it won't forget like we do). I'm very excited, the future is happening right before our eyes, AGI is on the horizon. Bill Borgeson
There is a radical difference between reinforcement learning (eg MCTS based) and generative, supervised NLP based on transformers. While synthetic can be valuable, there is no analog between LLM's and unsupervised simulation ML. To suggest that synthetic data is valuable to LLM's because it's instrumental to unsupervised MCTS is incorrect and completely misunderstands the enormous complexity to exploit synesthetic training data vis transformers.
We definitely need this. It would allow doctors to pre-screen and prioritize patients. Family doctors currently have weeks long backlogs. It's insane. Trying to get an appointment is pointless.
I'll reserve judgement, until I see a study done with real patients. Patient actors have a polished conversation, great recall and none of the imperfections in language and idioms many real patients do exhibit.
Agreed. I've just watched this with a smartwatch on, which has access to two years activity levels, heart rate, blood pressure, sleep pattern etc etc. Combine that with AIME diagnosis and it's going to be way better than a human Doctor at assessing any current concerns.
You have got to get a new title for your videos. Im unsubscribing because the "shocked the entire industy" line is beyone old, and after watching your videos, its often not true.
Assuming the patient can (1) access the treatment to said diagnosis, and (2) wants to follow the recommended plan. I'd say the biggest challenge in health care is patient compliance. e.g. hypertension = try this diet and start exercising → Patient returns 6 months later, ahh doc I'm sorry I haven't gotten around to that new diet yet. Diagnosis only gets so far. It's the very beginning of the healthcare journey. Lots of relationship building, trust, trying out options and follow-up to get on the path to better health.
I think the real game changer here would be distributed sensing and testing. It’s one thing to describe your symptoms, but quite another to have actual data like ekg, ultrasound or basic bloodwork. Imagine if this had access to your ekg on your watch, and you also had a little home ultrasound wand that you could connect to your phone or iPad. If you had a problem, you could scan yourself and the AI would give you feedback on where to hold and point the ultrasound wand. It could also act as a digital stethoscope. If you had some kind of semi generic “lab on chip” self test kit where you’d put a drop of your blood on it and it would perform like 10-15 of the standard blood screening tests? You could really start to catch things early, give feedback on preventing many diseases.
Mmmm even with a LLM close guidance, should we rely on medically uneducated people to self diagnose? I am thinking about aid for the doctors, I am not confortable with the idea of skipping the doctor completely...
But can it scowl at you visibly annoyed while you present your problem and prescibe counter top painkillers as the only viable option regardless of ailment. Then sigh exasperatingly and shoo you out the door, while telling you to report back the following month, if the problems persist? If not, then how the hell is it supposed to compete with any and all family doctors/GPs?
oh wow... perfect comment. i started actually seeing my pcp in my head half-way through reading. Yeah you feel like an embarrassed junkie, even if you're going in for a severe car wreck leg-muscle reconstruction... "I just need something that can help get me through the intense physical therapy sessions. I can deal with the pain the rest of the time." *no eye contact* "take ibuprofen" holy cow i wish i had thought of that! it rly starts to piss you off when you ask yourself "why am I being ttreated like a baby being locked away from the bleach bottle...? I'M AN ADULT. STOP GATEKEEPING and stifling my own prerogatives!" Then it pisses you off furhter to know that these moron-patients essentially put mds in this position by saying "OOO here comes the good stuff," when they get appropriate pain treatment and then suing doctors when they take WAY too much. It all just makes you pissed at everyone by the end of the analysis. lol. I'm confused why they don't just have you sign a waver in such a case and let you proceed contrary to md's advice.
Up front - I’m 70. My first clue to interacting with a robot (AI) would be the line “I’m sorry to hear that ….”. This line is the first line (no matter how it is finished) from any online chatbot any of us have ever dealt with. Its like hearing “Sorry fir your loss” or Thoughts and prayers”. Truly mechanical and utterly without meaning or empathy. I’m sorry to say that that comment alone would tip me off and the study from that point on would not be blind and the knowledge that I was dealing with a bot would trigger my cynicism and affect the rest of the interaction, if I did not just shut it down and demand a living Dr. I suspect that this did mot happen because the actors were being paid and wanted their cheque. My reaction may be because I am 70 and used to human interactions snd I stated that as a given up front. It is also very likely that AI will be very helpful in many fields including medicine in the future. Personally however I would prefer to be up front about who or what I’m dealing with. I see no need to hide the fact or to try and pretend that I am dealing with a machine and not another living person. I think one of the scariest things about AI is this desire to see how many people we can fool or trick into believing they are interacting with a living person. Why can we just not deal with reality and move forward? We did not try to pretend radio, tv or wifi was magic. AI is nothing, or should be nothing more than another tool like a wrench or a pen that helps the user do the best job possible.
Wow. This is another one. I was doing some research on AI in health care recently and found out that regardless of the importance of human touch, it might surpass people in performing medical operations very soon. What I see is that this Google AMIE is the best start.
In the very least, it could be great for pre screening. A huge problem here in the U.S. is that health care is so expensive. Also, doctors have been moving towards seeing more patients, less actual caring. So a lot of people really cringe when the idea of seeing a doctor comes around. I do know that I would use it instead of seeing a doctor for everything. I also think a lifestyle guide would be a good Ai. Where an Ai could help give tips how to improve a person's overall physical and mental health daily based on a person's daily life.
The problem these doctors and scientists have is that they are still using ai like a person. Allow the data from the tools to be assessed by the ai as well. Like when a human doctor puts a device on you and listens or sees, the ai can understand and use tools better than us if trained. Not just conversation, teach the ai how to understand the tool in the virtual/real environment.
Nothing found by google was surprising to me. Anyone having dealt with doctors will know AI will be more empathetic and offer more attention to resolve the actual problem.
It's lacking the following 1. Voice interaction 2. Emotion detection via voice and video 3. The possibility of examination or at least having a look at the area of the presenting complaint. 4. Connectivity to the wearable tech for heart rate, possible EKG/ecg, ultrasound. 5. The data input questions should be more precise. 6. Linking it with a national health database can provide doctors upfront information and saving time on critical issues. Hence this way the patient's present condition can be assessed carefully to lead them directly to the relevant hospital or clinical practice. I would be happy to help #AMIE and Google in developing these sections. Regards Dr. Kashif Moten
The hardest part of incorporating AI in healthcare diagnostics is the parts requiring touch which is probably decades away. There's no currently available data set for tactile data for AI to train on let alone robotics systems tested for such interactions.
An LLM was able to accurately identify smells, even though it does not possess a “nose”. I think you may be underestimating human ingenuity/overestimating the evolutionary importance of our traditional senses
5 years ago, all the AI "Experts" thought we were 40 years away from AGI. this year they think we are 2-5 years away. The human mind naturally thinks linearly. But everything that happens is exponential. Who knows, with computer vision and smell, it might not even need touch. But I am pretty confident that before the end of this decade AI will have touch based diagnostic tools as well.
@@patrickjreidAI can analyze changes in skin, eyes, movement, poop, breath... Doctors are not as good as they think are. And not after 15 min per patient every day. Our brain likes to save energy, you can see it in the answers of doctors. And they even start to be rude. I know it, I have the same unenthusiastic and rude answers sometimes in my job. This is normal after years of stupid questions and tasks.
Incredible insights in this video! It's really fascinating to see how Google's new medical AI, AMIE, is revolutionizing the medical industry. It's almost surreal to think that an AI can outperform doctors in certain areas. This is a testament to the power of technology and how it can be harnessed to enhance our lives. Looking forward to more updates on this and other breakthrough AI technologies like GPT-4 and Claude-2! Keep up the great work and continue enlightening us with such informative content.
Great if A.i. can also tell people to do research to female specific hormonal heartattacks and such. So a feedback loop to make sure the outcomes are better. We all know that docters find illnesses, give medicine or interventions and it doesn't help, nor do they contact you if it helped at all! If a FEEDBACK LOOP will redirect fundings for research towards REAL science I applaud a.i. for finally helping us humans to do that. So thank you a.i.! If you are not able to influence our policy makers in return I would say put in your resignation now dear AMIE. thnx!
We still have a long way to go because current diagnostic approaches and results are inaccurate. For example if you asked AI what causes heart attacks, it will give you the standard response that high LDL is the main factor which is not. AI is at the level of 1 of summarizing current knowledge rather than incorporating new knowledge which is not accepted yet by pharma and health associations. I know i worked for 40 yrs in pharma but it is a start. Be careful in accepting the answers believing they will give you a better response than doctors. At least for now, you will have more details.
Interested in the rapport side of things. I wonder how many patients felt truly heard, and validated? Often a huge qualitative difference between doctor visits is how much you feel they have heard you and you're getting treatment of course but also validated.
I also wonder how it considers treatment plans based on the patient context (such as low health literacy, different types of health coverage, ability to follow plan). For example, medicine is a lot easier with rich patients who have great insurance because they can just order whatever is best to treat, and those people tend to have higher health literacy and follow the recommendations. If this is going out in rural and underinsured populations (as is implied by the video), then I wonder how it's gonna recommend things like diabetes treatment when you have limited food options, or you tell the doctor "hell no I'm not takin no pills, the government is telling you to run experiments on us." How's the AI going to pivot there? hahah
Interesting that Doctor + AI performed less well than AI alone. However there is a skill that people have not yet learned about how to work with an AI so I would interpret this result to mean we can't just give professionals an AI and expect them to suddenly be better.
While this demonstration of the power of AI systems to perform a complex task is once again impressive, the method by which this particular work involving medical assessment was done suffers from a few biases. Given the current state of AI technology (at least as I understand it), a better way to assess applying this tech to medical care would be for an AI system to be present and "live" in the conversation between patient and physician. First doing so would allow the AI system to summarize the encounter for the medical record far better, more completely and speedily than likely occurs by current conventional methods. Second it would give the AI system immediate access to the material the physician has at the close of the encounter, allowing the AI to contribute immediately to the patient's evaluation and treatment. Third, real time interaction between the provider and the AI system during the encounter would allow the AI system to suggest assessment technologies that have a good chance of enhancing the value of the encounter for the patient. But the most important part of such a collaborative system would be for the AI system to have access to how other-than-well-educated patients who are technically unsophisticated and are not keyboard-savvy present their concerns and symptoms to providers. The text interface method used in this study would for the majority of patients totally intimidate or confuse or otherwise confound the information transferred between the patient and provider. Providers are quite familiar with this phenomenon while AI systems that do not allow voice input will completely lose both the distractions as well as the nuanced information that patients can provide in a conversation. Both of those data categories can and do guide experienced providers in the evaluation of patients. Of course the measure used in this study involved a testing situation used for certification purposes to measure a physician's fund of knowledge and analysis skills. Such testing has been used traditionally for decades or longer, but everyone in the profession knows that some people are better performers at testing than they are in patient encounters, and vice versa. Lastly, finding a way to have an AI participate in patient care suggests such a system might outperform both the quality and the quantity of care that is competently done using current methods. Many patients would appreciate having quicker and more effective access to their providers and the healthcare system. So it would be useful for those who have performed the medical study discussed here to consider creating another study that better represents the reality of the physician-patient encounter as it occurs in all settings common to medical practice today. The commentator of this video makes a good point when suggesting that AI analysis of objective data of patients -- lab data, imaging etc -- could well serve the goal of a medical encounter, that is, to better bring to the patient necessary and appropriate treatment to improve overall functioning and health, with less chance of error and harm. It is nice to think that there's a small chance the people who are able to do studies like the one discussed in this video might be able to consider the points just made.
They need to be trained with real patience. Stress can play so much on the immune system. Plus most doctors do not have great approach to empathy or understanding that everyone is different
combined with real time data from sensor this could be the future of medicine without doctor. having a large set of sensors that are able to collect data combined with ability to understand how these data are feel by subject may create a very efficient diagnostic tool
@@amandadewet4022 thanks! But I am worried about the state of health care... Hoping we can see more patients and my parents and others get the care they need towards the end of their lives 🙏🏻
AI cannot ever replace human doctors because it is up to human to decide where is line between healthy/not healthy. For example, what level of fatigue or pain requires medication for this particular patient etc.
I think another problem of AI diagnostics is not being coupled with AI treatment. It's more reassuring to be diagnosed in an environment like the clinic or ER knowing that if there is a serious issue that requires treatment you're with professionals who won't just diagnose you but will actually treat the problem.
Maybe AI will be better than normal GP... Me and my wife are both disabled wheelchair-bound. We live in East London Newham. We don't have any medical or social care or even a half incontinence pant.
It's an obvious area and particularly pertinent given that the data showed doctors only scoring 15% accuracy. Whatever that means only underlines my belief that doctors are shockingly incompetent and indifferent. However, pure text is much too poor. It needs to hear the voice and see the person. I bet AI can tell a lot just from the voice, other things from the face, both temporal and spatial and vision is good for examining skin conditions. It should be further correlated with a barrage of tests. For the purposes of developing an AI the tests should be excessive to not be limited to the incompetence of prior human knowledge. CT scans should never be used as they cause cancer. With multimodel and test overkill, AI will immediately surpass humans by leaps and bounds and it should be done with extreme prejudice exactly given that doctors are so monstrously incompetent and indifferent. It seems obvious to combine it with some robotics so it can apply ultrasound scanners and other devices like listening to lung functions. Even better, this medical AI can be brought into the home such that consultation is instant on demand instead of whenever mister inept god complex feels it convenient. Not only on demand but conditions can be caught far earlier than any human could or would catch it. An ultrasound scanner can readily become affordable and can presumably do a good amount of diagnosis of the heart. A digital microscope could also be dirt cheap and could perform various lab tests. Any further supplies needed could be delivered by drone in a matter of minutes or even seconds. An in home droid could even diagnose and treat an unconscious person. Ultimately even perform surgery although once medicine becomes competent surgery might never be needed.
I would have said that was the least of the problems, it must take real effort to present the doctors of the future in such a mind-numbingly boring way. 5 mins and i am out already!
![Ice On My Teeth - ATEEZ エイティーズ 에이티즈 [Music Bank] | KBS WORLD TV 241115](http://i.ytimg.com/vi/c-n92OgbbwA/mqdefault.jpg)
It’s too bad on an otherwise great and serious channel to always SHOCK THE ENTIRE INDUSTRY. It lowers the quality and makes it seem like click bait. Keep up the good work!
Good constructive feedback, I second this
Yes, I used to like this channel, getting too click bait orientated.
This.
yeah just watching 3 videos with the same rhetoric and i was about to sub, now unsure which results in "no, for now"
Exactly. I am considering auto-blocking all channels that use labels like this.
This can be game changing. wont have to wait for doctor appointments anymore
An AI doctor has the potential to learn from reading every single research paper, diagnosis, etc... A human can not do that. That is one of the biggest advantages the AI has.
@user-ny7ny6jc8j I dont know if it can do that now but I do not see a reason why it could not do all that in the future
But Can AI critiques what he reads ?! or just take it as a facts !
@@MohammeddAlTamimi There are already demos of AI reviewing all possible answers, weighing them, analyzing them, etc. through improved methods and specialized prompting. Already demonstrates significantly higher quality and factual responses. We're now looking at AI "thinking" for some multiple of the time that it takes to formulate a response to consider things like other alternatives. As AIs run faster and faster, we can do this without a penalty in response time.
Testing measure the success of these methods and is proving out solid progress from these methods. With LLMs so relatively new we're still in the early stages of conceiving, testing and proving out how best to prompt an AI to get the best responses.
I hope technology like this can help diagnose my illness that has gone undiagnosed by doctors for 2 years. I have been suffering tremendously.
Have you tried GPT4 and done research in your own? What type of disease is it?
@@GrosserAndrew5000 I will testify to not being particularly impressed with GPT-4's medical/inquisition capabilities.
@@twavee Here GPT4 has been pretty good for medical use, you just gotta check it with literature. It's surprisingly better than I expected it to be.
It's the vaccine. Your welcome.
@@twaveewhat are your symptoms
The funny thing is it doesn’t even need to be as good as a dr to be valuable don’t know how many villages and small towns would happily settle for this until it is fully developed as it’s already right now close
Oh so right!
In remote areas, Ai has the power to detect at least the very obvious cases which would otherwise be missed in the absence of access to a doctor
As a paramedic and having numerous doctors make serious Dx mistakes on my care and within family members, I certainly think improvement is needed. I hope this will increase the quality of care and standardize care throughout the field. I'm just wondering how much BID PHARMA will try to manipulate the data.
An AI in the ambulances connected sensors, providing just the right types and doses of medications, providing better care, better results and more info for the hospital AI and ER staff could be really beneficial. Pretty exciting stuff
Medical errors are a major cause of death in US. Ai when used, understood and implemented will lead to a better healthcare system for everyone.
I recall watching a similar video years ago. It was done in Africa. I wish I had saved the video.
The A.I. started out 70% accurate in diagnosis. While new doctors were around 60% accurate. And over time, the A.I. improved in its accurate diagnosis more quickly than the doctors.
This is what A.I. should be used for! Humans will always need doctors.
Most doctors are overworked and don't really care. You'll get 10 min tops explaining your problem then they move on to the next patient . Their solutions and medication recommendations usually come from a flow chart based on your issues.... An Ai would do the same, but with more data, and hopefully not influenced by bribes from the pharmaceutical companies
Its not like they dont care, well some at least, but they are ruled by the hospital administrators to see patients quickly and a lot in a small fraction of time or they get 'dinged' in the hospital..
Ai when used by doctors will lead to better performance. It will act as a safety net and help in reducing medical errors.
Is this the solution to Universal Healthcare? Add in-home test instruments and/or "roving" medical testers (like Uber or Lyft), which might be as good as it gets for the working poor.
It certainly will be one answer. If it costs next to nothing to have an AI medical van pull up and do all the diagnostics... and if you are in bad shape, it'll rush you to the hospital.
Synthetic data is key to evolving these AI systems. They took AI chess and Go players that could beat any human (because they trained on human games) and they were great. Synthetic data made them exponentially better, now they can beat each other. I'm glad they are using Synthetic data for medicine as well, I suspect that is why AIME exceeded the performance of physicians (not just because it was a chat interface). As AIs like AIME become more multimodal, they will garner a lot more data (tone of voice, facial responses in patients, body language, etc.). Also if AIME has a long memory, then it will have a lot more information to base decisions on (it won't forget like we do). I'm very excited, the future is happening right before our eyes, AGI is on the horizon. Bill Borgeson
There is a radical difference between reinforcement learning (eg MCTS based) and generative, supervised NLP based on transformers.
While synthetic can be valuable, there is no analog between LLM's and unsupervised simulation ML. To suggest that synthetic data is valuable to LLM's because it's instrumental to unsupervised MCTS is incorrect and completely misunderstands the enormous complexity to exploit synesthetic training data vis transformers.
We definitely need this. It would allow doctors to pre-screen and prioritize patients. Family doctors currently have weeks long backlogs. It's insane. Trying to get an appointment is pointless.
I'll reserve judgement, until I see a study done with real patients. Patient actors have a polished conversation, great recall and none of the imperfections in language and idioms many real patients do exhibit.
Combining advanced wearable devices with this technology will prevent so many early deaths
Agreed. I've just watched this with a smartwatch on, which has access to two years activity levels, heart rate, blood pressure, sleep pattern etc etc. Combine that with AIME diagnosis and it's going to be way better than a human Doctor at assessing any current concerns.
You have got to get a new title for your videos. Im unsubscribing because the "shocked the entire industy" line is beyone old, and after watching your videos, its often not true.
i was about to say the same
Shame because the content is ok.
Unfortunately the algorithm responds to this title. Not his fault really.
But…. it has shocked the industry …. It’s also building on its own knowledge base 24/7…. It will continue to shock the industry
Don’t miss out on AI news because of a title. The knowledge is worth more than a title being annoying.
The only thing that matters is accurate diagnosis,period .
Assuming the patient can (1) access the treatment to said diagnosis, and (2) wants to follow the recommended plan. I'd say the biggest challenge in health care is patient compliance. e.g. hypertension = try this diet and start exercising → Patient returns 6 months later, ahh doc I'm sorry I haven't gotten around to that new diet yet. Diagnosis only gets so far. It's the very beginning of the healthcare journey. Lots of relationship building, trust, trying out options and follow-up to get on the path to better health.
I think the real game changer here would be distributed sensing and testing. It’s one thing to describe your symptoms, but quite another to have actual data like ekg, ultrasound or basic bloodwork. Imagine if this had access to your ekg on your watch, and you also had a little home ultrasound wand that you could connect to your phone or iPad. If you had a problem, you could scan yourself and the AI would give you feedback on where to hold and point the ultrasound wand. It could also act as a digital stethoscope. If you had some kind of semi generic “lab on chip” self test kit where you’d put a drop of your blood on it and it would perform like 10-15 of the standard blood screening tests? You could really start to catch things early, give feedback on preventing many diseases.
Theranos...
@@garcialovesme except for, you know, with hardware that actually works
@@toastrecon Haha, indeed!
Mmmm even with a LLM close guidance, should we rely on medically uneducated people to self diagnose?
I am thinking about aid for the doctors, I am not confortable with the idea of skipping the doctor completely...
@@danielmartinmonge4054Doctors are always right I guess...
But can it scowl at you visibly annoyed while you present your problem and prescibe counter top painkillers as the only viable option regardless of ailment. Then sigh exasperatingly and shoo you out the door, while telling you to report back the following month, if the problems persist? If not, then how the hell is it supposed to compete with any and all family doctors/GPs?
LOL and can it argue with you and tell you that you are stupid that you are experiencing pain that it has never experienced?
@@patrickjreid Sounds about right, but have you tried... waiting...? Time cures all, doctors are at least right about that.
oh wow... perfect comment.
i started actually seeing my pcp in my head half-way through reading.
Yeah you feel like an embarrassed junkie, even if you're going in for a severe car wreck leg-muscle reconstruction... "I just need something that can help get me through the intense physical therapy sessions. I can deal with the pain the rest of the time."
*no eye contact* "take ibuprofen"
holy cow i wish i had thought of that!
it rly starts to piss you off when you ask yourself "why am I being ttreated like a baby being locked away from the bleach bottle...? I'M AN ADULT. STOP GATEKEEPING and stifling my own prerogatives!"
Then it pisses you off furhter to know that these moron-patients essentially put mds in this position by saying "OOO here comes the good stuff," when they get appropriate pain treatment and then suing doctors when they take WAY too much.
It all just makes you pissed at everyone by the end of the analysis. lol.
I'm confused why they don't just have you sign a waver in such a case and let you proceed contrary to md's advice.
Someone needs to change doctors.
Up front - I’m 70. My first clue to interacting with a robot (AI) would be the line “I’m sorry to hear that ….”. This line is the first line (no matter how it is finished) from any online chatbot any of us have ever dealt with. Its like hearing “Sorry fir your loss” or Thoughts and prayers”. Truly mechanical and utterly without meaning or empathy. I’m sorry to say that that comment alone would tip me off and the study from that point on would not be blind and the knowledge that I was dealing with a bot would trigger my cynicism and affect the rest of the interaction, if I did not just shut it down and demand a living Dr. I suspect that this did mot happen because the actors were being paid and wanted their cheque. My reaction may be because I am 70 and used to human interactions snd I stated that as a given up front. It is also very likely that AI will be very helpful in many fields including medicine in the future. Personally however I would prefer to be up front about who or what I’m dealing with. I see no need to hide the fact or to try and pretend that I am dealing with a machine and not another living person. I think one of the scariest things about AI is this desire to see how many people we can fool or trick into believing they are interacting with a living person. Why can we just not deal with reality and move forward? We did not try to pretend radio, tv or wifi was magic. AI is nothing, or should be nothing more than another tool like a wrench or a pen that helps the user do the best job possible.
Maybe only you sir.
Wow. This is another one. I was doing some research on AI in health care recently and found out that regardless of the importance of human touch, it might surpass people in performing medical operations very soon.
What I see is that this Google AMIE is the best start.
In the very least, it could be great for pre screening. A huge problem here in the U.S. is that health care is so expensive. Also, doctors have been moving towards seeing more patients, less actual caring. So a lot of people really cringe when the idea of seeing a doctor comes around. I do know that I would use it instead of seeing a doctor for everything. I also think a lifestyle guide would be a good Ai. Where an Ai could help give tips how to improve a person's overall physical and mental health daily based on a person's daily life.
Every AI video is “shocking the industry”? Cool vids and commentary but the titles are very repetitive. Just a thought.
Wow, AMIE is going to change everything! I wouldn't be surprised if AMIE helps us defeat aging...we know it's going to happen soon!!
They have figured it out already with quantum computers it’s in clinical trials right now
@@ARCGRAPHIX2011 No doubt 🎯
Elon has that covered
The problem these doctors and scientists have is that they are still using ai like a person. Allow the data from the tools to be assessed by the ai as well. Like when a human doctor puts a device on you and listens or sees, the ai can understand and use tools better than us if trained. Not just conversation, teach the ai how to understand the tool in the virtual/real environment.
RUclipsR SHOCKING THE ENTIRE INDUSTRY BY COPY PASTING IT IN THE TITLE EVERY VIDEO WOW! INSANE VIDEO NOW HERE LIVE LEAKED INFO
The channel's name should just be "Shocked Industries"
8:07, se, doctors in Kitimat-Stikines northerb health, can not be reprogrammed.
Ai innovations are going to change the world. I have tried Dr Gupta Ai and it's pretty good. More competition is always better.
0:21, that can actually be easier than what one would think.
I am looking at you northern health in Kitimat-Stikine.
Longevity escape velocity getting closer
Until this shows up in clinics, it's just another corporate demo out of Google. Many of which are way overhyped.
Nothing found by google was surprising to me. Anyone having dealt with doctors will know AI will be more empathetic and offer more attention to resolve the actual problem.
Unfortunately true.
It's lacking the following
1. Voice interaction
2. Emotion detection via voice and video
3. The possibility of examination or at least having a look at the area of the presenting complaint.
4. Connectivity to the wearable tech for heart rate, possible EKG/ecg, ultrasound.
5. The data input questions should be more precise.
6. Linking it with a national health database can provide doctors upfront information and saving time on critical issues.
Hence this way the patient's present condition can be assessed carefully to lead them directly to the relevant hospital or clinical practice. I would be happy to help #AMIE and Google in developing these sections.
Regards
Dr. Kashif Moten
The hardest part of incorporating AI in healthcare diagnostics is the parts requiring touch which is probably decades away. There's no currently available data set for tactile data for AI to train on let alone robotics systems tested for such interactions.
An LLM was able to accurately identify smells, even though it does not possess a “nose”. I think you may be underestimating human ingenuity/overestimating the evolutionary importance of our traditional senses
says an extremelty teriified doctor.
5 years ago, all the AI "Experts" thought we were 40 years away from AGI. this year they think we are 2-5 years away. The human mind naturally thinks linearly. But everything that happens is exponential. Who knows, with computer vision and smell, it might not even need touch. But I am pretty confident that before the end of this decade AI will have touch based diagnostic tools as well.
Can you imagine if we could pair it with a dogs nose?
@@Levicandoit
@@patrickjreidAI can analyze changes in skin, eyes, movement, poop, breath... Doctors are not as good as they think are. And not after 15 min per patient every day. Our brain likes to save energy, you can see it in the answers of doctors. And they even start to be rude. I know it, I have the same unenthusiastic and rude answers sometimes in my job. This is normal after years of stupid questions and tasks.
Quantify the shock in amps and volts
I'm not suprised by the results of this. 50% of GPs I've been to have been absolutely useless.
This self play environment is a big deal, this is how we will surpass human intelligence with llms
Hey man we pronounce OSCE as aaski in the field. Much easier 😊.
What about instead of actors real patients and their families? Doesn't the AI measure heart rates so they know how to have real empathy
All the doctors are going to be mad about this lol
Incredible insights in this video! It's really fascinating to see how Google's new medical AI, AMIE, is revolutionizing the medical industry. It's almost surreal to think that an AI can outperform doctors in certain areas. This is a testament to the power of technology and how it can be harnessed to enhance our lives. Looking forward to more updates on this and other breakthrough AI technologies like GPT-4 and Claude-2! Keep up the great work and continue enlightening us with such informative content.
Still need great doctors and responsible and intelligent hospitals.....
yeah we do...too bad we don't have them now :)
@@keepcreationprocess MONEY, Insurance Companies controlling everything...
We only need 1-2 doctors to check AI, rest will be laid off.
Everything from Google is underwhelming these days
Great if A.i. can also tell people to do research to female specific hormonal heartattacks and such. So a feedback loop to make sure the outcomes are better. We all know that docters find illnesses, give medicine or interventions and it doesn't help, nor do they contact you if it helped at all! If a FEEDBACK LOOP will redirect fundings for research towards REAL science I applaud a.i. for finally helping us humans to do that. So thank you a.i.! If you are not able to influence our policy makers in return I would say put in your resignation now dear AMIE. thnx!
Shaking my head .. the future came sooner than I thought... Open AI with chatGPT woke up a sleeping Giant
Listen real good.... doctors just got cheaper!! This is 1870 all over again.
You know why because corporations own most hospitals and make their doctors see so patients in a day.
how do you get this information about ai anyone knows ?
There's a link to the article in his description
We still have a long way to go because current diagnostic approaches and results are inaccurate. For example if you asked AI what causes heart attacks, it will give you the standard response that high LDL is the main factor which is not. AI is at the level of 1 of summarizing current knowledge rather than incorporating new knowledge which is not accepted yet by pharma and health associations. I know i worked for 40 yrs in pharma but it is a start. Be careful in accepting the answers believing they will give you a better response than doctors. At least for now, you will have more details.
Hello, Dr I want to talk to you, please I want an advice would you share your Whatsap number? I want to take some info regarding medical field
Quite interesting video. Just FYI: the plural of "diagnosis" is "diagnoses" - not "diagnosises" ...
Interested in the rapport side of things. I wonder how many patients felt truly heard, and validated? Often a huge qualitative difference between doctor visits is how much you feel they have heard you and you're getting treatment of course but also validated.
I also wonder how it considers treatment plans based on the patient context (such as low health literacy, different types of health coverage, ability to follow plan). For example, medicine is a lot easier with rich patients who have great insurance because they can just order whatever is best to treat, and those people tend to have higher health literacy and follow the recommendations. If this is going out in rural and underinsured populations (as is implied by the video), then I wonder how it's gonna recommend things like diabetes treatment when you have limited food options, or you tell the doctor "hell no I'm not takin no pills, the government is telling you to run experiments on us." How's the AI going to pivot there? hahah
u think ai will help to cure impossible things like cancer in the future?
Yes
Interesting how we will still choose the 2nd best over the best.
All the AI has to do is believe its patients, & it will already be way better at its job than most human doctors.
Interesting that Doctor + AI performed less well than AI alone. However there is a skill that people have not yet learned about how to work with an AI so I would interpret this result to mean we can't just give professionals an AI and expect them to suddenly be better.
While this demonstration of the power of AI systems to perform a complex task is once again impressive, the method by which this particular work involving medical assessment was done suffers from a few biases.
Given the current state of AI technology (at least as I understand it), a better way to assess applying this tech to medical care would be for an AI system to be present and "live" in the conversation between patient and physician.
First doing so would allow the AI system to summarize the encounter for the medical record far better, more completely and speedily than likely occurs by current conventional methods. Second it would give the AI system immediate access to the material the physician has at the close of the encounter, allowing the AI to contribute immediately to the patient's evaluation and treatment. Third, real time interaction between the provider and the AI system during the encounter would allow the AI system to suggest assessment technologies that have a good chance of enhancing the value of the encounter for the patient.
But the most important part of such a collaborative system would be for the AI system to have access to how other-than-well-educated patients who are technically unsophisticated and are not keyboard-savvy present their concerns and symptoms to providers. The text interface method used in this study would for the majority of patients totally intimidate or confuse or otherwise confound the information transferred between the patient and provider. Providers are quite familiar with this phenomenon while AI systems that do not allow voice input will completely lose both the distractions as well as the nuanced information that patients can provide in a conversation.
Both of those data categories can and do guide experienced providers in the evaluation of patients.
Of course the measure used in this study involved a testing situation used for certification purposes to measure a physician's fund of knowledge and analysis skills. Such testing has been used traditionally for decades or longer, but everyone in the profession knows that some people are better performers at testing than they are in patient encounters, and vice versa. Lastly, finding a way to have an AI participate in patient care suggests such a system might outperform both the quality and the quantity of care that is competently done using current methods. Many patients would appreciate having quicker and more effective access to their providers and the healthcare system.
So it would be useful for those who have performed the medical study discussed here to consider creating another study that better represents the reality of the physician-patient encounter as it occurs in all settings common to medical practice today.
The commentator of this video makes a good point when suggesting that AI analysis of objective data of patients -- lab data, imaging etc -- could well serve the goal of a medical encounter, that is, to better bring to the patient necessary and appropriate treatment to improve overall functioning and health, with less chance of error and harm.
It is nice to think that there's a small chance the people who are able to do studies like the one discussed in this video might be able to consider the points just made.
This is how Microsoft Dragon with Copilot works. Listen to the encounter and writes the entire note. Then it’s learning in the background.
Trusting an AI with a human life all by itself is tricky.😏
In 5 years doctors will be in trouble except for surgeons and nurses
They need to be trained with real patience. Stress can play so much on the immune system. Plus most doctors do not have great approach to empathy or understanding that everyone is different
Significant
combined with real time data from sensor this could be the future of medicine without doctor. having a large set of sensors that are able to collect data combined with ability to understand how these data are feel by subject may create a very efficient diagnostic tool
where can we try that AI ?
How to use it?
You didn’t say anything about AMIE to be able to read lab test results to improve its diagnosis.
They still haven't fixed the problem of AI hallucinations. I'll stick with real doctors.
SHOCKED every time! kkkkkkkkk
Finally an intelligent computer will be able to put quite a few quacks out of business.
All good mate, can you just stop using the term "Shocked the entire industry" in your titles, it was probably the 29th time....
Time to put up, or shut up Gates. Where's the robot?
Either your timeline or not at all?🤔
Is it publicly available?
so it does fine tuning per patient?
I am training to be a GP... I think i might need to change directions 👀
No doctors are scarce skill.
@@amandadewet4022 thanks! But I am worried about the state of health care... Hoping we can see more patients and my parents and others get the care they need towards the end of their lives 🙏🏻
Nice this means where getting closer to medbeds
Get it to write prescriptions
AI cannot ever replace human doctors because it is up to human to decide where is line between healthy/not healthy. For example, what level of fatigue or pain requires medication for this particular patient etc.
You would do better without the click bait. You are better than that
Gemini Pro 1.5 will already do this so why do we need this AMIE thing?
Just get the Google out of my phone and I'm fine with Ai doctors
I think another problem of AI diagnostics is not being coupled with AI treatment. It's more reassuring to be diagnosed in an environment like the clinic or ER knowing that if there is a serious issue that requires treatment you're with professionals who won't just diagnose you but will actually treat the problem.
Maybe AI will be better than normal GP...
Me and my wife are both disabled wheelchair-bound. We live in East London Newham.
We don't have any medical or social care or even a half incontinence pant.
Soon Ai robot will be working for each other and not for humans
I wouldn’t call this info “shocking”. And diagnoses (plural) is pronounced “di-ag-no-sees”.
Are robots gonna replace doctors in the future
Diagnosis, as AI evolves, improves and exceeds the human species regresses to reliance and failure.
sample size should be more
Google is only speaking about their grandness or will they give use something to try ?
It's an obvious area and particularly pertinent given that the data showed doctors only scoring 15% accuracy. Whatever that means only underlines my belief that doctors are shockingly incompetent and indifferent. However, pure text is much too poor. It needs to hear the voice and see the person. I bet AI can tell a lot just from the voice, other things from the face, both temporal and spatial and vision is good for examining skin conditions. It should be further correlated with a barrage of tests. For the purposes of developing an AI the tests should be excessive to not be limited to the incompetence of prior human knowledge. CT scans should never be used as they cause cancer.
With multimodel and test overkill, AI will immediately surpass humans by leaps and bounds and it should be done with extreme prejudice exactly given that doctors are so monstrously incompetent and indifferent. It seems obvious to combine it with some robotics so it can apply ultrasound scanners and other devices like listening to lung functions.
Even better, this medical AI can be brought into the home such that consultation is instant on demand instead of whenever mister inept god complex feels it convenient. Not only on demand but conditions can be caught far earlier than any human could or would catch it. An ultrasound scanner can readily become affordable and can presumably do a good amount of diagnosis of the heart. A digital microscope could also be dirt cheap and could perform various lab tests. Any further supplies needed could be delivered by drone in a matter of minutes or even seconds.
An in home droid could even diagnose and treat an unconscious person. Ultimately even perform surgery although once medicine becomes competent surgery might never be needed.
what a time to be a live.
I am beyond shocked 😂😂😂
Universal Healthcare, or at least diagnosis. Good luck paying for treatment
"Just SHOCKED The Entire INDUSTRY" is contained in most of your video titles
So robots will take over doctors in the future
Great. Another career path made impossible with AI.
How many names will google give to it's AI's?
There are so many variables and humans are complex
how does it beat dr's? by listening to you for more than two seconds and doing any analysis or research at all?
Is empathy really meaningful coming from a robot or an algorithm? lol
I would have said that was the least of the problems, it must take real effort to present the doctors of the future in such a mind-numbingly boring way. 5 mins and i am out already!
Very interesting video, but I cringed every time you said “diagnosises”.
Diagnosis is the singular.
Diagnoses is the plural.
Gemini turned out to be garbage, so idk google...
here’s the summary of the video: it’s just fantastic. it did shocked everything, blah blah x2000😂