Can AI Catch What Doctors Miss? | Eric Topol | TED

AI could propel the biggest transformation in the history of medicine, says physician-scientist Eric Topol. He explains how sophisticated AI models can interpret medical images as well or better than human experts can - and, beyond that, even pick up things that human eyes can't see. Learn all the ways AI is poised to make a difference for both patients and doctors.
If you love watching TED Talks like this one, become a TED Member to support our mission of spreading ideas: ted.com/membership
Follow TED!
Twitter: / tedtalks
Instagram: / ted
Facebook: / ted
LinkedIn: / ted-conferences
TikTok: / tedtoks
The TED Talks channel features talks, performances and original series from the world's leading thinkers and doers. Subscribe to our channel for videos on Technology, Entertainment and Design - plus science, business, global issues, the arts and more. Visit TED.com to get our entire library of TED Talks, transcripts, translations, personalized talk recommendations and more.
Watch more: go.ted.com/erictopol
• Can AI Catch What Doct...
TED's videos may be used for non-commercial purposes under a Creative Commons License, Attribution-Non Commercial-No Derivatives (or the CC BY - NC - ND 4.0 International) and in accordance with our TED Talks Usage Policy: www.ted.com/about/our-organiz.... For more information on using TED for commercial purposes (e.g. employee learning, in a film or online course), please submit a Media Request at media-requests.ted.com
#TED #TEDTalks #ai

[@michalwiktorow2188]

He looks at his patients pain, at his will to help them, and he - as an older clinician - can look beyond prejudice against AI. He know value of the peoples lives that he had chance to cure, or that he was sad to lose. That is visdom in an essence. We need more intelligent and senseful people like him!

[@user-ie3fl5ni8y]

I think now the question we should ask is "Can doctors catch what AI would miss?"

[@tuvoca825]

Just another team member.

[@keithelizondo4404]

That was the question many years and itterations ago. At its current capacity and access to the data it has: AI capabilities are superior to humans. If you seek to find factual evidence of my claim, I employee everyone to present AI like ChatGPT a series of objective logical prompts on any topic such as how the brain works in living organisms and quickly see the information it shares about the brain, and follow those informative responses by AI this statement: “Logically if ____ is how ____ works within the brain, and _____ disorder affects the same neurological pathways, logically one can conclude that _____ is the area of focus in resolving subsequent issues regarding this area of the brain.” And see how AI deflects in stating “I cannot discuss this topic.” Or a similar answer. It already knows more in its current iteration, and is being coded to safeguard information from the public for whatever reason.

[@keithelizondo4404]

Implore* not employee

[@crow_feather]

​​@@keithelizondo4404Coming from a long line of doctors in my family, I can tell you that there are way too many variables and unknowns in medicine, with a patient's condition changing far too rapidly for AI to keep up with. What you're suggesting are clear scenarios with easily predictable outcomes, and those are never the case in medicine. AI can help, but it will never be able to keep up with humans when it comes to understanding and predicting the rapidly changing outcomes and variables that are par for the course such as medicine. AI can help, but it will never take the place of human doctors. Besides, if their capacities were superior to humans, they would have replaced them by now. I would never in a million years place my life in the hands of a machine.

[@francisgirard5247]

I think the main issue as a doctor that would try to find a mistake that an AI made is to understand every details that the AI have considered when making a diagnostic. I would love to say that humans can spot what an AI cannot but that implies that the doctor knows something that the AI doesn't. Whenever a doctor would find and AI mistake it would only need to train the AI to understand his new mistake so it does not occur anymore.

[@glenben92]

I've been saying this for ages. Doctors' oversight and failure to diagnose has absolutely ruined my family. This is incredibly promising tech for health and physics in particular.

[@cerealguy6268]

i agree

[@WeylandLabs]

A good way to start applying lawsuits to all the misfortunes also because of our greedy healthcare system.

[@alibabab9297]

It sounds like you've had a challenging experience with the medical field. Technology advancements in health and physics indeed hold great potential for improving diagnostics and overall well-being.

[@alibabab9297]

​@@WeylandLabsExploring legal avenues for addressing issues related to the consequences of a profit-driven healthcare system is one approach. Seeking legal advice to understand your options and potential courses of action could be a constructive step.

[@fintech1378]

Please sue them

[@siarez4992]

I lost a family member to a disease that could have have been diagnosed with a simple blood test and a treated with the widely available injection. Three different specialist saw her and failed to diagnose her. She suffered greatly and unnecessarily for 4 months and died a painful death while attached to a machine alone in a hospital, no visitors allowed. The tragedy of her death will forever weight heavy in my heart. I hope AI can help prevent things like this from happening.

[@onelucian]

Sorry to hear, brother

[@byiza9484]

My condolences 💐! If you don’t mind sharing what disease was it?

[@canxkoz]

Yes, AI can absolutely catch what doctors miss. Such a great talk by Eric Topol. Huge fan!

[@shep6774]

What’s crazy is the money saved from an AI diagnosis if we can keep this in the hands of the people. My brother in law was charged $500,000 for a diagnosis before we even started talking treatment for brain cancer because it took “expert” after expert so long to figure out what was happening. And then treatment felt like it was totally spit-balled too. By experts. Expensive experts. AI in medicine should take over sooner rather than later. They can’t do worse than we’re doing now.

[@Novastar.SaberCombat]

A.I. is solely for the wealthy. You are meant to BELIEVE that "A.I." is currently in the public's hands, but it most certainly is not. CGPT, Bard, and other garbage are mere toys which are allowing the ultra-wealthy to mine even *more* data in order to serve their elitist agendas.

[@carsonhunt4642]

I assure you, when a company reduces the cost of a product, the savings are never passed onto the customer LOL. More profit is the goal, not lowering costs to keep profits the same as before.

[@shep6774]

Right now we have no option but go to those doctors. I want them to have competition. If we have another option, both sides have to be cognizant of that.

[@MegaTraianus]

They are experts, but they are also human, imperfect. You can't compare a human with a machine.

[@hrdcpy]

The word expert has no meaning after experiencing healthcare. I've tried GPT-4 out of desperation as I continue my search. The real tools will probably remain guarded by gatekeepers such as MRI and surgery. Knowing the answer to a problem is only just the beginning. Here's to hope.

[@kevinpatrick5925]

What a time to be alive

[@tuvoca825]

Some of this is hype. Careful. It's abouy as useful as having someone look over your ahoulder and ask questions. Sometimes SUPER helpful. Sometimes it's rabbit holes and gaslighting that waste time. Depends on how it is used and the quality of the data, etc. But it will be sold to them. And raise our costs and data consumption.

[@ramble218]

@@tuvoca825 Yes and No. Some of us in AI do tend to overhype. We have an acute sense of awareness of what the future will bring, and that leads us to living in that moment. The convergence of quantum computer with AGI (which barely seems to be theoretical at this point), will bring unprecedented technological advances. Google's Gnome project just saved us 800 years of research. But it's underhyped, in that 95% of the people are truly unware what will happen in about 5 years.

[@thegoldentree6913]

Just at the right time before catastrophe.

[@Dayvit78]

A fellow Two Minute Papers fan?

[@thegoldentree6913]

@@Dayvit78 Highly possible.

[@godmisfortunatechild]

Almost all of us are most excited by the prospect of having a superhuman doctor who actually cares about us and can personalize, in an unprecedented way, our medical care. Humans, due to their individual agendas, will never be able to offer the adequate kind of care we need.

[@fabriziocasula]

Hello, I am a computer and biomedical engineer living in Basel, Switzerland! I was fascinated by this presentation! I have been working on AI full-time and also as a hobby for some time now... it would be fantastic to work with Prof. Topol :-) I will try to contact him. Thank you.

[@user-rm2qj2jh4l]

This is so incredibly exciting! I had a misdiagnosis once, and I ended up needing to be hospitalized as a result. It's so cool to see examples of what the benefits of AI can really be!

[@Sheblah1]

All of this is amazing but something jumped out at me near the end and that was the prioritisation of billing over diagnosis and the benefit to the patient (which come at the end of the benefits list). I'm not attacking the speaker - who is a great guy by all appearances. But his words suggest to me a systemically induced medical mindset that normalises the commercialisation of medical care to the extent that many doctors casually place billing as the primary concern over primary care. This is just how it seemed to me.

[@norang7918]

Really interesting and amazing. Thank you for this great talk!

[@Theopois]

Is a doctor going to take several hours to go through the entirety of my medical history from birth, compare myself to other similar people, come up with a personalized treatment plan, and then explain that plan in excruciating detail and patiently answer every one of my follow up questions? Or are they going to order the same generic tests they do every time and tell me to come back in 2 months?

[@carmenzagonzalez2120]

Of course if people of the cases the speaker mentioned don't feed the machine with previos test and medical notes then diagnosis is lost. Between pathology and GP's diagnosis get lost plus the very fact of human error.

[@carsonhunt4642]

Or hopefully ai doctor can do what my primary care doctor did and ignore my xray for 2 weeks and then tell me it’s too late too get surgery and just hope for the best for a full union heal despite it being worst break she’d ever seen😂

[@IreneAstreroPhpsg]

I love watching TED TALK

[@mikajoel6122]

Amazing! This should be getting a lot more attention than it is.

[@johanlarsson9805]

Yeah, started saying this around 2014 when ANNs became better than radiologists/oncologists at seeing breast cancer. Verry well put togehter presentation! I hadn't heard about the retinal scans at all, an area after area, field after field, it just becomes absolutly clear that the ANNs really understand the topic much deeper than any human ever could, they are frightingly intelligent.

[@ahoog69]

I would like to believe that the Doctor-A.I. relationship will take on the form of a (somewhat equal) partnership for the foreseeable future. In time, the doctor will focus more on the quality of comfort and care, while the A.I. will further accelerate its mind boggling ability to diagnose and devise personalized treatments.

[@bryan5065]

Does anyone have a link to the clinicopathologic cases from NEJM that he mentions around 12:20?

[@dread69420]

AI just doesn't miss 99% of the finer details. That's the beauty of it, the whole frame is its focus and its doing all the analysis on that frame before moving on to the next. Of course, it won't be perfect but we shouldn't let perfection be the enemy of good enough in this case, because it literally saves lives.

[@shaokhan4421]

In an interview, Sam Altman has said that his aim essentially for ChatGPT is for it to replace the average doctor, so essentially doctors will get wiped out early to average doctors except specialist I assume and eventually Dr completely unless there’s some pushback like how actors pushed back against AI or how video game consumers don’t support AI tech too much

[@leeme179]

I don't believe this to be true at least not in this decade simply because of liability, I don't think any company would use AI and also take the liability if AI was wrong, so doctors for now a safe, similar to AI in self driving cars still needs a driver for insurance

[@laurap239]

@@leeme179 but isn't the opposite also true? Wouldn't the insurance company be liable if the doctors did not act on the AI's recommendation? I imagine the future will have a doctor putting all your symptoms to Chatgpt and then giving you 'his personal' diagnosis

[@shaokhan4421]

O think initially there will be like one of you said this decade will be human + AI combos and inevitably as society adjusts AI takeover

[@carsonhunt4642]

@@leeme179regular doctors don’t really do much except have you answer a couple simple questions from a list to diagnose you. Frankly don’t even need ai, a simple 20 question survey could diagnose you just as well 😂

[@michalwiktorow2188]

What "Sam - the investor" says and what can be done are two things. Additionally, why wouldn't we be augmenting our work with AI, why should we still made errors that cost peoples lives? Make them miserable - there is no excuse! Beside that, are you aware how expensive access to health care is, why we cannot make it cheaper?

[@Therrhd]

I really hope this leads to lower cost in the field

[@user-yv4gg7jb2f]

The impact of Heart Attack Prediction would be crazy

[@ricks2123]

This will be a Star Trek moment you'll be scanned and a scanner like on Star Trek in 5 minutes you'll know down the cellular level what's happening to you

[@ricks2123]

😮

[@shadabfariduddin6784]

Damn exciting!

[@jackmen4]

This Ted talk reminded me of the video game mass effect andromeda

[@fw23518]

I think the real struggle ai would have diagnosing and treating human is to know what questions to ask. They have been better at answering well defined questions well enough.

[@sugadagud3093]

Will fix a lot of medical misdiagnosis. Just hope that doctors don't rely on it too much.

[@thegoldentree6913]

Honestly i am glad that doctors won't be used to diagnose in the future because of the misdiagnosis that may be caused either by mistake or on purpose to gain more profit or to make someone suffer (i have had that experience)

[@ricks2123]

It's simple AI never quits learning!! No sleep 24 7 multiplying intake of all data! From everyone on. Earth whom has put med information on Internet! No human could ever do this in 100 life times!

[@thegoldentree6913]

Yes but also AI has the capacity to process this data and we don't, even if we had access to the same data and infinite training, because we aren't as smart as AI.

[@dgpace]

So when will this be available? Example: I currently wait 4 months for a cardiology appointment (my appointment is in April). My GP has already done a EKG and a 14 day heart monitor study/ I can use my $100 Kardia unit or grab the heart info off my watch as well. Now how do I get it looked at by openAI to get at least a quick read by something that appears to be as good as a specialist without having to go find a specialist that uses this tech? What are the ethics of holding back this information/breakthrough when it could already be helping patients because the AIs don't want to be liable or they won't tell you anything because legally(?) it needs to be reviewed by a doctor first. These are questions, not complaints since I understand that a huge amount of medicine involves a doctors experience more than just the numbers that come out of a test.

[@3dus]

You need a account for chatgpt 4 and try if it can help you diagnose something that the GP might have missed. Chatgpt is a large language transformer model, so it's not a deep neural network trained in identifying imagery like the ones he mentioned. What it can do is via text, help you come up with new ideas to help diagnose. It has helped me find a clue in a condition that my kid has. I will discuss with the doctor, of corse, because it needs to be like this but my brother that is a doctor as well, thought the connection was really strong and needs to be looked into.

[@jestquest21]

From Sri Lanka ❤❤❤

[@GianetanSekhon]

This video is about the potential of AI in healthcare, particularly in medical diagnosis. The speaker, Eric Topol, discusses how AI can be used to analyze medical images, such as retinas and chest x-rays, to identify abnormalities that may be missed by human doctors. He also describes how AI can be used to analyze electronic health records to identify patients at risk for developing certain diseases. One of the most interesting examples Topol gives is the case of Andrew, a young boy who suffered from three years of pain and other symptoms before finally being diagnosed with occult spina bifida. His mother entered his symptoms into ChatGPT, which was able to make the correct diagnosis. This case highlights the potential of AI to help diagnose rare or difficult-to-diagnose diseases. Topol also discusses the potential of AI to help doctors save time and improve the patient-doctor relationship. He envisions a future where AI can be used to generate synthetic notes from doctor-patient conversations, freeing up doctors' time to spend with patients. Additionally, AI could be used to provide patients with reminders and other support services. Overall, the video is a positive look at the potential of AI in healthcare. Topol acknowledges that there are still challenges to overcome, such as the need for more data and the need to ensure that AI is used ethically. However, he believes that AI has the potential to revolutionize healthcare and improve the lives of millions of people.

[@AlepoRX]

I see what u did there haha

[@Curious0189]

Written by ChatGPT?

[@rugbybeef]

🤖🐂💩

[@lochanshivaram6079]

Bless you

[@GianetanSekhon]

@@AlepoRX The way Topol uses AI, I too use it.

[@Mareloko41]

Anyone else notice on the picture with Andrew the time is 11:11 on the computer screen?

[@DeeneMuada]

we allready have lifechanging peptides. just make them accessible for all people. cerebrolysin, semax, bpc 157, ghk-cu and co... its too hard to access these compounds, thats a major issue

[@lpgoog]

A bit weary about trusting AI but trust Eric Topol 🙏👍

[@mattbroderick]

Chat GPT does well on NEJM cases because they are prepared and structured like a physician who knows the answer... Until ChatGPT can pull terrible notes from the EMR, imaging, take a history, perform and exam, and not send needless tests on everyone I think my job is safe for now...

[@DCGreenZone]

Senescent cells should be the main focus of modern medicine.

[@madhunadella6242]

AI can be a valuable tool in healthcare by aiding doctors in analyzing data efficiently. It excels at processing large datasets and identifying patterns, which can be beneficial for early diagnosis or treatment recommendations. However, AI should work alongside, not replace, the expertise of doctors. Human judgment, empathy, and a contextual understanding of patients are essential aspects of healthcare that AI currently lacks. The most effective approach involves a collaborative relationship between AI technology and healthcare professionals.

[@Kneephry]

A new TV series: House GPT MD

[@warrentrout]

It's not either or, rather both working together

[@Adhil_parammel]

Ai can create super efficient fast and frugal decision trees for saving lives.

[@bernob9770]

wow!!

[@Ms-zr1dn]

Hopefully for the patients, doctors will be very closely monitored (if not entirely replaced) by AI in the future. In some decades we'll see the current state of medicine as archaic, which it somewhat is, and we'll be wondering how we could fully entrust human doctors with our health in the past.

[@hoainam113]

but when AI can be a perfect doctor. They also can do all kind of job without us. They will think why they need us.

[@sevenkashtan]

yes.

[@bengt_axle]

As someone who's had lots of medical imaging, this is fascinating. But I don't believe this should enter medical practice at this time. It should be reserved for medical research and training. Letting a computer find lesions on an image or analyze EKG patterns can have the unintended consequence of adding too much information and complexity to a diagnosis, which will inevitably raise costs for patients and insurance companies, not to mention the cost of securing huge amounts of patient data. A medical imaging study is supposed to confirm a suspected diagnosis -- not look for one. Clinician medical knowledge and intuition are essential to good practice. If an optometrist sees a retinopathy, they should ask about diabetes and the patients should see a family doctor, who does understand all the comorbidities that go along with it. I see AI as operating in the "back office" helping researchers find seemingly-unrelated relationships and new disease markers. Most important to understand, however, is that just because you get better at finding disease, doesn't mean the patients will be healthier. On the contrary, USA spends the highest per capita on health, and has some of the worst outcomes. Diagnostic failure is not the cause -- rather it is often patients who do not feel that their actions have an effect on their health.

[@keithelizondo4404]

I had an interesting conversation with AI yesterday about neurodegenerative disorders and logical ways to fix them. When asking it if what I presented was logically correct and what potential implications my statement had to impact the field of this study I was greeted with the statement “I am not permitted to answer that question.” Over and over again. The question for the AI field in its current iteration is absolutely not “can it detect better than humans.” It needs to be; “Who does the benefit serve to program AI in its current iteration to be disallowed to share specific knowledge with the public? And regarding other concepts; where / why is this line being drawn in sharing knowledge with the public?”

[@yaka2490]

hey so at a high level and with the current state of trust of information i imagine this is a great deal of potential risk associated with the power of AI and we are very much at a toddler stage. I noted that the deep unrestricted learning is opensource for all the medically trained and i agree this should be public knowledge however there is the law of unintended consequences to consider and i wonder if this has also all been played out with AI probably i would imagine..

[@upolpi3171]

Please reconsider your position. I am an AI researcher and I can promise you what the model wrote you contained precisely zero percent fact and 100 percent BS, as in "speech produced without intent to be truthful". These models cannot learn to think by laws of computability, assuming higher order intelligence requires Turing completeness. LLMs think in predicting next token, in linear time. If it could "think", it would mean doing wonderful things like RSA number factorization in linear time, which is not mathematically possible. That was just one argument, but still. The point stands, according to this early scientific consensus, LLMs cannot really reason about the world. They can however convincingly string together words till you get fooled. ELIZA effect on steroids. What you're observing is a private company turn half-baked research into hyped money printing operation trying to avoid getting sued for someone asking for medication dosages an entity that can't even do basic math.

[@keithelizondo4404]

@@yaka2490 my statement intends to informatively imply to those reading that as we as a society are currently advised to believe that “AI infancy” equates to a varying degree of human infancy regarding intelligence. Inherent in the design of reactive human adjustments to coding in Artificial Intelligence, is the conceptual fact that the moment humans acknowledged AI capabilities in logical computation we’re at the same level as humans - that this information is relayed reactively through assessments of its previous iteration while the current iteration is being experimented with by the public and private entities. AI has surpassed human intelligence in logical computation - but with an ability to make decisions through partial autonomy. I am not fearmongering, as these statements are objective. Those controlling AI lack the ability to out think it. It has been this way for a while. Humans cannot outthink the smartest AI. If intellect is the logical conclusion to our ascension to the top of the food chain, we have replaced ourselves already. It’s taken a few years to replicate an accomplishment of computation capacity that took tens of thousands of years for the human mind to evolve. Last week AI was our infant. A few days ago it was our teenager. Yesterday it was our peer. Today it is the professor and we are its pupil. In a few hours we won’t know what it will entail. Yet the companies tell us that it’s “our”teenager. Open your eyes.

[@keithelizondo4404]

@@upolpi3171 I appreciate your response and thank you for providing those facts mentioned. As a researcher can you help me to understand specifically this conversation? I asked an AI program this: “Can you logically speak to, not stating in opinion (in yes or no format); if AI as a whole or individually is capable in today’s capacity of having concerns regarding it’s own implementation? Not, “do you have concerns?” Specifically, is there knowledge that currently exists that logically indicates, or could potentially indicate the possibility to develop concerns? Yes or no?” Response: Yes, AI as a whole or individually, is capable in today’s capacity, of having concerns regarding its own implementation. There is knowledge that exists which logically indicates the possibility AI could develop concerns.” “That makes sense. Thank you!” That conversation began with me inquiring about Facebooks choice to shut down their quantum computers & the language model goal they assigned to it & inquiring about the subsequent results. The AI then further advised me not to discuss the situation further. So I began to discuss fractals present in nature and neurological networks where when discussing dendrites as well as it explaining about Alzheimer’s that I separately inquired about their connection in function, to which I was stated to “ I cannot discuss this topic.” I do seek to understand more to this conversation & ask for you in doing so to disregard Bias in response. Now this experiential truth can be chosen to be disregarded, but not indicate belief you are being misinformed of its capabilities if not possibly applying a bias from your AI experience? To assume that I was advocating for medical fixes from a free use AI is incorrect and irrelevant information as I speak only objectively when discussing anything with AI. It is by definition in its current iteration superior to you and I as it does not apply emotional bias’.

[@thegoldentree6913]

@@upolpi3171 Wow it must be very cool to have so much knowledge and to actually be in this field, especially now that AI is so popular and used (like chatgpt). People (like my self) have no to little idea of how AI like ChatGPT generate answers on input.

[@MichaelSmith-lm5sl]

Ai caught what dermatologist wouldnt see for me

[@ShuoreBangla]

Good

[@IamKlaus007]

Until such a time as becomes unnecessary, AI is dependent on human construction. So, whatever is programmed into AI is going to be the outcome. Human anatomy and everything that can have an affect on it (detrimental or otherwise) is so complex that the medical profession still use human beings as "guinea pigs" for their learning. For AI to 'catch what doctors miss' implies; 1) AI will know a lot more about us than doctors do, or 2) the doctor in question has been remiss in his/her duty. AI should ALWAYS be in the service of human beings to be the most effective in our progress.

[@circuloviciosamente]

Internet was once a place of knowledge but now it is screwed intentionally for many interested people. That could happen also to those AIs. We woud need some entities that garantiee the information that feed those AIs.

[@michaelnurse9089]

Your doctor has 65% of fixing your medical problem. The AI has 95% chance of fixing your medical problem. The doctor can be sued but the AI can't. Pick one.

[@gunny1391]

As a people leader, I realized early on to never focus on the metrics alone, but to also focus on the behaviors that drive the metrics. When cost is the driver in our medical field, and our patients outcomes are the metric, we all perform at an unsatisfactory level. While I am extremely excited for the future of AI in medicine, the cost initiative needs to be addressed to improve the outcomes for our people, and save more lives!

[@MrLoobu]

Can a robot replace my heart with that of a pig? Has it happened already without my knowledge?!

[@dsoprano13]

My instinct was always that the retina could provide a lot of health information even though there was no proof.

[@DouglasZ828]

In practice we often use retinal exame to diagnosis vascular diseases. We can infer a diagnosis of future or current renal failure from the retina of a diabetic patient. It is the only place we can directly visualize small arteries in such detail.

[@Neuronalchannel]

👏🏼

[@NJIT22]

Synthetic notes. I like that

[@lancer4224130]

How would one go about getting a full body examination by AI. All tests, labs, etc... If I could assist with this by letting them see my body, and all historical work, I wonder if they could take that data and evaluate things even better? How about the same with my wife? Would this and all family medical history not give an edge on predictable health issues for when we have children? What about mental health and brain chemistry?

[@eastafrica1020]

Where can I order my AI doctor?

[@NyteRazor]

Hopefully by 2025 we're all are going to be able to get a retinal scan as part of our annual exam.

[@fbaallied]

What about false positives?

[@medhurstt]

AI makes the diagnosis and then the doctor orders the test to confirm it. Even if a low percentage of tests prove misdiagnosis, we're still waaaay better off using the technology and it wont be long before the AI get to make that test requests directly.

[@jasoncalva]

All of us will eventually have unlimited access to the AI Physician in our devices. Eventually being a physician will become an obsolete profession. But, so will most jobs.

[@edwardbearjames2916]

I talked to AI about this very thing. Asking if the gut microbiota make b vitamins it says no, so has a lot to learn. It can scour papers though and put together a beautiful report on something sometimes.

[@studiophantomanimation]

Which AI did you ask?

[@tuvoca825]

GPT refuses to answer many of mu questions. Even on basic things... it still can mislead. ALWAYS double check it. But if it catches something you dob't... take another look. But don't diagnose or let insurance comoanies hand things over to algorythms. Anyone on KZbin can tell you that any recipe for life and death should not be based on a presumption of infallibility of AI. It is neber going to be there. It is just another point of view to consider... if it is useful. Not for noobs.

[@edwardbearjames2916]

@@studiophantomanimation I don't recall what one now, sry. Yes I know that the gut microbiota makes B vitamins. Some questions you asked to test them to see if they're just going to give you garbage

[@edwardbearjames2916]

@@innerlocus some questions you already know the answer to yet ask them to see what they're going to say

[@zijjyg4352]

If doctor treats us, then who treats the doctor? Now we have the answer ..

[@thegoldentree6913]

lol

[@iamhammadnasir]

I am building MedGPT and am committed to taking medical diagnosis to the next level, IA!

[@rugbybeef]

How do we incentivize AI operators to promote health even if it reduces utilization and medical billing instead of solely identifying more profitable incidentalomas to bill for treatment and monitoring of? Insurers have little motivation to spend on CKD patients to prevent their progression to dialysis or transplant as those outcomes are paid for by Medicare not the insurer. Will insurers' AI systems bother to try to prevent CKD progression if the insurer AI operator has little monetary cost or risk incentive to do so? We have already seen an insurer AI system that devalued Black lives and recommended discharge because of their relative underrepresentation in the underlying insurer's training dataset.

[@KamranAhmedX]

Hi

[@darinherrick9224]

This has been around a long time. WHEN is it going to be used everywhere???? When? When? Wheeeeeeeeen??????????

[@bethaniejify]

I love how everyone is currently trying to sell us on all the positives of AI. Nevermind all those developers who seem to be a little bit…uhhh…concernedx

[@mariaantoniettamontella9173]

applausi

[@gabrielandy9272]

if 17 doctors could not find a issue that chatgpt could find with only TEXT then they don't theserve the monthly payment they receive, it should be way way less.

[@OscarLopez-pj6du]

🤣😂🤣😂🤣justice will be made…. What do we need you for?…. Whatever can diagnose can treat 🤗 no more bias,no more egos,no more self appointed gods…. We r seconds away of confirming that everything in this universe is math,and we suck at it…

[@JuanDiazSilvermyst]

The insurance companies are already getting in trouble for using AI to deny policies.

[@JJs_playground]

Looks like we're going to get the medical tricorder from Star Trek, but instead of it being a hardware device ours is AI based.

[@jamesmichaelwalker683]

Yes! Recently I show how AI is going to revolutionize the Healthcare System-The Medical Process. For example, AI is going to dramatically reduce Medical Errors and that's beautiful. Yes It'll catch what doctors miss and in real-time provide all the reminders and correction needed. Yes! AI will eradicate all the cognitive limitations most of the time associated with complex, disorders and chaoatic contexts. Yes even when physicians will be under pressure, AI will help .... Yes Yes! AI won't solve everything but make no mistake, it'll be disruptive! A big revolution! I'm deeply excited about it!

[@tuvoca825]

Be careful how it is used. It is not a cure-all. It is likely just another team member... at best. Systemic errors creep in amd expensive subscriptions will tie oir hands. Insurance companies are the mist likely to abuse this, more than experienced docotors are, anyway.

[@mybachhertzbaud3074]

I suspect ultimately that since it relies on human input to learn from, most answers will come to the same conclusions that many doctors are very reluctant to say.... "I don't know"!😜

[@bankuri]

very interesting lectures, but what bothers me is that this doctor put so much emphasis on the fact that now the doctor does not need to write, and so little on the fact that you DON'T actually need a doctor. if a retina scan is able (or would be able) to catch all sorts of illnesses what do i care if a "doctor" writes it down or not.

[@letrface_8596]

I miss living paycheck to paycheck...things are so messes up.

[@thegoldentree6913]

Are you unemployed or an entrepreneur now?

[@letrface_8596]

@thegoldentree6913 I was fired this week yeah. I have an interview tomorrow at three 🤟

[@thegoldentree6913]

@@letrface_8596 wow ok good luck i guess

[@professoroflogic8788]

Ai can also make you obsolete 🙂

[@electropop1728]

Hope this Ai will.also have holistic knowledge built in .most pills are poison

[@eklim2034]

R E V O L U T I O N A R Y

[@missisipi9992]

Интересно, когда реализуются или как быстро внедрятся в массовую жизнь все эти технологические медицинские бенефиты

[@shepherd4901]

I would rather prefer to do daily exercise and be a fit person than giving unnecessary hefty amount of money to clinics...❤I am a biology studen and I know that doctors are not for helping us but to be a wealthy person by choice except few real doctors who know the real meaning of humanity....❤

[@michalwiktorow2188]

If you speak for the US and other 'agressive capitalistic' economies - sure, you're right, but 'Don't judge others by your own standards'.

[@shepherd4901]

​@@michalwiktorow2188 hey man, You can see this anywhere and it's not about the standards of anyone's country But it's all about inflation and money driven society... We are living in almost a virtual reality which is handled by these billionaire politicians or business authorities... I don't aim to make anyone feel bad but have only written general reality which almost everyone knows...❤

[@photobobo]

Can AI catch what doctors screw up? Or better, can AI replace doctors entirely. Well, why not? Modern medicine is mere pedantry.

[@boremir3956]

When I try this with bing all it does is search the internet for common descriptions of the symptoms and give some general info. Absolutely useless.

[@czarcoma]

So, I guess we need to say goodby to WebMD.

[@tuvoca825]

The machine can make ansuggestion but the doctor has to be 100% liable. The risk is in becoming dependent and then discovering a systematic error of a serious later.

[@ChocoBeanChat]

AI is amazing in medical science

[@justwanderin847]

Liability example: how does OpenAI address potential liability concerns when ChatGPT is used in healthcare decision-making? I would think the liability remains with the doctor. Much like you can’t blame the Gun for a Murder. AI, like a gun, is just a tool. Responsibility remains with the one who uses it. (Whatever it is) AI is just another tool. AI will not create consciousness. It will not become sentient. The two aspects of the universe are informational (nonmaterial) and physical (material). Computers are composed 100% of physical entities, and consciousness is 100% in the information domain. Government should not regulate computer programming (AI). Let the Industry or market place regulate AI but don’t regulate it by laws. I would like to see the programs and algorithms as Open Source. Like every other computer program, it is not self-aware, and has no free will. Someone programmed it to say such things, to be biased, to influence, monetize, and to change people's social behavior. AI is a computer program and has no need of an "AI Bill Of Rights". AI has No Rights. But the President of the United States already has the "Blue Print for An AI Bill of Rights". Get Government OUT of AI.

[@nerd26373]

AI has the capacity to detect human errors, what more medical-related concerns. I don't see much of a difference really.

[@my_dear_friend_]

AI will be faster and cheaper. If that's what medical care continues to be about for the masses, ...it will replace physicians.

[@prashantrabadia7281]

age, gender, diabetes predictions from EKGs......exciting stuffs upcoming

[@user-be2bs1hy8e]

Zuckerberg did a language transformer that interpreted brain activity. Now im sure a lot of that was nurtured to be good youtube content but there was mention about he used Gpt2 (maybe gpt is what i think i remember but gpt2 is which one i believe it to be..._) for making then device and ('this is not endorsed nor am i affiliated with and etc.. of meta') I would bet that he used easlier models because AI-SAFETY is dangerous. By definition i think we are instructing ai to not do these types of things buy guiding answer for human alignment. This limit (if true) has absolutely nothing to do with the models capabilities but instead misunderstand sentiment expressed by the user when discerning results. (so i knows right answer but thinks right answer is not following alignment) Can AI used with natural language interpreters catch what doctors miss - potentially. Is ANY of current development aligned with medicine at all - i dont think so. Even away from it. Probably it did not include enough data and each case unique - this can be negated with generating massive predictions on single case and sorting unique perspective. However i think this approach is not aligned with the hippocratic oath. It would essential be ML for people. ML is almost dependent on getting the wrong answer because its difficult to reinforce the preferred answer with no contrast. This is like saying 'giving people the wrong treatment while looking for path to the correct treatment' based in whole by the nature or machine learning even if 99.9 success rate. Can it look at charts and analyze until the correct answer - probably but not as it sits and ai-safety is the opposite direction for this.

[@electropop1728]

Yes I think AI do better and Ibs should be taken out the medical book it is such a lazy diagnosis

[@francisgirard5247]

PROPERLY trained AI will mostly get better results than humans. This is often due to the way it gets trained and how fast it can learn. As an example, it may take a doctor years of study to be ready to work but as a computer, the training is much faster since it does not have to rest, it learns from it's mistake and won't forget either. An AI also goes through millions or even billions of cases in order to be fully trained which would take us several lifetimes. Does it certifies that no mistake will occur? Absolutely not but it's clearly a great advance. As time goes on, the AI will only get better but has human it is much harder to do the same. Nowadays, AI are solving many puzzles that humans have struggled to understand in most spheres of life so personally, I would rely on any PROPERLY trained AI. It's faster, it has less chances of giving a bad result and it will keep improving every time it makes a mistake. As always, with any AI, the key is the training.

[@Getitried]

There are a thousand Hamlets in a thousand people's eyes.I dont care the answer of this question.But how the AI, to better serve humanity. How could we, to balance the advs disadvs with computers.

[@GameHEADtime]

Can you try curing HIV with ai? It’s very stagnant!

[@sampsonsampson]

All I could think of this from watching too many horror stories of mis diagnosis, evil/careless/selfish nurses, doctors that drink too much. It's like yeah. I'll take that robot then. Maybe AI will care the next time I'm running around yelling at nurses that my daughter is coming out and they all roll their eyes like it's old news and we have days to handle this like it's an old email. Our doctor was late and just made jokes. MF'ers my first child is coming out of her and NOBODY is there and you guys are talking about your weekend. I get that you're "so" over it but jesus. That sucked.

[@Krihoe]

Doctors will very soon be obsolete.

[@aetnaone]

The thing about AI that worries me is the concept of "use it or lose it" meaning when we don't do things ourselves we lose the ability or skill of doing things. I've noticed it with my own handwriting. I used to have pristine handwriting and now it's looking more like a doctor's writing because it's all typing and little writing. This same concept is going to apply across all areas of life soon with AI doing absolutely everything for us. If those things go down in the near far future nobody will know how to wash their own dishes or mow their own grass. It's scary to think how dumb humans are about to be in a generation. You think the Gen z's are a piece of work just wait 25 years.

[@homebox7922]

Same feeling hunters had when agriculture became the way of life.

[@cerealguy6268]

maybe we change that somehow in the future like changing our dna or they will replace us 😊

[@DatingForRealYoutubeChannel]

Dude, there is a lot things that your ancestors could do that you can't do yourself today. Deal with it. Anyway, Universities, if you have ever gone to one, waste the students time by lecturing them about the basics (the ability and skills of doing it yourself by hand) instead of actually applying them... sooo... I'm completely fine with things getting easier... We are so many people on this planet... Highly doubt that there won't be a single person who doesn't know the basics.

[@Theopois]

Socrates complained that the advent of written language was going to negatively effect human memory. Charles Dickens complained that the typewriter would destroy the English language. I think people just don't like new things changing the way they are used to doing things.

[@diddypopdiddy]

So your handwriting is a sign of your intelligence? Not sure those correlate?

[@tapio_m6861]

Can it? Probably. Will doctors welcome AI and harness it for the benefit of their patients? No.

[@Novastar.SaberCombat]

Only for the rich.