- 8 minutes ago
At the India Today AI Summit 2026, experts explored how AI is reshaping hospitals, diagnostics, and patient care. They emphasised that technology can improve access and efficiency but cannot replace human responsibility.
Category
🗞
NewsTranscript
00:00All right. Thank you so much. And I think I'd like to begin with the most basic question, Dr. Treyahan.
00:07Guilty as I plead, but for a lot of us now, ChatGPT is becoming the first go-to doctor.
00:15And any initial symptoms, the first thing we do is click on ChatGPT, put it all out, ask whether we
00:22should go to a doctor or what medicines we should take. Really wanting to hear from you on this, the
00:27word that you put out for everyone listening here.
00:31So there are multiple aspects to this. One, of course, it translates into so-cho-mat, you have some feedback.
00:44Okay. The danger of that, the other aspect is that you get so involved in the possibilities of what you
00:56may have that the anxiety levels go up so high and then till you have found the right help for
01:04it.
01:05The third is that it is a good tool to have at your disposal to say, yes, you need to
01:15go to a doctor or you need something.
01:18Okay. So if you actually just did not have any feedback, you may just console yourself, let me see how
01:27it would go away or what happens.
01:29And especially in our field with cardiac, you know, so many people lose their lives just thinking it must be
01:37acidity, maybe, oh, it can't be happening to me.
01:41I've got too many responsibilities. So there are many, many things that make you lose that urgency.
01:47And I think a tool like ChatGPT, which was Google was already there, is a ready reference for your symptoms
01:56and you have a friend in that.
01:57So I think it's a very positive thing to happen. But at the same time, it can trigger anxiety reactions
02:03which are unfounded till you find the right help.
02:08And it is an AI tool at the end of the day. It hugely depends on the prompts that you're
02:13giving it, possibility of it also guiding you in the wrong manner.
02:22I think this is something that we all know, right? It's that the AI tool is based on the data
02:27that it gives and the data that you have fed it.
02:29So, for instance, I've got a long ChatGPT thread just for my mother. I have input all of the information
02:37of her blood tests, what happens with her appointments,
02:40what's going on. And I've also built out the personality of my chat specifically to look at case studies, to
02:47look at studies of the best cardiologists in the world,
02:51to think about diabetes, to think about high blood pressure. So the more data that you give it, and we're
02:56seeing that with a number of AI tools,
02:59the more it can inform not just your own opinion, which we see that reinforces in a number of tools,
03:06but it can actually give you at least real-time options, which otherwise you can't always contact a doctor in
03:12the middle of the night.
03:13And so far, at least it's been a component of the thinking. I don't think it should be everything at
03:20this stage,
03:21but for me, it's been a component that I can keep building out and feeding information to,
03:26so it can be more thorough and more effective than it would be just as a random AI tool.
03:34And, you know, Vivek, sometimes for doctors, it can get mechanical to answer too many questions.
03:40And there are also those who I know who go on to say that when I put the same technical
03:47terms that the doctor has spelled out for me on ChatGPD,
03:50it helps me understand much better what the problem is, because I can ask it 10 questions, and it breaks
03:55it down really easy.
03:57Absolutely. Yeah, definitely that is a great tool to have.
04:00And when you mentioned that specific to a doctor, what happens?
04:08These tools will actually save a lot of time for the doctors and for hospitals itself.
04:14We strongly believe that there is a lot of inefficiency sitting in healthcare systems today.
04:20About 60% of inefficiency sitting in healthcare systems today.
04:27India has one of the lower doctor-to-population ratio and beds-to-population ratio when you compare it with
04:34the rest of the world,
04:36which means you would expect Indian hospitals to be brimming with occupancy,
04:40given that there is a supply-demand gap.
04:43But surprisingly, when you walk into private hospitals, you see that occupancy is not that high.
04:50Private hospitals in India is somewhere at 60% occupancy,
04:54which is a manifestation of the affordability gap, of the accessibility gap, of trust that is there.
05:00Given that occupancy of private hospitals is on the lower side,
05:04you would expect that throughput is extremely high.
05:08But you know what we see in every hospital?
05:10You see people sitting and waiting in different areas of the hospital,
05:14which is a manifestation of the underlying inefficiency in the healthcare system.
05:18And that is where we see chat GPT or LLMs or AI as a broader category playing a huge role
05:26in eliminating that inefficiency, which is basically data movement, data summarization,
05:32answering things from data, and providing suggestions to people in a timely manner
05:38to be able to eliminate all of that.
05:40All right, my colleague Sonal also joining us.
05:42Well, somebody was missing from the panel.
05:45Well, it's not me. It's AI, isn't it?
05:48Right?
05:49I'm not missing at all. We've been having a lot of conversations.
05:51So I thought it would be a good idea to ask AI a question, no, on health.
05:56So on my way here, and by the way, one thing that AI should solve for is traffic.
06:01To all those people sitting here, please solve for traffic before anything else.
06:06But on my way, I asked chat GPT this question, which I'm sure all of you want to know as
06:11well.
06:12I typed in and I said, hey GPT, Elon Musk says that AI will replace doctors.
06:19What do you think?
06:21I'll let the chatbot give me an answer and then I'll come to the humans.
06:26Okay?
06:27You want to know what chat GPT said? Yeah? Okay.
06:30It said, AI will dramatically change medicine, but full replacement of doctors is highly unlikely in the near future.
06:40It already performs very well in narrow tasks like reading scans and analyzing large data sets.
06:47However, medicine is not just diagnosis.
06:50It involves judgment, ethics, empathy and accountability.
06:56Patients trust humans with life-altering decisions and that cannot be automated easily.
07:03More realistically, doctors who use AI will replace doctors who do not.
07:10I have to say, we have to give a round of applause to GPT.
07:13That's a very good answer.
07:16Not only applause, it's very reassuring to us.
07:20But the last line, Dr. Trihan, it says, doctors who use AI will replace doctors who do not.
07:27That's correct.
07:28That is absolutely correct if you look at history of how evolution has taken place in medicine, technology and the
07:37use of AI.
07:38AI is not new to medicine, by the way, because we've been using algorithms, machine learning for many, many years.
07:46Robots were created in the late 1970-something.
07:52We brought the first robot to India in 2002.
07:56So, you know, all this stuff has progressively gone in a direction where it's true that if you do not
08:04use...
08:05See, errors are known.
08:07Medical errors are known.
08:09It is known that 100,000-plus lives are lost in the United States, which is supposed to be a
08:16well-oiled health care system.
08:19Okay.
08:21I don't...
08:22I'd try to even think if human errors, how many people would have lost their lives in India.
08:27Okay.
08:28So, what you're saying is over-dependence on AI has the same danger as over-dependence on yourself.
08:39See, somebody said very rightly, the illiterates of the world are not going to be people who have gone to
08:49college and are professors.
08:51The illiterates of the world are going to be those people who cannot unlearn and learn again.
08:58So, it says it all.
09:00As far as...
09:01This is the new wave.
09:02If you didn't learn internet, if you didn't learn...
09:06I mean, it's so convenient to use a cell phone.
09:09So, that was the second revolution.
09:11That's very nicely said.
09:12There are also several gaps, Dr. Triyan.
09:14And I want Vivek to get it on this one.
09:17The biggest problem with India at the moment is not that we lack AI innovation.
09:22It is that we lack unaffordability.
09:25Right?
09:26We are trying to bridge that gap.
09:29With what you are doing at Narayan Hospital, have you managed to use AI to cut down cost?
09:34Because that's the key question here.
09:37Absolutely.
09:38So, fantastic question, first of all.
09:41And that is one of the main reasons why AI remains in pilot stages in many institutions.
09:47It is because you are not thinking of return on investment when you are investing into an AI project.
09:54Because finally, at the end of the day, who pays for this AI?
09:57That is the question.
09:58And the person who pays for this AI, where ROI is not clear, in which case it is a gamble.
10:04So, the person who pays for this AI gamble is the patient.
10:08Can the patient afford to pay for that gamble?
10:10That is the question.
10:11Which means, if you are trying out fancy AI projects one after another, where your end objective and your path
10:18to the objective is not clear,
10:19they are most likely going to fail.
10:21Because you are going to have to increase your costs to the patient.
10:24So, it is extremely pertinent that the institution that is implementing AI have a clear objective on what this is
10:31going to solve.
10:32And not that it is a fancy toy that I get to play with.
10:35Once your objective and purpose is clear, and your route to ROI is clear, which is basically the financial benefit
10:41that it is going to deliver,
10:43then you will be able to deliver it at the right cost for the patient.
10:46Yeah.
10:47But Anjali…
10:48Also, one thing to understand from the point of view of surgeries, because there has been a lot of talk,
10:52that when it comes to surgeries in the future, would largely now be taken over by robotics, machine…
10:59That's the tricky part, Nuswesha.
11:01India is not one India.
11:02India is one, two, and three.
11:04In India, one, in people like us, in societies like these, we talk about robotic surgery, we talk about affordability.
11:12In other parts of India, the point is of access, which is what I wanted to come to Anjali for.
11:17So, how is that shift going to be?
11:18Yeah.
11:18How will you manage that?
11:20A lot of AI innovation seems to be working for affluent India.
11:24But what about access where it's really required?
11:27I think that works for AI across the Global South, and this is what the AI Summit is about, right?
11:34So, for the first time, you have a Global Summit focused on impact, led by the Global South,
11:40and specifically looking at access and looking at equity.
11:43The previous panels have talked about this, but I think it's important to remember that low-bandwidth environments
11:50is what we should be designing AI for.
11:53And that needs to be intentional design in thinking about vernacular languages, voice-operated apps.
12:00How do you think about low-cost in the apps actually landing in these places?
12:05And all of that can only happen in the beginnings of the stages, right?
12:10So, what you're seeing in India and a number of other countries is that you've got silos in products and
12:16innovations that are happening,
12:17and they haven't, from their initial stages in terms of healthcare, aligned with the government's national architecture of health.
12:26And if you don't do that intentionally from the beginning, what you end up doing is then just bolting on
12:33afterwards to the health system
12:35instead of building an interoperability within that health system.
12:39That's what addresses your low-resource environment, but it also addresses where AI needs to hit, which is the community
12:47level.
12:48The community level is not at your tertiary hospital.
12:50It's at the community health clinics, and they need an expansion of access to healthcare and to human resources, which
12:57AI should do.
12:58Yeah, but AI is also all about accuracy, right?
13:02And Dr. Geeta, coming on this point, accuracy is only as good as the data.
13:08How far are we on the India data set?
13:10And I want to take that question to Dr. Trihan later as well.
13:13Is he comfortable even applying, or even Vivek for that matter, applying those same AI tools,
13:18which are not built on Indian genome body types, it's not built for us, it's largely resting in the West?
13:25Yeah, so before I answer the training set required for the accuracy,
13:30I would like to make a small comment about the previous question as well, if you know me.
13:35So essentially, how do we enable AI-enabled tools to make the underserved population get to the higher infrastructure at
13:47the right time?
13:47And that, I think, is the profound use case that can help Indian economy,
13:53because I really believe there is a healthcare divide in India, just like we used to talk about IT divide.
13:59There is a healthcare, really, there is a healthcare divide from access and so on.
14:03And AI can be a super tool for this.
14:05And for example, at Viramai, we've actually made this possible with AI-based triaging for breast cancer screening, right?
14:14So where women who were earlier detecting cancer by hand are now able to come into the hospital for more
14:22detailed diagnostics,
14:24just with AI kind of automated, I would say, for a red, yellow, green, by just taking a few pictures.
14:31So it's really possible to do this.
14:33Now, coming to the data problem, I sort of flip it the other way and say,
14:38as a founder or developer of the AI models, right, we have to foresee this, right?
14:43Where is it going to be used?
14:45Have I trained the model sufficiently enough so that the intended use cases, intended areas,
14:52targeted users are able to use this accurately?
14:56And I think Vivekji mentioned nicely that if there is an error in AI, who is paying for it?
15:02The end user.
15:03So if I wear the founder hat or AI developer hat, we have to look at the user and develop
15:10it for her, right?
15:12So that our accuracy will be much better for the targeted user.
15:17And what that means is we have to train our models with diverse data sets, whether it is there in
15:24the current AI data stack or not,
15:26in the hospitals or not, it's our duty to actually ensure we get to the data.
15:30Again, in our case, we really had to look at all of these 5 million data points that we had
15:36and collect the data.
15:38It took me two years to collect 256 patients' data of thermal images, mammography, ultrasound, biopsy.
15:47And later, of course, it became as much before.
15:48Ma'am, but you're talking about usage.
15:50The question is also about responsibility.
15:52Exactly.
15:53If the algo gets it wrong, who's responsible?
15:56Is it the doctor?
15:57I think most importantly in the question of mental health, because these days a lot of people are depending on
16:02chat GPT to also be their shrink.
16:04And two points really here, there have been cases where chat GPT has led them in the wrong direction.
16:10And there have been, unfortunately, cases of suicide.
16:13Secondly, again, the biggest point that we've been discussing since morning of how much of you is getting mapped by
16:20AI
16:20when you're literally sharing everything about your mental health with it.
16:24I think we should not depend on chat GPT for diagnosis for sure, like Trihan said mentioned.
16:30Trihan will have other stories to tell about patients who enter his clinic.
16:35Yeah, we need to make sure that AI is sufficiently clinically validated and regulatory cleared to be a medical device.
16:42Otherwise, you know, taking advice from it is really not worth it.
16:47But Zuesha, we've been speaking so much about AI and data sets and all of it.
16:51I want to talk about AI of the everyday and AI of the now right now.
16:55Dr. Trihan, I'm guilty of being one of them, but I have to ask you, are you also bombarded with
17:01patients who come to your clinic now
17:04with detailed diagnosis from GPT and Gemini?
17:08We started this conversation with exactly that question.
17:11But I want to know if they have an opposite diagnosis from what you're saying.
17:15How do you respond?
17:16No, see, it is a very good thing to happen.
17:20People used to depend on Google.
17:22Now it's chat GPT.
17:24And you always like to know a well-informed patient or their family.
17:31It's not because most patients are a little older.
17:33They don't use that much, but they come with their kids who have actually mastered the art.
17:38So it's actually, in a way, some people think it's a challenge or a headache.
17:44It's not.
17:45It is something that stimulates you to be able to mull through all the diagnosis yourself
17:51and also explain to them in detail that if a family is caring and intelligent about what they are heading
18:01for,
18:01it's much better from our point of view.
18:03And when it comes to mental health?
18:04Mental health, you know, this is something, a very tricky subject.
18:09Like Geeta said, you don't outsource your mental health to chat GPT.
18:16If you have anxiety, then maybe it's a good tool to talk to.
18:23You know, people, you can have a full conversation.
18:26But I don't think you should go as far as treating it like your therapist.
18:32It's not equipped to do it.
18:34Because of the fact, when somebody pointed out that, look, all the data points are, which are today current,
18:41only X-rays, mammograms are done in India, the data that has laid on a layer of accuracy on,
18:52like if you look at chest X-rays, which are the most basic X-rays done, that the AI tools
19:00have actually detected nodules,
19:02which are way, way beyond the human eye.
19:06Okay?
19:07So that is, I think, a great contribution to a point.
19:10Now, what are the dangers of that?
19:13Say, like, if you have a nodule, then you say, okay, what do I do?
19:18I must go to a doctor.
19:19Then the doctor says, I'll do a biopsy.
19:21And then they've taken you to a different loop completely, where it turns out to be nothing.
19:28So you have to lay the experience on it.
19:32That's why one of the things is that doctor's experience will tell you that if you detect something,
19:37the first thing to do is revisit it six months and a year later.
19:42If it is changing in size or character, of course, you must act.
19:47If it is not, if it is a five or two, one or two millimeter nodule that was detected,
19:53you should watch it rather than go and do an operation for it.
19:56Okay?
19:57That's, those are the responsibilities of how you're going to use AI.
20:02For doctors, that's why we talked about the doctor who is not AI enabled.
20:07Yeah.
20:08Will be always at a disadvantage.
20:11You will not ever reach the point of precision with AI is a real good tool to achieve.
20:20We have last few minutes.
20:21I have to ask you this and we'll quickly go around the panel as well.
20:24Do you know that in China at the moment, there are vending machines where they put a doctor inside
20:29and that doctor is supposed to answer all those questions and give you the drugs instantly.
20:34That is one extreme.
20:36But on the other side, what is the one thing?
20:38And let's begin with Dr. Gita there.
20:40What is the one thing you really hope AI will solve for in this lifetime?
20:45I'm not talking about future possibilities.
20:47If AI were to solve for one thing medically in India at the moment, what would it be?
20:54It's bridging the healthcare divide, right?
20:56I really feel that preventative care, you know, it's an amazing tool to have and not just automated.
21:03It should be in the hands of a health worker at least because, you know, it's really pathetic when you
21:09go to the rural areas and they don't have a doctor to go to.
21:12And get the healthcare worker a little bit trained on how to use the tool and really if you can
21:17get the affordable, high quality healthcare to everyone, I think that's something we should definitely aim for.
21:24Everybody talks about AI with jobs.
21:26Doctors don't have to be worried.
21:27We are anyways at a, you know, we need more doctors.
21:31Is there a way in which AI can solve for that gap?
21:34Absolutely.
21:35Absolutely.
21:36It will make doctors extremely more productive.
21:38Not just doctors but the healthcare infrastructure that we have as a country.
21:42Even after all of these years, we spend very little on healthcare as a percentage of GDP and there is
21:47a lot of ground for us to cover.
21:49AI will help us cover that ground extremely quickly.
21:52And finally about AI replacing doctors, that is not going to happen in the near future because as it was
21:57pointed out in the panel earlier, we all want accountability to be there with a human and not with a
22:03computer program that is not visible to us and we don't know who has developed it.
22:06In fact, I quickly want to share a recent example of how hospitals are now slowly going to have AI
22:12receive the first call instead of a receptionist, hear out your symptoms and then suggest a doctor fix up that
22:20appointment and then the patient will come according to what AI suggests is the doctor.
22:25So if it's a skin problem, then it's a dermat.
22:28Again, hospitals are headed that way.
22:30How safe?
22:31Not safe?
22:32We are really treading that path.
22:33Angela, you want to take that?
22:34So just to speak more broadly, for AI and healthcare and for many other sectors, it's not a technology story.
22:42That's not what I think.
22:43It's a governance story, it's an institution story, and it's a trust story.
22:47So that's what we need to be focusing on in India and the broader global south.
22:52And once you understand how to strengthen those aspects, the technology can actually be a part of the system.
22:59We have to be thinking about systems itself and not just one solution by one solution.
23:03Dr. Trihan, what would you like AI to solve for?
23:07Would you like it to predict the next pandemic or solve for air pollution?
23:11So there are two things.
23:14One, these are early days.
23:15And one is localization, localization to India.
23:21So when there are models existing which actually are predictive models for areas which are not mapped in India,
23:29I think that you'll always have a flaw in it.
23:31So what you can actually use today is to mine data very rapidly and lead to drug discovery,
23:41new modalities of treating different diseases,
23:46and the fact that India will play a major role in the world in drug discovery
23:52because we are lucky to have a very diverse gene pool.
23:56Not many countries have that because we had the Alexander's armies from the west,
24:02Genghis Khan and his gang from the east, and then the Mughals came from the north.
24:09So that whole gene pool mixing, they were very generous with their sperm.
24:13So there is a huge, huge diversity in our own population.
24:18So that's why I'm saying that there has to be a great opportunity for India to harness AI
24:26to actually help the world also because not, like I said,
24:31there are pure gene pools in most parts of the world.
24:35So that's one advantage.
24:37The other thing is that India needs to get into innovation drugs.
24:43We are paying a very heavy price because all the innovation is being done overseas.
24:49We are very happy being a pharmacy to the world, but copycat pharmacy.
24:54So it will help us to catapult us into, like we leapfrogged, telephony.
25:01India is still the most sophisticated, you know, telephone system right now,
25:07the wireless telephone.
25:08I think we can be the same at a very high level for the pharmacy of the world,
25:14not copycat.
25:16Very nicely said.
25:17I'm afraid that's all the time we have.
25:18But I think it's safe to say that when AI is all about probability,
25:22remember health is all about responsibility,
25:24and we need doctors for that, don't we?
25:27More than just AI.
25:27And the other conclusion, through all the sessions that have happened,
25:30news anchors may not be losing their jobs.
25:32Doctors are certainly not losing their jobs.
25:35Not yet, at least.
25:36Thank you so much, all of you, for being here today,
25:38and thank you for the audience as well.
Comments