Google Cloud Speech-to-Text API (Chirp) + Gemini 2.5 Flash Non-Thinking. Treat it as a searchable index of what was broadcast, not a quotation record.
00:00:17take, we'll put electrodes on, yes, how
00:00:21is modern cutting-edge science done? The most
00:00:24accurate answer to this question is to try it out yourself, everyone knows.
00:00:28järele proovida, kõik teab.
00:00:34At the Institute, in these laboratories,
00:00:36the smartest solutions are being developed on how to use
00:00:40smart devices and the data
00:00:42contained in them in medicine. We are testing one such today.
00:00:45The purpose of this study is
00:00:49to validate these smartwatches and
00:00:52three parameters: pulse frequency,
00:00:56number of steps, and energy expenditure. How accurately these watches measure these.
00:01:00parameters, do I understand correctly that you suspect that these
00:01:04watches are inaccurate? No, we don't suspect they are inaccurate, we
00:01:07actually want to know how accurate they are. Smartwatch manufacturers
00:01:11are very different and we are not quite sure how
00:01:14reliable one or another watch is,
00:01:18so here we take watches from different companies,
00:01:21we put
00:01:24them on the test subject, on their arm, and then
00:01:28either within a couple of hours or...
00:01:30or in some cases within 24 hours, we collect
00:01:33the data that comes
00:01:37from the watch and then compare it with our so-called
00:01:40gold standard, i.e., steps taken on a treadmill and heart rate
00:01:44measured by ECG. Wires
00:01:47attached, oxygen mask on, and we begin
00:01:51the scientific experiment. So, do you feel
00:01:53normal? As much as one can feel
00:01:57normal in this situation, yes. Following
00:02:00the instructions of Institute researcher Kristjan Pilt and his colleagues, I perform
00:02:04routine activities, I walk, I run,
00:02:07I lie down, I ride a bike, I solve
00:02:10puzzles,
00:02:12mhm, so that was a three-year-old's, yeah,
00:02:16yeah, yeah,
00:02:18mhm, and I even clean, and we have
00:02:22here on the floor a bit of
00:02:25paper from a paper shredder, which needs to be swept up with a broom,
00:02:29and then if... once you've swept it up with a broom, then you have to sweep it apart again
00:02:32and then sweep it together again. This is indeed
00:02:35such a fruitless task, but the important thing is the process you're
00:02:39doing, so it took two minutes again, okay. At the same
00:02:42time, the smartwatch records various data
00:02:46about my health.
00:02:48What is the bigger goal of this? Why are you
00:02:52doing this? The bigger goal would be that
00:02:55people, who daily wear their
00:02:59pulse. so already collect health data
00:03:02about themselves, that they would be able to
00:03:05transmit it to the health portal,
00:03:08and if they then, for example, go to a family doctor or
00:03:12to an occupational health doctor, then these
00:03:15doctors would already be able to foresee what the
00:03:19dynamics in these parameters have been,
00:03:22and these doctors would not then have to
00:03:25send these people for initial examinations, but in fact, the examination would already have been performed by these
00:03:29people themselves. One could say the philosophical
00:03:31background
00:03:34in the medical sector or healthcare sector is very similar
00:03:38to other service sectors, and
00:03:40if we think back twenty years ago, that
00:03:44when I went to the shop and asked from behind the counter, "Give
00:03:47me that bag of sugar and that pack of
00:03:50butter," then that time is irreversibly gone. Today, I'm
00:03:54on an airplane, I've had to buy the tickets myself, I
00:03:58have to check in my own luggage... No one helps
00:04:01me anymore, and the same trend
00:04:04is in healthcare, that we have to give
00:04:07people more tasks to deal with their
00:04:10own health. Professor Peeter Ross explains that in the
00:04:14future, our smartwatches will be connected to Estonia's national
00:04:17health information system. If this succeeds, it will save
00:04:20a huge amount of resources. We can
00:04:24collect data now
00:04:26in the smart age, not once a day and not
00:04:30once an hour, but practically once
00:04:32a second, which means that we
00:04:36get a broad
00:04:38and comprehensive overview
00:04:42of a person's health, and we can also
00:04:46get this data throughout their life, and not only when
00:04:49I go to the
00:04:53doctor, who has a white coat on, and my
00:04:56blood pressure is already rising and my blood pressure is already rising. The point
00:05:00is that if currently repeated specialist visits burden
00:05:03the medical system, then in the future
00:05:07some appointments could be canceled and instead, people would monitor
00:05:11themselves at home. themselves, or rather, the smartwatch
00:05:15would do it for them. This way, doctors' valuable
00:05:18working time can be saved, and in total, millions
00:05:21of euros. Do you mean that this very smartwatch on
00:05:25my wrist, that this device will diagnose me
00:05:28in the future and tell me when I should go to the
00:05:32doctor? I would be very cautious with the word
00:05:35"diagnose," but it will give advice. It will certainly give
00:05:38advice. Well, I think a lot of
00:05:42it already gives advice, that if by the end of the day
00:05:46you've only taken 500 steps, then you should go out
00:05:50for a walk, and that is the simplest
00:05:53example. But what we do with these
00:05:57studies is we look at these daily patterns,
00:06:00how the pulse quickens, how much a person walks at different times during the day,
00:06:04and based on that, over a relatively
00:06:07longer period of time, we hope
00:06:11to be able to predict when... when a person's health
00:06:14starts to deteriorate, not specifically to diagnose, but
00:06:18then, if activity decreases,
00:06:21for some reason the pulse quickens in unusual situations,
00:06:25then there would be advice through the
00:06:29smartphone, through an avatar, that you should go
00:06:33see a family nurse or an occupational health doctor, or
00:06:36such prevention is the goal.
00:06:39What diseases the smartwatch literally
00:06:42detected in such a way in the early stages in practice in the future?
00:06:46What is the challenge of our society
00:06:50is obesity and lack of movement,
00:06:54and from this non-communicable
00:06:58diseases arise,
00:07:01which are a great burden both for individuals and society, namely cardiovascular
00:07:05diseases, diabetes, various mental health disorders,
00:07:09so the range is very wide, just like I..."
00:07:12"have been tested today by more than 80
00:07:15test subjects. Although a larger analysis is still ahead, major
00:07:19and minor deviations have already been identified, where
00:07:22makes a mistake and for example
00:07:25miscalculates the amount of energy expended, but other
00:07:29things have also emerged during the studies, yes, we had various
00:07:33things happen here, that for some
00:07:37subjects we detected arrhythmias from the ECG signal and we
00:07:40advised them to contact..." their family doctor so that they could
00:07:44get advice and see a cardiologist and
00:07:48see what was wrong with their heart
00:07:51function, and some subjects also
00:07:54left some activities unfinished which were simply a little
00:07:58too overwhelming for them, that there's nothing to do, our goal was
00:08:02not to complete the performance, so to speak, but to still monitor
00:08:06your body and stop at the moment when
00:08:09your feeling tells you to.
00:08:29Institute of Molecular and Cell Biology show how
00:08:33many hard drives worth of information are needed to conduct
00:08:36one modern scientific study. Tartu
00:08:39University researchers, led by bioinformatics associate
00:08:42professor Kaur Alasoo, are mapping, with the help of data from more than half
00:08:46a million people, how genes affect our
00:08:49metabolic processes. Our main interest is
00:08:53how we can generally develop better,
00:08:56faster, and more effective medicines. What is the current problem?
00:09:00Why does drug development often fail? Firstly, it's because
00:09:03pharmaceutical companies are very good at making these small molecules, but then they go
00:09:07into clinical trials, and it turns out that the molecule works well, but the protein was
00:09:10wrong, meaning the target. The drug's target that we want to influence was wrong,
00:09:14and that's why it doesn't work, for example. Or the other problem is
00:09:18that it might work, but it has many other
00:09:21side effects, it's not safe. Often
00:09:25it is believed that every gene in our body performs only one specific
00:09:28task. Alasoo and his colleagues' study, however, proved
00:09:32that one gene can simultaneously affect hundreds of compounds and
00:09:35molecules. In practice,
00:09:38this means that if you try to suppress only one indicator with a drug at a time,
00:09:42the person may still not get well. What do you
00:09:46do then? We try to find these
00:09:49so-called targets within cells, to find targets that
00:09:52can be targeted with drugs. A human cell is like a Lego house, which
00:09:55consists of blocks, the blocks are so-called proteins, and then
00:09:59different blocks have different functions, that is, some are made like
00:10:02walls, some might correspond to then some are made into windows,
00:10:06doors, a roof, and so on. For example, if we want to make it so that there is more
00:10:10light in the house, that if our problem is... that the disease is that it's too dark
00:10:14in the room, then how to get more light, then we would try
00:10:17to find which proteins we should influence so that there is more
00:10:21light, if we think about that house, no matter which block we remove,
00:10:25then as a result there is more light in the house, and there is most light when we
00:10:29only remove the roof, then there is a tremendous amount of
00:10:32light, but now rain comes in, for example, yes, so
00:10:36that is precisely the problem, isn't it, that we get a side effect, that there is a lot
00:10:39of light, but rain comes in and it's cold too, that is,
00:10:43maybe medicating, it's not such a good idea to target
00:10:46the roof. Tartu researchers studied the genetic
00:10:49data of more than 600,000 people
00:10:53based on information from the Estonian Gene Bank and the UK
00:10:56Biobank, and compiled, figuratively speaking, the necessary
00:11:00instructions for assembling a Lego house, which helps
00:11:03to distinguish the true causes of diseases from accompanying
00:11:06symptoms. Where does artificial intelligence come into play? We
00:11:09actually would like
00:11:12to have an artificial intelligence model that could
00:11:16predict for us what
00:11:19a specific genetic variant
00:11:22does, if we change it. I don't know, on the first
00:11:25chromosome, position 1 million 321 to a
00:11:29C, firstly, if anything happens at all, and if
00:11:32it does, what happens, that is,
00:11:36we would like the model to be able to predict this for us, not that we have to do an experiment and measure
00:11:40every time we want to know it. Firstly, the model predicts, and secondly, the model can also
00:11:44tell us why it predicts, meaning the model basically explains
00:11:47biology to us. What help is this to an ordinary person?
00:11:50How does this help me if we have such good
00:11:54models? We can, as a whole society, develop
00:11:58medicines more quickly and efficiently,
00:12:01and then with that, either treat or
00:12:04prevent various diseases. On the one hand, it helps to make drug development
00:12:08significantly faster, cheaper and more efficient. On the other
00:12:12hand, it would create an opportunity to focus on new diseases,
00:12:15the development of drugs for which has not previously been considered
00:12:19reasonable for economic reasons. For example, rare diseases, or especially
00:12:22rare diseases, where drug development
00:12:25starts from the actual needs of each patient,
00:12:28we need to find that needle in the haystack, how to find that one change among
00:12:32the three billion. In reality, the difference between two people is a million, we have to find
00:12:36the one that causes the disease, and that is where, if we have very good
00:12:40models that can predict what each
00:12:43genetic difference does, then it is significantly easier to find
00:12:46these variants, and then we can also think about what the therapy could
00:12:50be.
00:12:53In fact, in the Estonian medical system, there are several
00:12:56areas where artificial intelligence has already gained
00:12:59a foothold. In Estonia's largest North Estonian
00:13:02Regional Hospital, intelligent machines work together with doctors
00:13:06on a daily basis. It
00:13:09is warm at first, don't be scared, you can close your
00:13:11eyes, it's more comfortable. We'll close
00:13:15the clips and I'll make breathing free immediately.
00:13:19Mhm. Okay, I'll come closer with my fingers.
00:13:21ligi.
00:13:26One place where artificial intelligence assists doctors
00:13:29is radiation therapy. The artificial intelligence tool precisely
00:13:33marks the contours on the image, i.e., where the disease is located and where
00:13:36the healthy organs are. This way,
00:13:40doctors can direct the treatment to the right place and better
00:13:44protect healthy body parts.
00:13:47Here you can see, well, I think there are a couple of hundred
00:13:51contours here, which are all
00:13:55structured. Of course, not everything is always needed, it depends on where
00:13:59the tumor is located. So in that area,
00:14:02it is already known what the risk organs are, then
00:14:05the radiologist technician must contour them before
00:14:08the doctor starts contouring
00:14:11the tumor volume. He has already done that work, yes, he has
00:14:15done that work. How fast is it? It takes a few minutes, so
00:14:19not long. For example, here there are 380
00:14:23layers, so it's... layers taken from the body in axial
00:14:27sections. Here you can also see them in other planes, here
00:14:30these layers are in the coronal
00:14:34plane. Simply put, if it's the prostate area, then we need to
00:14:38contour the rectum,
00:14:41the sigmoid colon, then we need to do
00:14:44the femoral heads, the bladder,
00:14:47and sometimes also the bowel bag, that is, the area that includes all the bowels,
00:14:51like basically here, you know. So
00:14:54everything that is in the human body can be
00:14:56contoured.
00:14:59How big is the role of this artificial intelligence in your work today,
00:15:03that it is basically a full-time
00:15:06assistant for you. Yes, it is daily, that
00:15:09basically all patients who come to us for a
00:15:13CT scan, who come to us for treatment, we use
00:15:16AI for all of them, so it's a standard that we
00:15:20take these AI contours and we
00:15:22then correct those contours. according to what we
00:15:26need, if it has done something a little wrong somewhere, then we correct
00:15:30it, we cannot leave
00:15:33it so that we don't check a single contour, we still check all contours
00:15:36in the relevant area. How much does AI make mistakes,
00:15:40how much does it give results that are not true? I can't
00:15:44really say numerically,
00:15:47but as for contours, there are contours that I
00:15:51practically don't have to correct, but there
00:15:53are contours that I have to correct a lot, or then do
00:15:57them myself from start to finish, perhaps, because it might be more convenient and faster for me
00:16:01to do it myself than if AI has done it. It
00:16:04all depends on the patient's anatomy. If
00:16:08the anatomy is unusual, for example, if the patient has been operated on,
00:16:12then the anatomy has already changed due to that, and
00:16:15AI may not
00:16:17detect it as well.
00:16:19Have you measured how great the effect of this artificial intelligence
00:16:23is, how much time you save or how much human work you save
00:16:27with it? I can confidently say that it saves at least
00:16:30about 50% of time on
00:16:33the contouring part. Generally, I can say that once...
00:16:38if the images, i.e., the CT scans, are taken with a very thin layer,
00:16:41then it could take three to four hours
00:16:45to correct or contour all the images
00:16:48of one patient, now I think we can
00:16:52do it with AI in about an hour, so that's a very
00:16:55big time saving,
00:16:58wow, that's still quite a lot, basically you can see four patients
00:17:02instead of one, yes,
00:17:03basically.
00:17:07Moving from radiation therapy to the radiology center, where
00:17:11artificial intelligence literally saves lives. If we now
00:17:14imagine that we have a patient who is developing
00:17:18a cerebral infarction, then
00:17:21their brain must be treated as quickly as possible to save
00:17:25as much brain as possible. So we are talking about minutes, basically? Well, the
00:17:28faster, the better, so Estonia has a very good
00:17:31ambulance service that brings patients to the hospital very quickly, and
00:17:35then all this process that we are talking about
00:17:39now, until the patient
00:17:41receives treatment, can take less than an
00:17:44hour, well, in an ideal situation,
00:17:47right? Which means that the patient comes to our hospital, a neurologist
00:17:51examines them and decides that they need a study, then they
00:17:55come to us for the study, then the study is done, which looks exactly like what is on
00:17:59the screen, that is, such a
00:18:02boring black and white image for the ordinary eye, but
00:18:05a radiologist sees a lot here. Where
00:18:08AI now comes to help is that it helps the human
00:18:12eye to make this thing
00:18:15perhaps easier to follow, meaning if we look at this colored
00:18:19image, then
00:18:22the computer tells us to
00:18:25look here, it makes it red for us, basically,
00:18:28it clearly indicates, it clearly says
00:18:31that the blood flow to the brain has decreased, so this is the place where
00:18:35an infarction starts to form. And then what else we do
00:18:39is such an angiography, meaning we inject
00:18:42a contrast agent into the patient which starts circulating in their blood vessels,
00:18:46and if there is a clot in the blood vessel somewhere, which is
00:18:50the cause of this infarction, then it is also able to
00:18:53mark that place for us, meaning we can quickly
00:18:57tell the neurologist, "There is a blockage in this
00:19:01cerebral artery. Due to this blockage
00:19:03of the artery, the patient has suffered
00:19:06a cerebral infarction. Such and such an amount of the brain
00:19:10is damaged, such and such an amount is still
00:19:13salvageable." And then the person is called
00:19:16who will immediately remove that clot from the cerebral
00:19:20artery. Do you notice that artificial intelligence in radiology is perhaps the most
00:19:23advanced, or do you have any idea compared to other
00:19:27fields? In radiology, artificial intelligence has perhaps
00:19:30been used the most, perhaps the longest. In some European
00:19:34countries, for example, AI already does
00:19:38the first reading of screenings, which means that a human doesn't
00:19:41look at normal images anymore, AI selects
00:19:45those that need human attention, and
00:19:49some are then left out of the pot, meaning this saves
00:19:53our resources, that we don't have to review all these things. In Estonia, we are
00:19:57not yet that far, mhm, in Estonia, every
00:20:00patient is still examined by a doctor. Yes, and in the case of screening, the important
00:20:04nuance is that it must be done by two people. Completely
00:20:07healthy people come to the examination, and two radiologists look at
00:20:11the image, one looks and the other looks. Let's talk about the other side too, how
00:20:15much AI, artificial intelligence, makes mistakes?
00:20:19It varies greatly by product,
00:20:21and it depends on how sensitive we make the AI. If
00:20:25we make it very sensitive, then it makes more
00:20:28mistakes. If we make it less sensitive, then it makes
00:20:32fewer mistakes, but then it misses
00:20:35things. What is being worked on today is precisely to find
00:20:38that golden mean, where it would find all those things, but
00:20:42not create too much noise. Where could radiology
00:20:45develop with artificial intelligence in the next 50
00:20:49years? Do you have any idea of what doesn't exist
00:20:53yet but you might dream of? Speech recognition is one thing
00:20:56that is elsewhere,
00:20:59outside Estonia, very normal in the context of radiology,
00:21:02that everyone dictates radiological answers, but since
00:21:06Estonian is so specific, there hasn't been very good
00:21:10speech recognition in Estonia until now. How would that work in practice, if that
00:21:13speech recognition was in Estonian, what could you do with it then?
00:21:17Most of the time I'm looking at an image, mhm, and at the same time. While I'm looking at this image,
00:21:20I start describing that
00:21:24the cortical gyri are widened, cerebral atrophy
00:21:28has developed, here he has a lesion, in the region of the
00:21:32basal ganglia on the left,
00:21:34so I look at the image, I can already speak, and the text then
00:21:38goes automatically, it goes automatically to you, but since we don't have that right now, then
00:21:41it's like this: I look, I write, I look, I write, I look, I write,
00:21:45and that's what actually takes up my time, exactly that.
00:21:49How to take doctors'. overburdening bureaucracy and
00:21:52paperwork away and give time for direct communication with patients
00:21:56will be the greatest charm and challenge of artificial intelligence
00:21:59in the coming years. Professor of Practical Ethics Kadri
00:22:02Simm sees great advantages in the application of artificial
00:22:06intelligence in medicine, especially considering that the financial
00:22:09resources of the healthcare system are increasingly constrained. However, according to her,
00:22:12the dangers and risks that have accompanied
00:22:15new technologies throughout history must be considered.
00:22:19Gunpowder can be used to make bombs and fireworks, and
00:22:23in the case of medical technologies, we also have many examples of
00:22:26how whether its effects are good or bad does not actually depend
00:22:30so much on the technology as on the context in which it
00:22:33is applied. A good example, this is not AI, but
00:22:37it's ultrasound technology, which is actually a relatively cheap technology, used for a very
00:22:41long time in pregnancy monitoring.
00:22:44And as a result, we now
00:22:47have a situation where, you know, there are many
00:22:50countries and nations where there are many more
00:22:53boys than girls, because during pregnancy monitoring
00:22:56with ultrasound, it is possible to determine the sex of the child. And if the
00:23:00cultural context and social context, you know,
00:23:03supports some visions that boys are better than
00:23:06girls, then the consequence of this is now that we are missing millions of girls
00:23:10and millions of women, and well, basically a social catastrophe
00:23:14because these women were never born. And yet the same
00:23:18technology here in Estonia has not caused anything like that, right?
00:23:22So when talking about the impact of technology, it is always really important
00:23:25to consider the society, the culture, or
00:23:29perhaps the specific medical field that influences it.
00:23:31In the case of artificial intelligence, Simm primarily sees a danger in the
00:23:35increase of inequality.
00:23:38how AI is trained
00:23:41on certain data, and that data comes from somewhere.
00:23:44In today's world, it's mostly English and Chinese
00:23:48data, which represents populations, or even more precisely,
00:23:52not the entire population. For
00:23:55example, in America, only those who have money to go to the doctor and who make it there,
00:23:59and then data is generated about them, right? And
00:24:02one concern with this is that, yes, that data is biased and the application
00:24:06of algorithms developed on the basis of these data
00:24:10in other societies,
00:24:13on different populations, may cause such problems. Simm
00:24:17predicts that as artificial intelligence grows,
00:24:21in the treatment of patients, doctors' technological
00:24:23awareness will become increasingly important.
00:24:27Of course, it is important that doctors trust and know
00:24:31how artificial intelligence works in their field. And, well,
00:24:35this is related to the broader question, isn't it, of medical education
00:24:38in general today, what should be paid attention to there first
00:24:42and how the advent of artificial intelligence affects how this
00:24:45medical education is given, what is generally taught,
00:24:48because data comes in so terribly much
00:24:51all the time. And people are never equal
00:24:54in their ability to analyze that amount of data compared
00:24:58to artificial intelligence, so something else is what
00:25:01that doctor should be doing there, so that doctors,
00:25:04and also other medical workers who are dealing
00:25:08with diagnosis using artificial intelligence,
00:25:11should know how
00:25:13it works. Now, a more complex problem is that
00:25:16often the developers themselves don't know how that answer
00:25:20comes out from the other end, but it's also interesting
00:25:24that, well, we have different legal frameworks, don't we?
00:25:27We have the United States, which is influential, we have
00:25:31the European Union, which is influential, we have our own Estonia,
00:25:35right, that in different legal frameworks,
00:25:38there are different opinions on this, for example,
00:25:41resuscitation, a resuscitation decision,
00:25:45right, that if there are no doctors nearby,
00:25:48no medical workers nearby, the question is whether to resuscitate
00:25:52or not. Again, if artificial intelligence
00:25:55or that system has the information that this person
00:25:58in no case wishes to be resuscitated, well, then
00:26:02it can be said that if it is not resuscitated,
00:26:06if artificial intelligence makes such a decision, then
00:26:10it is the right decision, or vice versa, right? Well,
00:26:13as long as a human is still ultimately responsible
00:26:16in some sense, that the doctor has put it into
00:26:20the system.
00:26:21I think that
00:26:25there is no clear black and white line
00:26:29between what artificial intelligence is allowed
00:26:32to decide and what it is not, if the final
00:26:35word does not remain with it.
00:26:38It is clear that we are living through a time of great
00:26:42changes and it is probably
00:27:33the main role in its development. The technological readiness
00:27:36is already there today. The question is rather,
00:27:39are you ready?
00:27:42The extensive news program of the end of the week
00:27:45will be on air in just a moment and after
00:27:49that at 10 o'clock a new episode
00:27:52of the crime series Professor will begin.
00:27:55"The criminology professor with a peculiar nature
00:27:58has found himself in a situation at the apex
00:28:02of an unpredictable chain of events.
00:28:05But the investigation of new cases continues
00:28:09with all its might anyway.
00:28:12The stylish British crime series Professor
00:28:15T. New season. Tonight
00:28:18at 10 pm on Estonian Television.
00:28:21Or watch it immediately on Jupiter.
00:28:24Rocket Junior is back. "Are there brave and clever
00:28:26experimenters in your class?
00:28:32Rocket Junior invites students from fourth
00:28:35to sixth grade to a science competition,
00:28:39where success is primarily brought by good teamwork
00:28:42and thinking outside the box.
00:28:45Winners will be flown
00:28:48to one of Europe's most exciting science
00:28:51centers. Put your students to the test
00:28:55and register for the competition today.
00:28:58More information can be
00:29:38found on the website raketjuunior.ee.
00:29:42I'll put it down before, I didn't want to break
00:29:45that big root. Tree
00:29:47roots grow directly under the eyes of scientists,
00:29:51well, that's something that a person usually
00:29:54never sees in their life.
00:29:58Pärnu Bay is the most important
00:30:01fishing ground for Estonian coastal fishermen.
00:30:05In order to see how the fish in Pärnu Bay
00:30:33are doing, we have come trawling
00:30:35with scientists. In this episode, we
00:30:39will meet American soldiers who are testing
00:30:42their tanks in the forests of Võrumaa.
00:30:45How do the distant comrades-in-arms cope
00:30:48with homesickness?
00:30:51We also talk to local entrepreneurs. They have
00:30:54a very big appetite, so very large quantities are always ordered. And we will find out what the residents of Võru think about it. The question is whether they can be trusted. Don't miss the show "Comrades-in-arms" on Monday evening at 8 PM on Estonian Television. Estonian heart health is one of the worst in the European Union, whatever we can come up with to make youth overweight and lack of movement a smaller problem than it is now. On the last day of the Russian elections, Moscow was hit by a large-scale Ukrainian drone attack. The rising price of car fuel makes filling the fuel tank more meager. The heating period is coming.