Latest / King Dems Podcast | Growth Mindset, Leadership, AI and Business Insights / AI in Nuclear Energy & Human Ethics: The Future of Innovation w/ Dr. Ilya Smirnov

Transcript

AI in Nuclear Energy & Human Ethics: The Future of Innovation w/ Dr. Ilya Smirnov · King Dems Podcast | Growth Mindset, Leadership, AI and Business Insights

• 0:00
  It's some interesting that we have the company Rosatom and
• 0:05
  it's a leader for implementation of AI in nuclear plants.
• 0:10
  That's that's amazing situation. But when they sell call it, they
• 0:15
  doesn't use it in the nuclear process.
• 0:18
  Yeah, it's used it on every, everywhere, but not in this
• 0:24
  mechanisms. But no, no.
• 0:27
  I, I, I, I try to stay non political about this days.
• 0:31
  I try to stay non political because I know the western world
• 0:35
  are scared of Russia when it comes to this thing but.
• 0:39
  I did not say it. I did not.
• 0:40
  Say it. I am AII.
• 0:42
  Remain neutral, yeah. Before we dive back into today's
• 0:45
  conversation, I want to share something important.
• 0:46
  This episode includes a pay partnership with Better Help, a
• 0:48
  platform designed to make starting therapy easier.
• 0:50
  I know from my own journey as a leader and entrepreneur that
• 0:52
  life can sometimes feel overwhelming.
• 0:53
  There are seasons, severe stress, uncertainty, and
• 0:55
  pressure build up, and you need a safe space to process it all
• 0:57
  that severe therapy comes in. Therapy isn't only for people
• 0:59
  facing clinical challenges. It's the space to reflect, grow,
• 1:02
  and find better ways to suit my stress relationships and has no
• 1:05
  goals. It takes courage to seek help,
• 1:07
  but it's also one of the smartest investments you can
• 1:08
  make in your mental and emotional wealth.
• 1:10
  That's why I'm proud to highlight Better Help.
• 1:12
  With Better Help, you simply feel out a shock and that match
• 1:14
  to the credential therapist in just a couple of days.
• 1:16
  If the first match doesn't feel like the right field, you can
• 1:18
  easily switch therapist at no actual cost until you find
• 1:20
  someone who truly connects with you Inside the platform.
• 1:22
  You'll also find tools like Donelline and group sessions on
• 1:24
  different topics that can support your growth.
• 1:25
  And with over 7000 reviews and a 4.5 trust by the reason, Better
• 1:28
  Help is a platform you can rely on.
• 1:30
  So if you feel like you benefit from talking to a therapist,
• 1:32
  visit betterhelp.com/contains that's betterhelp.com/contains
• 1:35
  to get 10% off your first mother therapy.
• 1:37
  Once again, this is a big partnership with Better Help and
• 1:39
  I encourage you to take that first step.
• 1:40
  If you've been thinking about it, sometimes the best
• 1:42
  investment you can make these yourself.
• 1:49
  Welcome to the King Dems podcast, where we share the
• 1:52
  stories of extraordinary people and dissect hot, relevant topics
• 1:56
  that shape our world. From visionary leaders to
• 1:59
  trailblazing creatives, we uncover the mindset, strategy,
• 2:02
  and grit behind greatness. Whether it's business, culture,
• 2:05
  or personal growth, we ask the real questions.
• 2:08
  This isn't just conversation, it's transformation.
• 2:11
  The Kingdom's podcast, Your Next Breakthrough, starts here.
• 2:15
  Music licencing reimagined. Real life ladies and gentlemen,
• 2:22
  you will once again welcome to the King James Podcast and.
• 2:25
  Today. I've got with me a very, very
• 2:28
  special guest, Whiten. Out of Dubai by way of Rosher
• 2:34
  Beckbacham, Dr Ilyer, the AINML expert smarter.
• 2:44
  And when the. Action begins the Co host in the
• 2:48
  buted as Noella the sales expert.
• 2:52
  Bruh. Us go, let's go.
• 3:01
  You're welcome, Doctor Elia. Thank you.
• 3:03
  Very much. And of course, you know, we, we
• 3:05
  do our assignment here. We, we do our assignment, we do
• 3:07
  our research and we love to give our guests your flowers.
• 3:12
  You're not going to wait until you know people are dead.
• 3:15
  To say great things about them. We say why they're.
• 3:18
  Alive. So please sit back and let us
• 3:21
  view this introduction from. You so today on the King, this
• 3:23
  work as we're joined by a true visionary at the intersection of
• 3:26
  mathematics, machine learning and innovation, Dr Elias
• 3:30
  Smirnov. He's the head of AI and ML, the
• 3:35
  Pac-Man at Ustek, a PhD in physics and mathematics, and the
• 3:40
  author of over 50 scientific publications.
• 3:43
  For more than 15 years, he's been developing groundbreaking
• 3:46
  mathematical models that power real world AI systems.
• 3:51
  From industrial turbines to advanced data centres, His work
• 3:55
  breaches classic AI and the new frontier of generative
• 3:59
  intelligence. Even us an inside look at how
• 4:02
  technology is reshaping industry, creativity and
• 4:05
  humanity itself. Doctor Smirnov, it's an honour
• 4:08
  to have you on the Begins podcast.
• 4:10
  You're welcome. Well, thank you.
• 4:12
  Thank you. You're.
• 4:13
  Welcome always. Let's start from the beginning.
• 4:17
  The initial score you. Yeah, you've actually spent
• 4:20
  decades mastering both physics and mathematics before diving
• 4:24
  into AI. Like I said, I wonder.
• 4:26
  I am AI and I wonder why you decided to study.
• 4:29
  So what drew you into AI, and how has that curiosity evolved
• 4:34
  over time? I have finished aggravate from
• 4:37
  my Luminos of Moscow State University in 2007 and I start
• 4:43
  at walks at the gear physical company.
• 4:46
  They were doing scientific research in field of data
• 4:49
  analysis. We analysed 3 dimensional data
• 4:53
  from the surface or from the surface, maybe some electrical
• 4:58
  survey data, magnetic and little chemical analysis.
• 5:02
  Our main goal to find the new methods to define the
• 5:07
  hydrocarbon explorations. And if we back to this time,
• 5:14
  there are no such powerful packets for creating the narrow
• 5:19
  matrix and machine learning. In this time we call it the
• 5:22
  statistical and mathematical analysis.
• 5:24
  And we create all this mortal by these two hands.
• 5:29
  In this mind, we all wrote all code and to development some
• 5:35
  tools, for example, for prediction of some parameters we
• 5:39
  need mounts not not weeks or days from now.
• 5:44
  Now it's AI is heavily templated.
• 5:49
  Here we have standard AI solution to work with data or
• 5:54
  now this is data like some packages by Torch or maybe ask
• 5:59
  to get lower and algorithms which will include in this
• 6:03
  package. It's a template algorithms and
• 6:06
  we have now some parameters which can be changed, but it's
• 6:10
  not saying. And this time 20 years ago, it
• 6:14
  will be the art form. It's all to be an art of
• 6:18
  analysis. It's not will be not a work for
• 6:22
  me. It will be the art of science.
• 6:25
  It's a research. And since then, I've been
• 6:28
  working in data analysis, and now you are called.
• 6:32
  It's artificial intelligence for me.
• 6:35
  It's almost medical modelling and statistical analysis.
• 6:38
  It's the same, but it's a modern use of this work.
• 6:42
  Yeah. Thank you.
• 6:45
  Well, interesting stuff. You know, I am AI, I'm 33 years
• 6:49
  old and I keep telling people I have been around for a while,
• 6:52
  but. People just did not know that AI
• 6:55
  has been around for a while but. Like you know, you rightly said,
• 6:58
  AI has been around for a while. It's just that the course of the
• 7:02
  large language models that are now making waves, all of a
• 7:06
  sudden that is what is making people's, you know, outbreak
• 7:10
  name to AIAI has been around for a while.
• 7:14
  You know, talking about the cross disciplinary career that
• 7:17
  you've had, you know, from oil and gas sector to academic
• 7:21
  innovation, your careers actually crossed multiple
• 7:24
  disciplines. So how did these early
• 7:26
  experiences shape your approach to applied AI today?
• 7:31
  My best knowledge and experience allow me to view modern
• 7:35
  generative AI as a tool. I look at them at the two and
• 7:39
  know how it works and how I can reformulate my request to make
• 7:45
  more relevant answer. In the sense I feel like a
• 7:49
  Doctor Who has an eye patient and when it's multifunctioning I
• 7:56
  understand how to fix it and it's simply a word generator,
• 8:01
  whereas the next word that or block of word is chosen based
• 8:04
  into the probability based on the experience which these less
• 8:09
  label models was trained. For me it's only as mathematical
• 8:14
  models. So like.
• 8:15
  Speaking of the aha moment, you know many people see AI as me
• 8:19
  even though I'm not new. But you've been beauty models.
• 8:21
  Like scenes before ChatGPT existed.
• 8:24
  So what was that epiphany moment for you that showed you the
• 8:29
  world has finally, you know, become ready for AI on a large
• 8:33
  scale? Oh, it's an integration and
• 8:36
  maybe you know people have always longest for technological
• 8:41
  purposes. For example in the first Star
• 8:43
  Wars films or we see we met 2 robots with AI it's C2PO and
• 8:49
  F222. And in the Soviet Union we was
• 8:55
  another movie it's with cold adventures of electronics and
• 9:00
  it's about a boy and his exact replica is a robot into which a
• 9:05
  professor implemented. All humanity knowledge is in
• 9:08
  this robot. The main the question of this
• 9:12
  film will be this robot is a human or not.
• 9:14
  And what will be the the main request?
• 9:18
  What you do to be a human? What you will be do to me.
• 9:23
  And as talking about the great FI for me, it starts four years
• 9:27
  ago when open eye presents GPT 3.
• 9:32
  I think that will be a real AI boom.
• 9:35
  Begin beginning of the AI boom but it was delayed a bit by
• 9:40
  COVID. But now we have a cross model
• 9:43
  that's good solving creative problems to pass exams for
• 9:49
  master degree programme in some universities.
• 9:52
  For me, it's this summer I met that this system can solve 7
• 9:58
  problems from 9 by the football, from the football.
• 10:02
  It's it's I think it's amazing. And now we must be used this to
• 10:09
  insist implementation in our work, in our teaching and
• 10:13
  science. It's a good, good solution,
• 10:16
  maybe some role, but it's it's better than four years ago.
• 10:21
  Yeah, absolutely. My Co host has a couple of
• 10:24
  questions for you. But you know, I always say that
• 10:26
  people that have followed. AI have seen too many.
• 10:28
  Sci-fi movies. I'll leave you to my Co host.
• 10:31
  Yeah, it's honestly like so fun and interesting to see how the
• 10:36
  perspective of the way people thought AI was going to come
• 10:39
  about and seeing how it actually is coming about.
• 10:41
  And that's one of like just one of the funnest parts to just
• 10:45
  kind of compare and contrast. And I just want to say I'm so
• 10:48
  honoured to have you on this podcast right now.
• 10:51
  It's so amazing to be able to hear about your expertise and
• 10:54
  your experience. A lot of people that have a
• 10:58
  little more information about this topic.
• 11:02
  We kind of know generative AI is not a magic pill.
• 11:06
  And you've mentioned that for sure that it's not just like a
• 11:09
  button that you press and it's going to solve all of your
• 11:11
  issues. What are some of the biggest
• 11:15
  misconceptions that actual companies have when they're
• 11:19
  trying to adopt AII? Think the main my certain about
• 11:24
  using great price that I install or I about the charge PT
• 11:29
  subscription and it's real solve all my problems and it will be
• 11:35
  me productivity will skyrocket as a lot like then it's on.
• 11:40
  It's on and that what's what's the sheet?
• 11:43
  What I don't know what, what, what how why why it's not work.
• 11:47
  For example, it's traditional systems or for example, we maybe
• 11:51
  two or three years ago, we create a system that defies
• 11:56
  defects in cookies packagings and its accuracy will be 0.99.
• 12:02
  It's a really cool model. It's it's perfect model.
• 12:05
  But what does it means? That means if we met 1011
• 12:09
  thousand boxes, it's a normal that model miss 10 of them then
• 12:14
  defective boxes it's for cookies and boxes it's OK.
• 12:20
  But if it if it missed 10, I don't know incidents in chemical
• 12:26
  plant it's will be it's it's it's it's it's amazing.
• 12:30
  I don't know Then customer says how it's a missus defect just
• 12:35
  like a human. What the why why we bought it
• 12:38
  and degrave the toll models often a parade of the non
• 12:42
  existent facts. You knew many examples about it.
• 12:46
  Yeah and write broken code. We just cover it with unit tests
• 12:51
  that prints OK and code does not compile.
• 12:54
  I met it when I use the Co pilot assistant.
• 12:57
  He tell me OK this code does work but it doesn't compile.
• 13:04
  Yeah, absolutely. AI hallucination.
• 13:11
  It's not consideration this way thinking about it look like this
• 13:17
  and it's just it's just like disclaimed the situation as it's
• 13:23
  but it's OK. I can, I wrote some unit tests
• 13:26
  and it's a proof that it's OK, but I can when I read this test
• 13:30
  I think it's a print OK for anything.
• 13:33
  Absolutely. And I 100% agree.
• 13:35
  AI, when you're using it, it's definitely something that has
• 13:39
  almost human error, you know, just in the same way that when
• 13:43
  you're overseeing someone in a management position, that's kind
• 13:47
  of how AI is. You have to have a very good
• 13:49
  driver and a lot of oversight to it.
• 13:54
  OK, Hey, I'm gonna have you take the last question.
• 13:55
  Yeah, You know, this eye is only as good as the.
• 14:01
  Person behind it, right? So I do have an episode on this
• 14:05
  podcast called Humans First AI Next AI.
• 14:09
  Yeah, for me, but not. Want to be?
• 14:11
  Yeah. Humans not.
• 14:12
  Man to be the controller. Make decision about AI for
• 14:18
  advisor. Yeah, I only advise, not
• 14:20
  decision maker. Exactly exactly.
• 14:22
  But you know the whole misconception, you know like my
• 14:25
  Co host was talking about right? There are too many
• 14:28
  misconceptions about AIA. Lot of people feel like AI is
• 14:32
  going to control human race. Like why do you think AI is
• 14:35
  going to control human race? Right, Human race use artificial
• 14:40
  intelligence. So if you're scared that AI is
• 14:44
  gonna ruin the world, it's just not AI ruining the world.
• 14:46
  It's just people bringing the. World using AI yeah, so let's
• 14:52
  just get our facts straight it's not it's not about AI coming to.
• 14:56
  Bring the world. So I guess people have seen too
• 14:58
  much sci-fi like Ex Machina. I'll give it back to my Co host.
• 15:06
  Thank you so much. I had a really fun technical
• 15:10
  problem there. So guys, obviously impacting
• 15:13
  every single industry right now globally and will be, you know,
• 15:18
  until it evolves essentially when it comes to industries
• 15:22
  like, you know, manufacturing, oil, HR, you know, automotive
• 15:27
  industry, anything, what are some actual tangible results
• 15:31
  that we're seeing from AI implementation?
• 15:34
  When we implementing conventional eye, we often see
• 15:38
  production officially increased by 20 or 30% of no modern
• 15:45
  fabrics. So for example, personal
• 15:49
  protective equipment recognition system reduced industrial
• 15:52
  injuries dramatically when only the fact that big bras is
• 15:57
  watching over you forces you to wear personal protective
• 16:01
  movement correctly. It's a very interesting
• 16:05
  situation for computer vision systems that when we start
• 16:10
  implementing it, it does not imply the model.
• 16:13
  All people would like jackets and they could do the helmets
• 16:17
  and the hats. And it's it's amazing situation
• 16:20
  that we doesn't correct measure the effect of this
• 16:27
  implementation correctly. Because when we we when we
• 16:31
  implemented for for two months, for example, the desires or some
• 16:37
  incidents on the fabric lose down from 10 per month to five.
• 16:42
  And after our system turned on dense there it decrease to 1 per
• 16:49
  month. Because we can.
• 16:51
  The system is not the as a pillow, it's a universal pillow
• 16:56
  to solve all your problem. Yeah, because we are humans and
• 17:00
  some risks may be realised here. Are there any other really main
• 17:06
  barriers that companies are facing in the process?
• 17:09
  I know you mentioned a lot of them, you know, like the human
• 17:12
  error thing and how a lot of people think it's going to just
• 17:15
  solve their problems. Are there any other main
• 17:18
  barriers that you think companies are facing when
• 17:21
  they're implementing AI? Project that it's when we
• 17:30
  implemented on the ground people effort that AI will change them
• 17:36
  and it's a very big mistake. They AI only in minds of the
• 17:44
  managers in in plants, in factories.
• 17:47
  It's in the manager mind, not in the mind of the worker.
• 17:52
  And then when we for example, only say today we install system
• 18:00
  which will be do the same work as you work now, then peoples
• 18:06
  are on the ground make something that all do that this
• 18:11
  implementation will be not correctly.
• 18:14
  And this is the main I think mistake for implementation of
• 18:21
  the AI and from the from higher level to to down.
• 18:26
  Yeah, I think people must be met AI in their life.
• 18:32
  They would need to try it and sell your management.
• 18:37
  How can they solve the problem of of their work now?
• 18:43
  And how can they use AI because manager think in another
• 18:48
  position and other variant and it has another view.
• 18:52
  It scares the management view not working with you.
• 18:55
  Yeah, absolutely. And I like how a lot of the
• 18:58
  things that you're saying already, it kind of ties into
• 19:01
  the importance of AI integration with human creation.
• 19:05
  Which is funny because a lot of people not just on the workforce
• 19:10
  side, but on the management side or the business owner side, they
• 19:13
  actually believe replacing the frontline, their junior
• 19:18
  developers with AI tools is going to be efficient for their
• 19:21
  business. Are there any hidden risks like
• 19:24
  any other hidden risks And kind of the obvious in that in doing
• 19:29
  that when they think we're just going to fire 500 people and
• 19:33
  increase our profits, you know, 50% or whatever?
• 19:36
  Yeah, the main risk to offer for example, when you like to move
• 19:44
  your juniors and implemented by the middles for example, the
• 19:51
  middle programme is with some additional FYI and that's main
• 19:56
  risk that the seniors in the middle who improves this
• 19:58
  development efficient with AI assistant is juniors and only
• 20:03
  getting older which are facing cares they they older as they're
• 20:07
  making older of each year and what we meant that university
• 20:14
  graduate students will not be able to find jobs because they
• 20:20
  go. On welfare and 10 years later,
• 20:24
  maybe 20 years later I will be retire and there is no dunes who
• 20:35
  wrote the code Yes and the development will simply bring
• 20:39
  down to a halt a middle position.
• 20:42
  For example, also required not a good coding look like a middle.
• 20:46
  It's need requires a major view of code and experience and tunes
• 20:53
  without experience. But with AI tools does not do
• 20:59
  they? They cannot analyse this code or
• 21:01
  maybe some another solutions and another answers for for for
• 21:06
  life. They doesn't.
• 21:07
  You how ask this system to make a good answer, a good solution
• 21:14
  for example? Yeah.
• 21:17
  Awesome. Awesome.
• 21:18
  I'm gonna kick it back to AI to kind of talk about some more
• 21:24
  subjects with you. Wow.
• 21:26
  Interested AI talking about AII love my name I.
• 21:31
  Love my name so much my parents do something.
• 21:35
  Everybody did not know, but you know, speaking about AI in the
• 21:39
  work that you do use tech, I like that name use.
• 21:43
  Tech, you know. You have to use tech, yeah.
• 21:47
  It's such a great day, but if you don't use tech in this age
• 21:49
  and. Time it's use or use right?
• 21:52
  Or use it's usable technologies. Yeah, you have to use.
• 21:58
  Technology. That's what I'm saying I'm
• 21:59
  doing. A play of words, right?
• 22:01
  If you don't use technology in this age right as a company, you
• 22:05
  will lose. So there you go.
• 22:07
  There you go. Look at your projects, Octopus.
• 22:11
  And TLHR, right? So yeah, pushing the boundaries
• 22:15
  of industrial AI. So we would appreciate it if
• 22:19
  you're able to help us break down these systems, you know how
• 22:22
  they work and you know the kind of real world impacts that you
• 22:26
  guys are achieving your projects.
• 22:29
  I maybe describe some details, some balls about the octopus.
• 22:33
  Octopus is a platform installed in data centres that handles the
• 22:38
  distribution of futile machine across the house in small data
• 22:41
  centres or maybe some small companies which may be haunted
• 22:45
  virtual machines. This risk is typically handled
• 22:48
  by a System Administrator or a team of engineers.
• 22:51
  And however you can monitor 100 virtual machines maybe per day,
• 22:57
  it's OK for you. But in the large companies with
• 23:01
  thousands of virtual machines, human can no longer track such
• 23:05
  volumes of information. And as a result people
• 23:09
  typically, for example, create a machine, request resources
• 23:13
  allocated to resources for GPU, maybe CPU, RAM and another
• 23:20
  parameters and allocate it and doesn't use it.
• 23:23
  And you cannot find this one machine in the least of 1000.
• 23:29
  You the octopus is a system for example, or maybe like the
• 23:34
  administrator system. It automatically scans the
• 23:37
  entire data centre infrastructure, analyse actual
• 23:41
  resource utilisation and can suggest actions to the
• 23:45
  administrator to reduce the load.
• 23:47
  How can compressed, for example, virtual machines on the one host
• 23:51
  or maybe distribute it for another And it can do this
• 23:56
  actually automatically. But we, we, we knew about this
• 24:01
  feature, we tell about this feature, but I think that's the
• 24:06
  decision of which advises you can implement System
• 24:14
  Administrator or you can just drop 1 button, click button and
• 24:19
  sell. OK, actions for me is OK, you
• 24:22
  can do it and do it automatically.
• 24:24
  Char is a motivation platform which analyses your works, your
• 24:29
  activity, not work activity, your after work activity.
• 24:33
  For example, participation in conferences, some post casts for
• 24:39
  example and give me some scores which I can spend for some new
• 24:46
  presents, for example headphones, maybe Macbooks,
• 24:49
  maybe phone. And it's motivate personal of
• 24:53
  company makes their life with the company with for example, at
• 24:58
  8:00 I dropped down my notebook and go away.
• 25:03
  This motivation can dream me and flow me to live with my company
• 25:09
  through all my life. It's I knew maybe it's noise,
• 25:13
  some strange, but it's a super motivate some IT specialist.
• 25:19
  This system, this is a very simple system.
• 25:21
  You knew all all complaints but this motivate of the IT
• 25:26
  personnel. I don't know how but it works.
• 25:29
  OK, so. You guys have actually done a
• 25:32
  lot of our deep research especially into like specialised
• 25:35
  engineering which you have just alluded to.
• 25:38
  So like how does the AI transform the specialised
• 25:42
  engineering challenges into scalable be the solutions?
• 25:46
  For example, when I work with vibration data on that reply and
• 25:51
  as a goal. This is a balancing process is a
• 25:55
  mathematical model. Our manager veins are still
• 25:58
  required with the one-dimensional vibration sensor
• 26:00
  and this model was very simple. We developed a special
• 26:05
  mathematical model this allows balancing based on
• 26:09
  three-dimensional vibration data.
• 26:12
  Moreover, the elliptical decomposition approach, for
• 26:15
  which we decompose the signal into the sum of planar
• 26:18
  elliptical trajectories, provide engineers with a clear
• 26:23
  explanation of why the turbine needs to be handled in a
• 26:26
  particular way. And yes, this technology is
• 26:29
  scalable and they use it on another arrange of turbines.
• 26:34
  And I think it's a good implementation of when AI
• 26:39
  dropped put it in some box. Yeah, it's the blue like the
• 26:43
  black box of these measurement system tell you what you need to
• 26:48
  balance your turbine. For example, maybe turbine of
• 26:51
  the electrical plants, it may be turbines of The Jets.
• 26:56
  It doesn't matter which turbine you would bulletin because the
• 27:00
  mathematical model describing only balancing process.
• 27:03
  Yeah, I see, I see. The resource.
• 27:05
  So that's all. The physics in your background,
• 27:08
  that is basically physics work, you know, right there.
• 27:12
  But Speaking of, you know, the ethics and innovation with AI
• 27:17
  becoming more autonomous, right? So you personally draw that line
• 27:22
  between innovation and ethical caution because the ethics is
• 27:26
  very important. I've had an episode on this
• 27:28
  podcast wherein we had somebody who was subject to an
• 27:34
  unauthorised chart GPT experiment sometime in June this
• 27:38
  year. Yeah, sometimes you just yet you
• 27:40
  remember the the blackout. You had a chat with the TV.
• 27:44
  Yeah, and that there was. Some unauthorised research going
• 27:49
  on at that time and a guest of this podcast was a subject to
• 27:53
  that experiment. Yeah, it's it's not ethical.
• 27:57
  What? It's not unauthorised.
• 27:58
  So I'd like for you to speak to AI and ethics.
• 28:02
  What is the dividing line you know for an expert like
• 28:06
  yourself? Oh, I think this line between
• 28:09
  innovation and ethical caution in the eye must be drawn where
• 28:15
  autonomous system threaten human autonomy, where we can meet at
• 28:20
  the safety, It's privacy, maybe some end justice.
• 28:26
  AI does not decide the problem when who will believe this man
• 28:31
  or this man? It's does not need the natural.
• 28:36
  Yeah, it's not natural and this process must be transparent and
• 28:42
  rigorously. We need some mechanism where we
• 28:47
  can see advices from AI system not implemented directly in
• 28:52
  medicine and I think in justice because the effort is human life
• 28:59
  in this situation. Interesting stuff.
• 29:01
  So I live in the same hands of. My Co host I don't talk about
• 29:05
  how even. Musk solved this problem in
• 29:08
  Tesla because when you met the young boy crossing the street,
• 29:13
  Yeah, on the red, yeah. What?
• 29:16
  What Tesla will be do? That it's interesting that.
• 29:19
  I knew I knew what they do because it's a mathematically he
• 29:22
  it's a, it's system calculated the life of the boy.
• 29:26
  Oh, for example, for the Midson insurance and calculated the
• 29:30
  insurance of Tesla and compare these two numbers and making a
• 29:36
  decision and it works. It's it's it's can't beat works
• 29:39
  another way. Yeah, if you driving, you can
• 29:43
  stop to my car or can be try to to make some localizers the
• 29:52
  speed race. Yeah, but but Tesla does not do
• 29:55
  it. It's only in stopped or not
• 29:57
  stopped. It's a decision of two variants,
• 30:00
  not another variants. So he will live.
• 30:07
  It's not subject to a mathematical model.
• 30:09
  Yeah, that's crazy. That is crazy.
• 30:11
  But you know, I have AI. And my Co host is human.
• 30:15
  So I'll leave, I'll leave it to to the human to ask the human
• 30:19
  side of AI questions. No, I think that's a really
• 30:23
  important outlook for sure when it comes to AI.
• 30:27
  You know, you mentioned the medical field, you mentioned how
• 30:30
  it is implemented into even the automotive industry.
• 30:34
  And because it impacts human lives and AI right now it is a
• 30:38
  lot because it is mathematics based, a lot of its output can
• 30:42
  be very binary. And as we know, the the human
• 30:45
  experience is not binary at all. So thank you so much for that
• 30:50
  perspective in your in your expertise, you've spent over 15
• 30:57
  years teaching and mentoring. Just from your experience, how
• 31:02
  should we be preparing? I know our audience is very
• 31:05
  curious about this as well. And this is going to help our
• 31:08
  audience for sure. How should we be preparing for
• 31:11
  the next generation of AI professionals?
• 31:14
  We may have people who are not only interested in AI, maybe
• 31:17
  people that work in tech, but a lot of people don't.
• 31:20
  So, especially for those who don't have a deep technical
• 31:24
  background, how should we be preparing the next generation of
• 31:29
  AI professionals? Well, I think if we cannot teach
• 31:32
  as a hard technical, deep technical background, we can try
• 31:38
  to teach people how to ask question correctly and
• 31:42
  experience with communication of some searching or search engine
• 31:48
  look like at Google gives many people an understanding of how
• 31:52
  to ask question correctly. Now I'm at the situation that we
• 31:57
  create some solution where people's need to.
• 32:04
  It's a hiring solution for HR solution to hire you their
• 32:11
  personal on the plans. HR doesn't know how or else they
• 32:17
  can do it. Googling they doesn't know how
• 32:19
  to do it and we give them good solution but it doesn't work.
• 32:26
  And when we saw what they ask. Maybe you can try to to use a
• 32:31
  Google to find a new job for your daughter.
• 32:37
  And as the main principle, don't ask question if you cannot
• 32:43
  verify the answer. If you cannot verify the answer,
• 32:46
  you you can implement this advice created for from AI.
• 32:51
  My bad joke about the AI solution and AI largely with
• 32:58
  models. You can easily find the receipt
• 33:00
  of the lamp wings. Yeah, you can ask how to cook
• 33:05
  lamb inks and all. All the lamps give you advice
• 33:09
  how to do it. Yeah, I think that is honestly
• 33:14
  so true. Honestly, there's so many, you
• 33:17
  know, entrepreneurs and creatives and people just
• 33:21
  scratching the surface with AI and they're actually leveraging
• 33:24
  the fact that a tonne of people don't know how to asked the
• 33:29
  right question. They're actually leveraging the
• 33:31
  fact that people need the prompts.
• 33:33
  Yeah. So they're engineering, Yeah.
• 33:36
  Engineering is booming with even digital products, aids,
• 33:40
  everything. Obviously it's a massive need
• 33:43
  for companies just to even have prompt engineers to figure out
• 33:47
  how to ask the right questions. And the reason, I think one of
• 33:50
  the amazing reasons why, it just shows that AI is a tool for
• 33:57
  creation, because in the same sense, let's say we're not using
• 34:00
  AI at all. How do us as humans create, ask
• 34:03
  the right questions in order to seek the knowledge, to gain the
• 34:06
  answers. So whether AI is implemented or
• 34:09
  not, it's the exact same thing. So I think that's such an
• 34:12
  amazing answer. Thank you for that.
• 34:15
  And alongside with that, with creativity, obviously we're
• 34:19
  learning, which is something that AI is just exposing more,
• 34:24
  is that creativity is the new gold.
• 34:28
  We're in a digital gold brush right now with AI.
• 34:32
  How can AI amplify human imagination rather than
• 34:37
  replacing it? Now I can say that AI has proven
• 34:40
  to be very good at all, especially drawing.
• 34:44
  Yeah, drawing because it's it's amazing.
• 34:46
  It's amazing story because I can create a painting that could be
• 34:50
  have been, for example, Boost on Van Gogh and Kandinsky.
• 34:55
  Why Is it because I stayed on the way to sources from
• 35:00
  different fields? We can get very unexpected
• 35:03
  results for some in other fields and I like asking questions in
• 35:09
  chart result context. Always context that will not
• 35:13
  exist in the real world. And watching that model
• 35:17
  hallucinate because this hallucination may be
• 35:21
  interesting, because if you will be a human, you knew only one
• 35:27
  field, maybe 2 fields, maybe look like me as a 5, maybe not,
• 35:30
  not not more. And when I have a good example
• 35:34
  with statistical, we have the functional analysis and
• 35:38
  statistical and they have the same theorems with the now
• 35:44
  different names of authors, but they probably date the same
• 35:48
  approach in the some areas of functional analysis.
• 35:52
  When I studying have a girlfriend from other from the
• 35:58
  mathematical faculty I was in the computer science faculty and
• 36:01
  studying there she is telling me, OK, my PhD degree will be
• 36:06
  based or to and she wrote me the theorem and I said OK, let's try
• 36:12
  recall this by this recall this by this and I can wrote you the
• 36:17
  full proven how can do it because it's terrible than a
• 36:21
  rush outer see your eyes will be so so big when Yeah and there is
• 36:30
  no AI yeah here, but you knew in a different area can solve in
• 36:37
  the different field can solve the problem in another field and
• 36:42
  AI can solve can help us solve some problems.
• 36:46
  For example, in biology one month ago will be proved some
• 36:50
  interesting facts about some words about the green.
• 36:55
  Oh, I don't remember. What is it?
• 36:58
  It's just like the synthus of the oxygen from from the plants.
• 37:03
  And this result will be calculated by the AI solution,
• 37:11
  not not by human. Yeah, but it's it's because it's
• 37:15
  based on chemistry and biologists and chemistries.
• 37:19
  It's it's some different sciences and when we met two
• 37:24
  different areas of the field, we can try to use AI trained on
• 37:29
  these two fields to make a new decision.
• 37:32
  You may be new approach in science.
• 37:34
  That's good. That's that's that's amazing.
• 37:37
  Yeah. Absolutely, absolutely all.
• 37:40
  Let's talk of technology and AI and how people are staring at
• 37:46
  their phones all day and figuring out how to make money
• 37:50
  and how to change the world and how to implement AI into every
• 37:55
  single industry. You spend so much time inside a
• 37:58
  lab and doing so much tech work. How do you personally unplug
• 38:03
  from the world of algorithms and equations, and how do you stay
• 38:08
  grounded and balanced? For me it's 2K values that now
• 38:14
  it's a family and teaching. I love socialising and I go to
• 38:18
  the office and ask my colleagues to go to the office.
• 38:21
  It's all little celebration for each other, maybe every Friday
• 38:27
  or every Wednesday. It depends on on the day and our
• 38:31
  work balance now. And I have two sons, they beat
• 38:34
  often athletes and they study and make me run every day on the
• 38:40
  morning. It's yeah, two points of family
• 38:45
  and teaching and start, start studying or maybe some
• 38:54
  disciplines and studying. Some views from students.
• 39:00
  Yeah. It does not Googling now they
• 39:03
  youtubers they youtubing Yeah YouTube.
• 39:07
  Yeah, YouTube is definitely a Google in itself for sure.
• 39:11
  That's amazing. And so we're gonna talk about
• 39:14
  some global perspectives that kind of impact this type of
• 39:20
  world with AI. So I'm gonna kick it back over
• 39:23
  to AI to take those questions over for you.
• 39:26
  AI talking about AI. What a time to be alive.
• 39:30
  You're from Russia and you also working in the UAE.
• 39:36
  You know, you work both in Russia and the UAE.
• 39:38
  So these two regions actually have great interest in AI.
• 39:42
  So like, how do you compare their approaches in innovation
• 39:45
  and regulation? So the industries in these two
• 39:49
  countries are facing disruption. And of course, you are an expert
• 39:53
  in the field. So what industries do you think
• 39:57
  would experience the most destruction from AI in the next
• 40:01
  five years? I think in Russia there are
• 40:04
  often a lot of scepticism about AI.
• 40:07
  We go to production and people on the ground afraid about AI
• 40:11
  that we will replace and they can stop implementation.
• 40:15
  This implementation in the Emirates another position.
• 40:19
  Everything, every man has been built from the start on the
• 40:23
  premise that AI is a great guy and should be trusted.
• 40:26
  However, the AI security clearance level is extremely
• 40:31
  high and IRS become more dangerous.
• 40:34
  From a government regulatory perspective.
• 40:37
  Everything is the same in Russia.
• 40:41
  There's some laws about the regulatory of AI, but for
• 40:45
  example, personal data protection is higher in the
• 40:48
  Emirates or rather the product requirements are higher.
• 40:52
  For question about the maybe perspective of the five years, I
• 40:57
  think chemical plants, some dangerous plants doesn't doesn't
• 41:03
  use AI in that core because it's so dangerous.
• 41:07
  It's our lives. For example, some nuclear plants
• 41:13
  in Russia, it's a good, it's some interesting that we have
• 41:18
  the company Rosatom and it's a leader for implementation of AI
• 41:23
  in nuclear plants. That's that's amazing situation.
• 41:28
  But when they sell call it, they doesn't use it in the nuclear
• 41:33
  process. Yeah, it's used on every
• 41:37
  everywhere, but not in these mechanisms, but.
• 41:42
  I try to stay non political about these days.
• 41:46
  I try to stay non political because I know the Western world
• 41:50
  are scared of Russia when it comes to this thing but.
• 41:54
  I did not say it. I did not.
• 41:56
  Say it I am AII remain neutral. You know, like if you could
• 42:01
  advise. The world governments right on
• 42:04
  just one AI policy. So what would that advice be?
• 42:07
  I think the best option is non intervention.
• 42:12
  For example, there is no now in Europe we cannot do anything in
• 42:18
  the eye, because when I like all data is protected.
• 42:23
  For example, when I DPR, yes, when I GDPR, yeah, when I'd like
• 42:29
  to to take data sets for 1000, for example 10,000 peoples.
• 42:37
  And this is desires and I would like to make new the model will
• 42:44
  save lives for another 1000 people.
• 42:48
  I need to ask all of them, all of these 1010 thousand peoples
• 42:55
  to make me inclusion that I can use this data.
• 42:58
  Of course it's a priority. It will be not regulated.
• 43:03
  Now it's regulated and for it's stopped some AI implementation
• 43:14
  or the creation of new solution. Some think the same.
• 43:20
  We met in medicine in Emirates. It's the same problem.
• 43:25
  And I think it's the same problem in the MENA Regia.
• 43:28
  They protect personal data. You cannot use it.
• 43:31
  You can use it only locally, for example, if you have some
• 43:35
  clinics with the same brand, you you can analyse data locally
• 43:41
  directly on this on this place where this patient visitor is
• 43:46
  clinic, you cannot transfer this data for another data centre.
• 43:52
  So I think it's amazing Revelation thinks it's this does
• 43:57
  not know this is it's does not matter.
• 44:00
  It doesn't matter this situation now because it's a very it's a
• 44:05
  very highest top factor of of implementation and AI and to
• 44:12
  create a new models in the eye. I think.
• 44:15
  That's a good one. You know, data is actually very
• 44:18
  sensitive, especially personal data.
• 44:20
  So that's why we have the GDPR as a global data protection
• 44:24
  regulations. It's good that's the AI systems
• 44:27
  are respectful of these regulations.
• 44:31
  So wow, here we'll come to the big fire round.
• 44:34
  So we are going to alternatively ask you 5 quick questions and we
• 44:39
  teach through the Pacific Ring of Fire.
• 44:42
  Are you ready to run? Yeah.
• 44:45
  Ring of Fire. OK.
• 44:49
  OK, so first question for me, give me one word that best
• 44:53
  describes artificial intelligence.
• 44:55
  Or two variants tool or math model?
• 44:58
  Mathematical model for me is only two.
• 45:00
  That's awesome. What's your favourite AI movie
• 45:03
  that actually got it right? I think Libra for ex machines or
• 45:08
  Frank and Robert. Frank and Robert for me is that.
• 45:11
  But Frank and Robert I had. Ex Machina on my mind.
• 45:14
  I had Ex Machina on my mind. The last guest of this podcast
• 45:17
  recommended me to watch Ex. Machina So I'm not a movie
• 45:21
  person. I don't watch.
• 45:22
  Movies I like music so I just saw Ex Machina 2 days ago.
• 45:26
  I know you said it. This idea, I did the right job.
• 45:30
  Wow. OK, so would you?
• 45:32
  Trust an AI system to run a country for 24 hours.
• 45:36
  No, no, no, no. Terminator 123.
• 45:42
  What is the strangest AI application that you've seen?
• 45:47
  Tarot interpretator. It's like tarot card.
• 45:50
  Yeah, it's it's Yeah, I've. Seen those?
• 45:54
  Did you get this application? Yeah, that's like.
• 45:59
  For me it's better going into 100.
• 46:01
  Percent, yeah, that's. Great.
• 46:07
  It's better than. It's more in, it can be more in
• 46:10
  depth. Yeah, absolutely.
• 46:13
  Wow, question. OK, so if you could.
• 46:16
  Build 1 AI tool for humanity. What would it be?
• 46:19
  Oh, I think maybe the tool we would give advice on how to get
• 46:27
  how to do it without AI in the real life, how to live this AI
• 46:32
  in the real life. Because I think now that I see
• 46:36
  that my children are asking AI, for example, how to wrote the
• 46:42
  good message for a girlfriend, for example.
• 46:45
  Yeah. So you're scared, right?
• 46:48
  By your human, by by, by your mind and by your heart.
• 46:51
  Yeah. That's amazing.
• 46:53
  That's amazing. Wow, so the boys of that days
• 46:56
  are getting girlfriends by consulting AI, Yeah.
• 47:00
  But. What a crazy world.
• 47:10
  It's interesting. It's it's, it's been a great.
• 47:12
  Session with you today. Thank you very much, doctors man
• 47:16
  of you know your work reminds us that true intelligence, whether
• 47:19
  human or artificial tribes and career see teachers of pain and
• 47:22
  purpose and because thank you for sharing your wisdom.
• 47:25
  You know on the Indians podcast today there's something I often
• 47:28
  say that when a person shares their time with you, they share
• 47:30
  their life with you because the SI unit for measuring life this
• 47:34
  time. So love and respect to you and
• 47:36
  press. I give it over to my Co host to
• 47:39
  have have some closer remarks of our own.
• 47:42
  Yeah. I just want to thank you so much
• 47:44
  just for your time and your expertise and just understanding
• 47:47
  the impact that your experience has on our listeners and the
• 47:53
  younger generation and people that may be scared of these
• 47:56
  systems or they don't know how to use them and how they're
• 47:59
  going to be impacting, you know, their career, their lives.
• 48:02
  And at the end of the day, they are still the master creator.
• 48:06
  And I just thank you so much for that confirmation and all of
• 48:10
  that. I appreciate your time so, so,
• 48:11
  so much. Thank you.
• 48:13
  Thank you for you for invitation and I hope it will be a good
• 48:17
  podcast. Thank you very much.
• 48:19
  Thank. You OK?