Welcome to Real Talk on AI and I care with your host, Dr. Scott Morris. And I'm co host, Dr. Rehan Ahmed.
In each episode, we discuss current news, adding a little AI education, and debate innovative AI tools and topics that will change the industry. Well, welcome to another episode of Real Talk. My name is Rehan Ahmed, and I'm joined by Dr. Scott Morris.
Scott, how are you doing this evening? It's good, Rehan. It's been a super interesting week. Lots of lots of thought provoking discussions this week.
About a lot of different stuff, but I think the big thing is this, and I wouldn't even say it's somebody in our industry. It's actually I've ran into a couple different engineers and a couple did a big data scientists this week. And are the conversations we've been having have been around data centers. And so I think I want to do that as the knowledge bite and talk about what are they?
You know, and so if you're cool with that, I'm going to go for that. No, fantastic. I think as we make that as we get deeper into the tops, now we're talking about infrastructure. So we talked about, you know, what is AI, what are algorithms, what are hallucinations, sort of bread and butter.
Now we're going into infrastructure. And we are a lot about data centers. I definitely read a lot about in the news. I'm not sure I actually know what it is or what they are.
So Scott, please. Yeah, let's go for it. So, you know, I was thinking about it this way. We spend if you think about this, not just contextually, but in reality, we spend our lives kind of in the cloud, right?
Well, in the cloud, well, what heck is the cloud? Well, the answer is there's no cloud. It's not in the sky. It's this massive complex of these humming buildings called data centers.
And you can think of it like a data center is. Or a complex of data centers, kind of like the brain of the internet, right? It's the great neural network of how the internet works. And it's basically the series of physical facilities that house thousands, tens of thousands.
And now they're getting into the hundreds of thousands of these network computers, we call servers that kind of store this information. They process, they distributed it all live time instantaneously. And so, I mean, think about this way every time we were watching a movie on Netflix last night, every time you stream a movie or send an email or check your bank account or do something like that. Some data center somewhere on earth is processing that request.
And so, you know, the big challenges these days is building one and it's not just like let's go buy some will up warehouse and, you know, throw a few laptops in there. I mean, these data centers need just three big things. One, they need massive power. I mean, I think we come in on that a couple weeks ago, you know, it's, it's predicted by the year 2030 that 80% of all power made in the world will facilitate data centers.
80% of all power. And you know, these facilities run 24 seven and they're getting bigger and bigger. You know, every single large data center, I think we come in a couple weeks ago, can consume as much electricity as a small city every day. That's just massive amount of power and where are we going to get?
Are we going to get it from coal or we're going to get it from oil? Are we going to get it from fission fusion solar wind? And the answer is, I'm not sure we have enough power for any of those things that any of those things are going to give us enough power. So number two, the other big resources connectivity, you know, when we talk about fiber and, you know, they need just these ultra fast, multiple redundancy fiber optic cables that move data around the globe in just milliseconds.
If you think about every one of these data centers, a large data center has something like 6,000 miles of fiber optic cables in the data center. That doesn't count the fiber optics going across the ocean floor or what's going up into space or whatever. I mean, the connectivity of these have to be massive and they have to be over fast. And then the third category sources security.
I mean, it's just you're talking hundreds of thousands of square feet protected by scanners and reinforced vault walls and something protecting against a nuclear blast. And I mean, these data centers, third. They have more resource than any city has. And, you know, but the big challenge about this is that building them getting harder and harder and the two biggest hurdles we have is they create so much heat from their processing that we need something to cool them.
Otherwise, they're going to melt, right? We used to say we're going to use these massive air conditioning units, but that's just as energy intensive and now they're experimenting with liquid cooling and you know, may even sink and data centers into the ocean. Or we talked about a couple weeks ago, maybe they're going to put them in space where there is no heat sink, right? It's just space.
You know, then there's the water footprint. I mean, these centers are using just millions of gallons of water every day to cool millions much more than a nuclear power station uses to create, you know, just just trying to. And that's really a problem if you're liking Colorado, we have a huge data center up in northern north of Denver. And the problem is we haven't had any snow in three months for the most part.
We're very drought prone. And so what do you do to cool these data centers or do people drink water? What's what's most important? I don't know.
I mean, do we balance that thirst for data with the thirst for the environment or maybe ourselves. That's a huge puzzle that we need to solve. And I don't think we're anywhere near that. You know, so I think that I think about this every time I click, you know, I go to Gemini and I click go look for this or I click my feed on my phone or whatever.
There's thousands of rows of blinking light on a computer and hundreds of fans cooling this to make my one acquisition possible. And we've talked about it before a single query to an LLM is significantly more power intensive than a Google search. Oh, it's a microwave. It's money.
A microwave for one minute every time you do a search. That's the power. And so if you just think about that scale, then that's why we're reading by data centers all the time now, because they existed beforehand. They were in office buildings and maybe some some larger ones.
But this is not because of Netflix. I mean, that's part of it. No, it's part of it. But the reason we're really reading about it in the past few years is to compute needed for AI is just putting this on orders of magnitude more than the trajectory we were originally on before this AI revolution.
Yeah, our Gemini GPT search is just every day, not searches, but utilizing it every day. I thought I was making a video the other day and it was a it's a five minute video. And it took five hours of compute time to get this five minute video. And I thought about that.
I was getting ready to prep for this. I was like, you know, I probably just basically burned a hundred gallons of gas in my vehicle to do that one five minute video. Yeah, yeah, and so when we think about some of the negative externalities of AI and what we're sort of living through oftentimes we talk about security and privacy and of course, those are super important and, you know, whether it really existed in species, also relatively important. But the energy usage, the compute needed some of the environmental issues that are at play.
I think that's really these are really important topics. And so thanks thanks for the the knowledge that Scott is very helpful and I hope our audience now has a has more insight on to what data centers are why they are much more than news. And can can speak about them much more confidently. Well, I wanted to talk about for our fireside chat to talk about my opium monitoring and my opium management and the world of AI.
You know, Scott, I wrote about this recently. I have I have three young kids and I'm a minus six. My wife is also my opic. And I see my kids all the time on their screens with some books in their faces inside a lot.
And you know me being a minus six and seeing patients all the time for issues of from my opium from my op my opic maculopathy retinal attachment glaucoma. You know, these are things that are getting more definitely there's a lot more interest. And now with the new lenses from Esselord that are out distalist lens for my opia treatment as well as the my site from Cooper and the Hoyer lens. There's more treatment out there.
And so I want to spend a little bit of time on how is AI impacting my opium management. And so, you know, and I don't know if you've seen this, but like, you know, one of the things that's that's there is who's going to progress. I think one of the things who's going to progress on my opia. I guess right now the answer is, how do we know, right?
Because we don't track the data and, you know, maybe AI is going to give us more, I don't know what would be the, I guess, the term progression prediction. Yeah, you know me exactly right. I don't know if that's a real term or not, but I think that's kind of it. It was like, I guess it'd be no different than glaucoma, right?
Is will these models when they get fed with real lifetime data, be able to start looking at, you know, hey, what is glaucoma prayers, but maybe more importantly, because glaucoma is one or two percent of population my opia is 40% and growing and growing. And obviously this is a bigger, a bigger issue in Asia and a lot of the research is happening there, but it's coming here. It's just it's, in my mind, they're just foretelling what's going to happen for the rest of our country. And so yeah, exactly right.
So, you know, there's nothing approved yet from the FDA in terms of any consumer to, you know, any software that we track that. I mean, how do we, is there like I think about as having another podcast will be kind of shortly that that I did we were talking about the future is probably IOP sensors inside the item monitor, ocular pulse amplitude. What, you know, I guess let's brainstorm around, I don't know of any device that actually looks at my opia that's not a for ropter. But, but then I want to think about as we get to these automated refracting systems that may be in a mall.
Maybe you're doing a fraction every week or every month or every three months or instead of going to a doctor's office. Well, you know, and full disclosure, I advise a company called blink has a company. It's called Globe by medical and have blink frames and we do exactly that. So really, we, we had the smart glasses and unlike sensors or like the medical glasses with chip cameras that are pointing out word focus or about augmenting reality, which I think is super cool.
Maybe sometimes a bit creepy, but otherwise very cool. These cameras point in words and a monitor really important parameters that are linked to my opip progression. So think screen time not just screen time on your phone like that because it tells you, but screen time like looking out of computer right now we're looking at a TV screen. And another important parameters outdoor time we know kids who spend more times outdoor in sun, particularly they have less myopic progression.
And then not only that attracts activity, it tracks neck position, it tracks behavior. So if you're more active, we also know that exercise is linked to less my open all these all these great things you'd want to do anyway, it tracks that we should be doing. And we do right. Yeah.
So so in addition to doing other things that are maybe not related to my opip like blink rate. But it tracks and it's great for kids and the really cool thing is to pair that. So if I'm not, you know, if I'm a following a kid who I'm worried about my opi and I want them in a stela's lens, for instance, I would love to put them in these blink frames because not only am I doing the treatment, but I'm also tracking really important behavior. And also these maybe light exposure, you know, I like it.
Yeah, we know how much less natural light different than LED light. It is. Yeah, we're talking about sunlight, you know, the schools in China have had windows in their classrooms to allow natural sunlight through the windows. Exactly for my opi not for anything else, not just so that for the kids to see better outside, but for my op my opi progression.
So it's really, it's really important. And the last thing to monitor is these lenses are dose dependent. So if you look at the FDA studies, you have to wearing these glasses are the lenses for a certain amount of time a day, right? Otherwise, it doesn't work.
And so the frames and monitor how long you're actually wearing these glasses. You know, right, I have to say I'd never even really tell you just said that I've never even really thought about what the impact of AI was going to be in my opi management, man, that the opportunities are endless like a. Oh, it's huge. Yeah, it's huge.
I mean, maybe this is a disease, you know, I was my opi is a disease. I think too many times in eye care, we go, no, that's just a refractive state. But you know, I always say this has a symptom. It has an objective test.
It has a diagnosis and it has a treatment that that's called the disease. That affects 40% of the population is not many other eye disease that do that other than cataracts. This may maybe we, you know, I'm not sure I've got in my lifetime, but maybe we stomp out people don't get more than a minus five. So yeah, yeah, I was a minus 10.
So, you know, the thought of never being that near sighted that's that's that's fascinating. Why else I think you know, so where do you think that goes with today was a day I had, I'm, I must have fit like four scleral lenses today and a bunch of gas perms lens today. So I was dealing with high my opus and funky corny as all day. I wonder what, where does that, you know, because I think about what AI is doing for the ophthalmic lens industry.
I wonder what it's going to do for the contact lens industry. Like will we, will we will AI design better or though K lenses or better scleral lenses or better gas perl lenses or you know, I wonder will AI have an impact on changing that parameter and will we find that fitting somebody half doctor flatter than we think really does now we can have that theory. So one thing I think it's great. We talk a lot about digital twins, especially for complex long term disease.
And I think we sort of talk about glaucoma as a case or maybe met macadet degeneration. I think my opus is a great opportunity for digital twins think about you have a young kid who may have, you know, we're talking about potentially a decade of treatment if they're doing some like apthropene or for a while. I mean, not short, not about for a month and then or or though K there's sort of combination therapies are going to spectacles. And that is going to be after being non responders.
The last thing you want to do is be on after being for two years and it doesn't work, you know. And so this is what I think the idea of a digital twin where you can go through a simulation in advance of initiating therapy to determine. So at least have a better conversation with the parent on hey, we were we ran your child's information. We put in all their case.
We put in their actual links. We put in the refraction. We put in the parents refraction. And the the system output various probabilities.
And then we've won we hit a long term Scott imagine not just from six to 16, but from six to four. You know, I saw a couple of months ago. I saw 30 year old high my own with a really bad Mac off Bretton Old attachment and who didn't who had surgeon, you know, it was never 2020. She's not seeing normally ever again.
And can you imagine say your risk of developing a retinal dashboard or a map of my app XC and D. That's that's huge. And so I think a digital twin for these long term diseases can be a world of difference for initiating therapy or not. Because right now I'm saying it's like, oh, that doesn't actually solve the problem.
That doesn't really solve the problem. Right. I mean, I've had lacy. I still growing.
I still growing. Yeah. I mean, you know, then I also wonder about we, you know, we were talking. I think it was last week of the week after we're talking about the future of, you know, it's not going to be our phones.
It's going to be some form of heads up display. I'm not sure it's going to be meta or whatever we're doing now, but you know, some evolution of that. And then I wonder, you know, right now we have the stonthus lens. But we'll someday because as we move to heads up display, we will be able to program a lens, one lens to alternately blur.
Instead of using atropi. I believe there's been some commercial and after looking to which companies doing this, but I believe doing a blur for ambiopia is selectively is something that's been thought about. I haven't seen it being commercially successful. Yeah, I certainly haven't seen a kid in that, but that is definitely interesting.
So yeah, I love the digital twin though, Rayhan. I hadn't really thought about that. That's an awesome idea. I think and especially because we need buy it.
It's hard to talk to patients about risks that are so remote and maybe hard to understand when they've all just been taught, you know, what, you know, glasses. That's what's the big deal. But you know, obviously we're talking about much more. And this is where patient education, I think is becoming more important.
And I can be helpful there to help for educate the parent and the child, especially. It's just that that five minute video cost was much to run my truck, remember. That's right. Other than that small detail.
Other than the being stewards of our environment. Yeah. Yeah, that's just that's you know, right on I thank you. I hadn't really like this kind of like expanding my horizons in terms of there's just things I'd never really thought about.
But then I think about if we're having that talk about my opia. Maybe presby opias no different either, right? I mean, maybe there's something we're going to do for presbyopia. I and being over 40.
Boy, that'd be really nice. And you know, the disease, the disease, the presbyopia, yeah, the disease aging. I also agree with that. But yeah, I hope this was helpful in terms of talking about a disease and it's getting so much interest.
And I now with the new lens is coming on out that there's something we could actually do optically for it in addition to contact lenses, after pain. So really excited about where this field to go is in terms of my opia management and AI. And hope it's been helpful for the audience to hear about it. All right, everybody.
So thank you for joining us. So this week as well. Obviously, we always say if you have any comments or questions, please feel free to reach out to Rayhan or myself. It's rayon our.
Again, or Scott as COD. We have the following it's at a I the letters a letter I in I care.ai. We are always happy to get suggestions. We you know, we we listen to the audience and we try to, you know, some of the things we've talked about the last couple of weeks here because people called in or emailed us and said, hey, we want to know more about this.
So something that's tripping your trigger or you like, hey, I just need to know more information. You've been listening to real talk an AI and I care your weekly podcast to keep you informed about AI technologies revolutionizing I care.