Both research and practice create an optimum balance in delivering the best solutions. Angelo reflects on Season 1 with topics of how Computer Science changes and impacts our lives.
Angelo begins this episode with reflections on history and what brought us to the AI Winter. Why do we need a balance between research and practice? You don’t want to rediscover what has already been discovered or settle for something that could be better if you took the time to research a bit more.
In episode 4 we meet Angelo’s friend Andy Lee who talks about computer science predicting our biological age. Andy actually met Greg Fahy who talked about longevity. The study focused on injecting the thymus gland with a growth hormone that produced regeneration effects. The effects were measured through the epigenetic clock known as DNA methylation.
In Episode 6, Jim Shalaby talks with Angelo about how COVID-19 changed healthcare forever. Patients don’t have to wait in waiting rooms, they don’t have to find transportation to get there, and the patient has access to the clinicians.
The hard problems associated with explainability in artificial neural networks, we talked about in Episode 8. Angelo’s friend Nikos explained to us about five classic problems, one of which includes data privacy. Another big issue is developing a machine learning system to create adversarial attacks on the existing system.
In episode 7, Angelo’s friend Manos shared how complicated it is for people to invoke their right to have their data removed from a system. Typically those systems have to schedule deletions to remove the data through tombstones and a process called compacting.
What is on the horizon and what should we be paying attention to? We are going to run against barriers of technology. For instance, Moore's law is coming to an end. What do we do about that? What is happening in the short-term and how do we get past this barrier to the next? And then how do we blow away all those barriers with moonshots like quantum computing?
Finally, wrapping up our first season, Angelo wants to reflect on gratitude. Gratitude for you our listeners. Thank you so much for joining us on this journey. We really want to hear about your thoughts. The show is evolving just as the world is and we want to make sure that we're covering topics that you're interested in.
We would love for you to follow, rate, and review the show on your favorite podcast platform so that others can find us too. Thank you so much for listening.
Our Guests - Thank you!:
Nikos Myrtakis on LinkedIn
Manos Athanassoulis on LinkedIn and Boston University
Jim Shalaby on Twitter and LinkedIn
Andy Lee on Twitter and LinkedIn
About the Host
Angelo Kastroulis is an award-winning technologist, inventor, entrepreneur, speaker, data scientist, and author best known for his high-performance computing and Health IT experience. He is the principal consultant, lead architect, and owner of Carrera Group, a consulting firm specializing in software modernization, event streaming (Kafka), big data, analytics (Spark, elastic Search, and Graph), and high-performance software development on many technical stacks (Java, .net, Scala, C++, and Rust). A Data Scientist at heart, trained at the Harvard Data Systems Lab, Angelo enjoys a research-driven approach to creating powerful, massively scalable applications and innovating new methods for superior performance. He loves to educate, discover, then see the knowledge through to practical implementation.
Host: Angelo Kastroulis
Executive Producer: Kerri Patterson;
Producer: Albert Perrotta;
Communications Strategist: Albert Perrotta;
Audio Engineer: Ryan Thompson
Music: All Things Grow by Oliver Worth
Angelo: This season we asked the question, can computer science make our lives better? Join us as we recap season one. I'm your host Angelo Kastroulis and this is Counting Sand.
First, we need to go back to a statement I made in episode one. And that was that I said, you have to have a balance between research and application. Both are important. Why is that balance so important? Well, you will not get the best solution unless you apply both. Why can I say that? Well, back in the second and third century B.C., the Library of Alexandria was established.
Now we don't know exactly how much information was stored there, but some estimates go as high as 400,000 scrolls. What made it important was not just that it was a library with a lot of information in it, which there were other civilizations that had libraries, maybe not at that time, but new works in mathematics, astronomy physics, natural science were housed there.
So when I say new works, I mean, there was new, original research being done. It wasn't just a library, it was a place of study. It had empirical standards that were applied and it was one of the first and certainly the most strongest homes for serious textual criticism. In fact, today, when we talked about in that same episode, how research papers come about and the fact that they're peer reviewed and criticized, that's key, because you're able to kind of go through some of the things in the paper that make it strong and make it reproducible.
The reason we do that is so that we can build on knowledge and before we were able to have something like this, a place where people could go study and take the entire breadth of knowledge, and then build on top of it and then keep doing that generation after generation. That's how you're able to springboard knowledge.
That value is that scholars could gather there and they could hyper-focus their knowledge so that a succeeding researcher could then build on top of it. And according to legend and our namesake, the Syracuse and inventor Archimedes invented the Archimedes screw, a pump for moving water while studying at the Library of Alexandria.
But what made it, again, interesting was that this was a parallelizable process. Now, you didn't have to start over again or even have one person moving knowledge. You could have people from all over the world come here, bring
their knowledge and in parallel, in different directions, move human knowledge forward.
Now, I want to also mention something and that is the value of peer review. If you remember in season five, we talked about the AI winter, and that was what happens when we don't do that. If we just believe everything we read, like for example, the criticism of the perceptron and we just kind of took that at face value and effectively halted all research in neural networks in that area.
It caused decades long, an AI winter, which prevented us from moving the ball forward in that. So that is certainly a lesson learned. Don't believe everything you read. Be critical of it and take a scholarly approach to things. Now, the one thing I want to say about cutting out research.
Obviously, if you cut out research and you go straight to application, you're either going to spend too much time and money getting to some mediocre solution or you're just going to rediscover something that's already been discovered. Tread ground that's already been tread. If you cut out application and you only do research, you're forever going to be in the pursuit of perfection.
So what happens in both of these cases? You waste time and you waste money. What else can we learn from computer science? I'm going to talk about three perspectives. First, I think there's the personal perspective. For example, did computer science make my life better or did computer science make another person who decided to become a computer scientist’s life better? I think so and I think they're grateful for finding a job they can do.
For instance, who would have ever thought that you can make the living just pushing buttons on a computer, that people will pay you to do that. But I think what I mean when I ask, can computer science make our lives better, we're talking about something different. Maybe something collective. That's another perspective.
Can we make the world a better place in ways that are hard to measure? For example, one of my partners was telling me about a project he was working on that detected birds around windmills. And if a bird that's a particular endangered species comes near a windmill, it could slow down or stop to make sure that the bird doesn't get harmed. Does that make the world a better place? Certainly.
And we can apply that to many other things—energy, general health. Which takes us, I think, to the third category and that is, our individual perspective.
Can we improve another person's health? Can we ease their suffering? Can it make us happier? We spent a lot of season one in this area here.
In episode four, I had a chance to speak with an old friend, Andy Lee and Andy and I talked a bit about a way of getting actionable information to change our lives. A way of having computer science predict our biological age.
This is what Andy had to say about that paper and the research that they're doing in that regard.
Andy: Greg Fahy is the first author and I actually had the opportunity to see him speak in a longevity webcast, just this week and got to ask him a couple of questions, which was pretty exciting. He started a study using growth hormone as a treatment to regenerate the thymus. And from that, reverse the biological age as measured through this epigenetic clock.
And then another approach looking at plasma and some markers in the blood other than the DNA methylation that can do the same thing. So through this, they were able to show reversal of age. So people would come in, you calculate their baseline by looking at their biological age minus their chronological age to figure out are they above or below coming in and then track them over a year of treatment and see are they further above or below their chronological age at the end?
And they were on average able to shift people. Reduce their biological age by about two and a half years over a one year treatment plan. So really turning back the clock and making people younger throughout this pretty exciting concept that you may be able to take a treatment and become younger.
Angelo: So computer science definitely was able to make a change in some people and even reduce their biological age—effectively turned back the clock. So in that case, computer science did change their lives, but notice what it can't do. It didn't make the decision for them. It wasn't a magical pill automatically.
It still requires us to do something. I think that that's the key that we should expect out of computer science.
In episode six, I had an opportunity to speak with a friend of mine, Jim Shalaby. And we talked about how COVID-19 changed healthcare forever. Telemedicine had been there had been around for decades, but it really hadn't taken hold for whatever reason, mainly, because of privacy, computer science, and interpersonal issues of trust.
But when it was forced upon us by COVID-19, things changed. Notice what Jim said.
Jim: Simple things like that really lessen that anxiety on the clinical side as to whether they're getting reliable data or whether they're going to be legally responsible for monitoring that data on a regular basis. It wasn't as bad as they thought.
It doesn't substitute for face-to-face, but boy that's convenient. That level of accessibility of the patient having access to their providers, certainly they don't have to wait in a waiting room, they don't have to find transportation to get there. They have almost immediate access to their clinician without going through the waiting room issue. The clinicians also liked it because if they had a quick question, if they needed a quick followup, they knew they could actually engage with patients without having to play phone tag.
They can schedule something and do it. So, I think it's now become a part of healthcare. COVID has really changed the way patients and the clinicians. And the reason I say clinicians is that I don't just include physicians. I include the whole allied staff that supports a physician, the nurse practitioner, the physician’s assistant, the pharmacist, the dietician, even the case manager and the social worker are using the same modalities to communicate with the patient and they've all gotten into a rhythm. It's not perfect, but they've incorporated it into their daily workflows, which is really promising. It opens up new avenues for improving health and really the primary thing is it allows a new way to assess outcomes of interventions.
Angelo: Again, computer science wasn't a magical cure for anything, but it was a tremendous enabler for us to be able to do something different and be able to still have really high quality care.
Like artificial intelligence and artificial neural networks. We find that they become so complicated that they become difficult to use, or they have side effects. For example, in episode eight, we talked about the explainability of machine learning and how hard of a problem it is.
I was joined by Nikos Myrtakis, a PhD student at the University of Crete. We talked a bit about artificial neural networks and how hard it is to explain complex models like those and SVMs and others. Notice what he said.
Nikos: So it is very easy to understand why the system came up with a specific decision. On the other hand now, we have more complex models, nonlinear,
highly parameterized models, such as SVMs, such as neural networks, random forest, etc. These models are not interpretable by nature. So what we do is to explain them in a post-hoc fashion.
Angelo: So while expert systems tend to have a little bit of explainability built into them, these other systems don't, but that was a fascinating solution using a kind of neural network that is more explainable to try to reverse engineer and figure out why a different machine learning model made the decision it did. But it's not without its problems.
In that episode, we talked about five problems that are classically part of explainability and AI, but two of those problems I'll talk about right now. One is privacy of data. Nikos explains.
Nikos: Because to explain anything, we need access to the original data. So imagine that you have a database and you're provided with a model that you get from a different company. So we give the company the data in an encrypted manner.
Another very subtle issue is the adversarial attacks. So imagine that interpretability might enable people, or programs, to manipulate the system. So if one knows that by for instance, having three credit cards can increase his chance of getting a loan.
Angelo: And that's another problem I hadn't thought about. What happens when someone tries to reverse engineer the system to kind of scam it. Whether it's the individual whose credit is being pulled or some other third-party system trying to learn how the algorithms work so they can manipulate it.
In addition to those privacy concerns there are many more, not related necessarily to AI, but to our data in general. For example, in episode seven, I had a chance to speak with a friend of mine, Manos Athanassoulis, who's an assistant professor at Boston university where he's researching trying to solve problems like, how do you make sure your data is actually deleted from a database? He explains it this way.
Manos: We're trying to solve a very specific problem, which is, when you task your data with a data provider in the cloud, you have rights based on recent legislation. You have the right to ask them to delete your data. And if you ask them to delete your data, then they have to comply within a specific amount of time. They cannot just keep your data forever because it's not theirs. It's your data, right? When you generated something, it's your data.
For anybody who's familiar with LSM trees, which is not necessarily, you know, thousands of people in the world, but there's a few. Essentially every we delete we insert the tombstone, but eventually we'll meet the corresponding file that contains the invalidated data. And then we'll discard this data. And this merging is, which is called compaction, happens over time depending on how much data you have and how much data you're actually getting into your system.
So what we do is that instead of actually propagating these compactions with random order or with an order that depends on, I think which files are more frequently accessed.
I think the first one we call it lithy, by the way, the deleted related paper, we call it lithy, which comes from the Ancient Greek word which means forgetfulness. And the goal is to forget things.
Angelo: In all of these examples, we saw either practical application or theory. Many times he saw them together. Speaking to professors like Manos or research Nikos is doing, or even Andy, taking those papers and applying them with real code in real situations, we see that computer science can make a huge impact in our lives.
And in fact, those are not the only ones researching, either creating research or applying research. New research is being created in all different directions on the same issue. For example on privacy, we don't just worry about deletes. There's research being done in all aspects of privacy and all aspects of machine learning.
And imagine if we could take all of that together and start putting it into cohesive systems. Taking this piece of what Manos is working on and putting that in our installation of rocksDB and then maybe adding something else from over here and putting that into our implementation and taking this knowledge and creating something new.
We also, I think, had a lot of cautions. In addition to the AI winter, we talked about a little while ago. There were so many others. In episode three, we talked about the problems of micro-optimizing. Instead of looking at research and doing the hard work, kind of pulling that into implement it we sometimes get afraid of it and we get lost in finding a library or trying to micro-optimize something to get a mediocre solution.
Or, kind of being stuck in an old way of thinking. In that episode, we talked about designing for performance instead of tuning for it. Tuning for it is the way
we used to think decades ago. Now you can't build a solution then tune it later. Now, that's not to say you can't build a solution and then optimize it, but it's to say that do not expect to get big gains.
You have to start with a technology and an architecture that lends itself to the problem at hand. You can't just use some general purpose technology and then expect to tune it for big gains.
In episode two, we talked a little bit about the dangers of artificial intelligence. Here’s a quote that I really like it. I'm going to say it again that, “all machine learning models are bad. Some are less bad than others.” And that's because we build bias into these models. And that, we talked about a bit in episode two, we also talked about the bad side of technologies like social media that are supposed to bring people together.
But in reality, they don't, they don't make us happier. They make us a bit antisocial. So I think from that, we learned that you cannot foresee what's going to happen. Sometimes there are unforeseen side effects to technology even though we might be well meaning.
I'd like to add one more lesson to this, and that is, think about gratitude. The people that have gotten you where you are. You may have met people like this when you went to grad school or maybe you've met them in your day-to-day job and you may have met individuals, maybe you're building a company.
And the people who have worked with you and helped you get where you are, the mentors who have taught you through the years, your employees, take a moment to make sure that you have gratitude for those people around you. That will really go a long way. And in fact, I think that that's the right way to lead.
In episode one, I think I said that Chios Island, which is where my family comes from, has a long history of entrepreneurship. And I mentioned that there's a story there for another day. So I think it's time to make good on that here. Here's the basic story and it's going to take you back over 150 years to March 1822.
In retaliation to a Greek revolt against Turkish occupation, the Sultan sent Kara Ali Pasha, an Admiral with 40,000 Turkish troops to the Island of Chios. They set at a fire and they killed all infants under age three, all males, 12 and older and all females, 40 and older, then they took 52,000 slaves and they murdered another 52,000. The population of the island was only 120,000. Only 2000 were left on the island. 21,000 managed to escape and there were probably about a thousand that were out of the island because Chios was a seafaring center. So
they had lots and lots of ships and they were in places like Marseilles, London, and Liverpool.
And so they happen to be in these areas. So some of them ended up settling in those areas and when they did it, they became members of that community and entrepreneurs there. They restarted and rekindled. For instance, in London, F Sotros Raleigh laid the foundation stone in St. Sophia's cathedral and he and his brothers formed the finance committee. Another famous Yoti moved to Marseilles with his parents and his son then formed the Baltic exchange in London.
And the story can be repeated many, many times until 1832 when they moved back to Chios and kind of started rebuilding for about a hundred years until 1930s. And when the Germans came in and occupied Chios and they went through their massacre and so then they had to, again, rebuild.
You have individuals who are entrepreneurs and they continue that legacy and rebuild. The reason I say this is because I think the story of Chios is one of rebirth. Adversity, sure, but then landing on your feet and starting over. I think that that has affected me in several ways. For example, I distrust five-year plans. Look at COVID. What happened to everybody's five-year plan? Instead, I think it's really about opportunity, about creating doors of opportunity, and then later deciding which ones to go through.
I'm not saying we never plan, but what can you foresee five years in advance? It's actually really hard to do that.
Season two is going to be about opportunity. What is on the horizon and what should we be paying attention to. We're going to be talking about, how do we improve existing techniques? For example, we talked about expert systems and rules engines. We've done a ton of those. How can we improve that? How can we continue to protect privacy while threats are actually evolving at the same time. How can we make better and faster decisions in a world where threats to healthcare and other things appear out of nowhere like COVID-19 did.
And now we're having to kind of catch up, how do we get ahead of that? And then we're going to run against barriers of technology. For instance, Moore's Law is coming to an end. What do we do about that? So there's two sides to that. There's a, what's happening in the short-term. How do we get past this barrier to the next?
And then how do we blow away all those barriers with moonshots like quantum computing? So going back to my theme of gratitude, I wanted to make sure I express gratitude to all my listeners. Thank you so much for joining us on this journey. We really want to hear about your thoughts. The show is evolving just as the world is and we want to make sure that we're covering topics that you're interested in.
And we also want to know what is it that you want to deep dive on? I'm your host, Angelo Kastroulis and this has been Counting Sand. Please take a minute to follow, rate, and review the show on your favorite podcast platform so that others can find us. Thank you so much for listening.