← Leading Organizations That Matter
Loading summary
Transcript50 lines
- [00:00]
Ray Spadoni
Foreign.
- [00:10]
Dmitry Charovic
Welcome to Leading Organizations that Matter podcast about leadership, organizational culture and how we find meaning and purpose in our work. I'm your host Ray Spadoni and today's topic is Artificial Intelligence Primer an interview with Dmitry Charovic have you seen the newest ads from Apple on their recent introduction of Apple Intelligence, their version of AI? They are strange, which I'll explain during this episode, but if history repeats itself, once Apple gets involved in something, it becomes accepted and adopted by the culture at a very high rate. Not that anybody needs Apple to make AI the next big thing. It already is. But I thought I would spend some time on the podcast speaking with an expert on the topic with the aim of breaking it down into smaller bite sized parts so that leaders and supporters of mission driven organizations might have a better sense of what it is and especially how it could impact them and their organizations. Dimitri is a member of the faculty at Suffolk University in Boston where he focuses on AI, machine learning and business analytics. He is working, as I mentioned during the interview, to bridge the gap between academia and industry and this means helping students and faculty understand what it is and how best to be prepared for for this quickly changing world. Dmitri has worked for his entire career in this field and he serves as a founder and CEO of AnyQuest, a company that specializes in generative AI platforms. His career includes leadership roles at prominent tech companies such as Progress Software and Computer Associates. He holds an MBA from the MIT Sloan School of Management and a Master of Science from the Moscow Engineering Physics Institute. Let's get into it.
- [02:23]
Ray Spadoni
Hello Dimitri, thank you for being on my podcast.
- [02:27]
Dmitry Charovic
Hey Ray, great to be here.
- [02:29]
Ray Spadoni
Awesome. Appreciate it. Could you talk a little bit about yourself? I gave listeners a quick overview of your background before we hit record, but I wonder what was it that led you to focusing in on artificial intelligence and how does a person become an expert in AI?
- [02:52]
Dmitry Charovic
Well, I have actually been doing AI for a very long time. I studied AI at school in the 80s if you can believe it. At that time AI was very different, but I actually was working with Mikhail Batvinnik. He was a world chess champion on one of the first chess programs and that was AI at the time. And then yeah. And then just throughout my career I worked at large companies, small companies, startup companies, and they were always AI related.
- [03:19]
Ray Spadoni
Great. Well again I appreciate you being here and in a bit I'm going to ask you a little bit more about AI just from a let's define it. You know what, what was it back when you got started and we're working on the chess programs versus what it is now. But given the aim of this podcast, it's really focused on leaders of mission driven organizations, typically nonprofits. And that would be healthcare providers, social service agencies and other kind of community based, community facing organizations, ones that fundraise, ones that take a broader view of the needs of the folks they serve and so forth. So based on the podcast I've done so far and the reaction to it, I, I have a sense that there are going to be some leaders who listen to this who think about AI as, you know, a curiosity, but aren't going to spend too much time thinking about it. Others are going to figure it's kind of an irrelevant sideshow to what they do. Others may say, hey, this is a great opportunity, this is the future and this is something we need to capitalize on. And then there are probably going to be some who think of this and associate it with threat peril, something that's going to harm themselves personally in their careers or potentially their organizations and their missions. So given the fact that our listeners may be all over the gamut on that point, what should they be thinking about this?
- [04:52]
Dmitry Charovic
Well, I think AI is truly transformative technology and it has been that kind of technology for the past 30 years, but even more so now. Like you asked me what AI was when I started, and maybe some of us remember at that time it was rule based systems, expert systems. So somebody had to sit down and define a set of rules. If this happened, then to this, if that happened, then to that. And then based on those rules, the computer could interpret them and then act in an intelligent manner. That was the state of the art then. And expert systems were very powerful and they from certain function, but ultimately not very successful because defining those rules was a lot of work, right? We had, somebody had to sit down and build the rule system and then they were never complete. There was always a rule that you missed. So companies, insurance companies, healthcare organizations, they, they now have expert systems and rule based systems that have literally thousands of rules and they're very difficult to, to maintain and manage. So that was AI then. And then the, the next breakthrough in AI was the so called machine learning. So somebody brightly said, all right, why do I define all these rules? I have all this data, I have all this patient data, I have all this customer data. Let me design an algorithm that I can give all this data and have it come up with the rules. So that is machine learning and that is what most of us were recording AI for the past, I don't know, 15, 20 years, machine learning, business analytics, where you can feed a lot of data to a system and it will come up with the rules, and the more data you give it, the smarter it becomes. Right? So that was like the second part of the AI evolution. And then what happened just recently is that we learned how to model a human brain. It's called deep learning. So there are these artificial neurons that model biological neurons that we all carry in our brains. And so we learn how to put together billions of those neurons and create something that thinks like a human, talks like a human, and learns like a human, and makes the same mistakes as humans. So this is the new iteration. And I guess what happened recently is that we have achieved the breakthrough in how smart these artificial beings become. Like, they can read text, understand text, they can generate text, and they are now not very expensive. So any company, large or small, can now have thousands of these semi intelligent beings participating in the business process. And that is transformative.
- [07:29]
Ray Spadoni
So is that the deflection point in this trajectory, the deep learning, the fact that it's been made available to the masses, if you will, because automation, the fast pace of technology improvement and so forth? I think everyone's quite aware of that, certainly in the healthcare industry in terms of how we organize our care, how it's provided, how it's funded, how we regulate. There's been a great deal of advancement over the decades and certainly over the past 10 years, but it feels like something's different over the past couple of years. Is that it? It's this deep learning piece?
- [08:09]
Dmitry Charovic
Yeah, I would say it is the general nature of the models. Like, you could always train a deep learning model, but it would be very specific. Like it could read X ray images and tell you if somebody has cancer or doesn't have cancer. But you had to spend, I don't know, hundreds of thousands of dollars training those models, and then they would be able to do just one thing, like diagnose a disease. But now with these models, they are generalist. They have read every book out there. They have read Wikipedia, they watched every video, like, watched every, listened to every podcast, and they know everything. And you can then customize them just a little bit to make them into doctors or marketing specialists or HR managers with just a little bit of tweaking very inexpensively. So I think that is the inflection point, this kind of big brain generalist model that can be inexpensively customized to do this or that work and that. And that makes it broadly accessible and very powerful.
- [09:05]
Ray Spadoni
Well, Certainly things like ChatGPT and the fact that Apple has launched Apple Intelligence, we hear about it a lot more. And so for the uninitiated, someone who doesn't have a strong technical background, you may be given over to wondering whether this is marketing. This is just the industry putting a new face on what has been the normal progression forward. But it sounds to me as though, no, there's actually something that has changed. It's the widespread availability and the fact that it's been made available, you know, via chat, GPT and other sorts of things to anybody.
- [09:46]
Dmitry Charovic
Yeah, that's, it's certainly, yeah, so many people can view it as just marketing and hype and there is certainly enough of that. Right. If you just, just listen to all the, all the materials out there and read all the posts on X and other places. But there is a core to it and the core is that this thing is incredibly useful. Anyone who spent some time with ChatGPT knows you can ask it any question and you will get an intelligent answer. You can ask it to draft an email and the email that comes back is a fairly good first draft. So it is, yeah, I think it's, in that sense it is truly a transformative technology. But then as always there is enough marketing around it that is emphasizing and deemphasizing certain points.
- [10:29]
Ray Spadoni
I mean, I think we've all experienced certainly with Apple that a technology will exist, but once they embed it in their operating systems and make it widely available and simple enough to use, then it goes to the next level. And at the time of the recording of this podcast it's, you know, we're just a sort of few months into the beginning of the rollout of Apple Intelligence. I'm not sure if you've seen the ads that they've been, that they've been showing. To me they're curious. I mean I'm, I'm used to Apple ads that are compelling and interesting and motivating and make me want to spend my money. But they are playing up AI as a quick sort of trick for the half hearted folks. People are calling it the George Costanza ads, you know, based on the character from Seinfeld, which is, you know, someone who's always going to do the least amount to try to get the most benefit. And so the fact that you can enhance your productivity and you can have it write an email for you and so forth, it's making me think that the rollout of AI to those who don't know that much about it is that people are going to only think of it as a shortcut. To doing real work. What is it really? Or tell me it's not just that.
- [11:52]
Dmitry Charovic
Yeah, so I think it's actually so yeah, I've seen Apple's ads and I have just as you have been underwhelmed by them and I think it is because Apple is viewing or trying to present AI as a tool. They're saying, you know what, it's a tool that makes you draft emails better. Like you, you are maybe not such a good writer, but with AI you can write a better email or you can correct the picture for you. You took this picture with this person, we can remove the person. So this is an approach to AI as a tool. It's like it's making what you have already been doing a little bit quicker, better, more efficient. And that's fine. But I don't think that is the, the ultimate promise of AI. I think the, the, the ultimate promise of AI is actually the ability for us to, to use AI as a collaborator. So basically as a worker, let's say as a software developer. As a software developer you enjoy writing programs. This is what you ultimately do. But then as you grow in your career, you start doing managerial tasks, more and more of them and then eventually you manage and do admin work 80% of the time and you write code only 20% of the time.
- [13:03]
Ray Spadoni
Time.
- [13:04]
Dmitry Charovic
And so what AI can do is, is basically become a collaborator that has taken that administrative load off your back. And so learning how to use AI that way, how to use AI as a team member, I think will be the ultimate, the ultimate achievement. And another thing is that AI is plentiful. You can have not one collaborator, but a thousand. So what do you do when you have a team of a thousand working for you? And that I think is truly transformative.
- [13:34]
Ray Spadoni
Well, as I mentioned earlier, none of us are strangers to automation. And the history of progress has given us the printing press that did something that the scribes used to do or the way that we make cars today versus the way they used to be made. This seems to be the first time we have a technology more aimed towards sort of quote unquote, higher end knowledge workers. Do you expect as you look up ahead this to be extremely disruptive in that regard for some folks who have been essentially shielded from automation?
- [14:17]
Dmitry Charovic
Yes, this is exactly what it is. It is automation of knowledge work. But, but there is a catch. This is not a replacement for human, this is a being that we have never seen before. So on one hand it is a very well read individual, this artificial being. They have read every book. But at the same time it is not very clever. It's making stupid mistakes all the time and it cannot plan for more than two steps ahead. It's a great actor. It can impersonate and make itself sound and look like almost anyone, but at the same time it is very gullible. You can trick it into doing almost anything and it lies all the time. There's this idea of hallucinations. It's not very truthful, but it is very convincing when it is not truthful. So there are certain limitations that we need to learn how to work with. But once we master that, once we are aware of those limitations, then we can automate certain tasks, but not all of them.
- [15:17]
Ray Spadoni
Well, it's a good caution to maybe take the output with a bit of a grain of salt though. It sounds as though much of the effort is to minimize variability and error.
- [15:33]
Dmitry Charovic
Yes, exactly, variability and error. But the important part is that you cannot eliminate it. There will always be variability. It's very random in the way it works, the way it behaves. The foundation of it is that it is speaking and writing text by predicting the next word. And then so it's just creating a selection of the most likely words and then randomly picks the next one and that word determines what the next word will be. So it is by definition very random. So just when you think, you know, you have controlled everything and you have made it more reliable, it does something unpredictable again. And just like humans, by the way, right? Humans are unpredictable and they make mistakes. And, and so it's learning how to work with this computer system, as is this very unpredictable but very useful, useful member of a team.
- [16:27]
Ray Spadoni
Are you a sci fi fan? Do you, Are you, You know, I mean, the next, the next step would be when it becomes self aware and then that becomes the end of, you know, civilization. And so many movies have been, you know, and plots have been, have been, have developed around that. It's, it's kind of humorous to think about the fact that we are entering a stage where the technology is becoming, you've said it a few times already, more human. It's, it's, it's kind of fascinating and, and strange at the same time.
- [17:01]
Dmitry Charovic
It, yeah, it's, it does sound, it makes it sound human. Like it's very good at mimicing and acting. So it, when it communicates with people, it makes sound itself like a human. So that's certainly what it is. But I would say the level of intellect is not yet as powerful as humans and it may never be.
- [17:23]
Well, I know that you work at.
- [17:25]
Ray Spadoni
Suffolk University in Boston in the various programs there and that you are working to as your bio states bridge the gap between academia and industry. And I think there are many in academia who are wondering about the impact of AI on their students and on learning in general. What's your take on that and what are you seeing at Suffolk?
- [17:53]
Dmitry Charovic
Yeah, so I think this latest crop of AI has been truly disruptive for education because the way they train these models and the way they evaluate these models is on standardized tests and they use the same tests that students use at school. So there is no problem or no homework that you can give to a student that the model wouldn't be able to solve. And that just changes the whole education process and has the disruption. And the education industry, the whole academia is not trying to figure out what to do with this. It's something that you have never seen before. And some schools are becoming protective. Like they're saying that if you use AI, it's cheating and because you're not learning, you're not getting everything that you could get from the education process. So you are not allowed to use it, period. But here at Suffolk, we thought long and hard about it and we decided that that's not the right approach because when students graduate, they will be expected to use AI and then we'll be expected to manage AI and invent new use cases for AI. So what we should do instead is prepare them to be productive with AI and know all the drawbacks are using it, know all the advantages thrown at big points. And so the decision was made to re engineer the education process here at school with AI in mind. So change the curriculum, upskill faculty, you know, make the tools available to students and all the business processes that we teach in marketing and finance, they will be taught with AI in mind. Like how would you perform this work when you have an AI collaborator?
- [19:41]
Ray Spadoni
So it's, you know, rather than viewing it as some type of a threat or problem, it sounds as though you're viewing it as it's the world. And so let's get you ready to live in this world and to be successful and manage and lead in this world. Which sounds more forward thinking. You know, I recall when calculators first came into schools and there was the debate of, well, you know, if, if we let students use calculators, then they're never going to learn math. But I think reality was that we allowed our brains to be used for higher order or different tasks than the fundamental computing tasks on a calculator. Is that a Reasonable analogy for what you see happening with students is that that some tasks may be replaced by AI, but in reality we're just going to allow ourselves to do other things instead.
- [20:35]
Dmitry Charovic
That is a very good analogy. So calculate is one of my favorite analogies. But also handwriting. It used to be that schools would not allow you to use a typewriter typewriter because apparently handwriting was essential. So at every step in the technology evolution, schools tried to try to be conservative. But I think it is just given the pace of everything accelerating, it is important to give students the skills they will need. And the main skill, I would say is that of. As any manager would. Right. For business school in particular, it's important that there's delegation, there's knowing what task to delegate, how to articulate the assignment, how to check the results. So a student that knows how to work with AI is also a good manager, I think.
- [21:22]
Ray Spadoni
Okay, before I ask you, Dimitri, about where this is headed and the future, let's not look up ahead just yet and let's just think about the current situation. Obviously what you mentioned before about the sort of access and the, you know, the inflection point as you described, it has happened. What else is currently happening that you know, the leaders of the organizations who listen to this podcast, what's happening right now that that folks should know about?
- [21:56]
Dmitry Charovic
Well, I think the. There are actually two phenomena that are happening. So one of them is that we are approaching, I think the limit of performance for the traditional transformer based models like GPT or, and Thrombotic and others. So it's becoming more and more expensive to get to the next level of performance. So we are kind of flattening out. But at the same time, the cost of compute, the cost of intelligence has been driving precipitously like it's. You can now spend a dollar and analyze all of Orcs of Shakespeare. So this kind of flattening of performance and then the, the reduction in price I think will create an environment where it will be very inexpensive to use AI at work and that will create new opportunities. Like for example, for mission driven organizations that you mentioned, you can assign a worker to every donor that you're working with, like a virtual worker that will report to them, that will collect their requirements, maintain the communication channel. So this is just one example, but there are just many opportunities to use AI to enhance and optimize whatever you're currently doing.
- [23:08]
Ray Spadoni
Okay, as you think about the next, let's say five years, what are you expecting? We're going to see how is this going to continue to evolve forward. And is anything revolutionary on the horizon?
- [23:25]
Dmitry Charovic
Yeah, obviously no one knows what will happen in five years. I don't think we will have. Well, maybe we will have something that would be close to AGI, artificial general intelligence. But even if we don't, we already have. Like, in the past two years, we have created enough that will last us the next five years just to learn and understand how to use it more efficiently. So I think that will be the focus of the next five years, figuring out how to use this technology productively in a way that is safe, ethical and efficient. So that is what we will be doing?
- [24:00]
Ray Spadoni
I think so. Is that AGI, Artificial general intelligence? Could you just define that for us?
- [24:07]
Dmitry Charovic
Yeah. So it's artificial general intelligence. It's intelligence that is the same or exceed human intelligence. And I don't think anyone can even measure it. We know it exists, and so we think that so many people hope it will be coming sometime soon. Like, if you listen to AI company leaderships, it can be like next year or the year after. I'm not as optimistic simply because I don't see the technology step that needs to happen in order for that to happen. But I think the models that we currently have, they will become more powerful and less expensive, and that is enough. It's like electricity. Electricity is more powerful and less expensive and omnipresent. So the same will be happening with intelligence, and we just need to learn how to use it in different domains.
- [24:57]
Ray Spadoni
Are you worried about anything? I noted your inclusion of the word ethical as you were describing. You know, what you're, what you're hoping for and what you're expecting. What are you worried about or what are the risks here?
- [25:13]
Dmitry Charovic
Yeah, so I think there is a, there is tendency, or not the tendency, a risk of delegating too much to AI. For example, I don't think AI should be making hiring or firing decisions because that is something that to me would be unethical. And AI cannot be held accountable for any decisions that it makes. So there will be always a realm of decisions and actions that only humans can do. And so that is ethical. But then there are other risks. Like there are biases in models. There is all kinds of things that people who deploy AI need to be aware of.
- [25:54]
Ray Spadoni
Well, you know, human nature being what it is. And going back to the George Costanza, you know, reference about, you know, folks taking shortcuts. I'm, I'm, I'm interested and curious about your notion that you don't want them making hiring and firing decisions. And you know, maybe there are other things that folks will task AI with that go up against or over the boundary of appropriate. How do we regulate this or how do we put governors or boundaries around this so it doesn't get out of hand? Or should I not be thinking about that? Is that too restrictive a way? Might that close us in and prevent progress?
- [26:39]
Dmitry Charovic
I think you're totally right to think about it and I think everybody should be thinking about it and the way to address it. This is so new. I think everybody and the society overall is learning about this technology. So as I guess as we explore different use cases, we should be also thinking through the risks and consequences. Here at school, for example, we have AI task force and one of its objectives is to define ethical guidelines for how AI should be used. And I think every company should have one. But then more broadly at the government level, you know, local and state and federal, there should also be some, some process. It's a fine balance. You cannot be too restrictive because then you may not let the technology blossom to its full potential. But yes, it's a fine line, fine balance.
- [27:35]
Ray Spadoni
I would imagine that the task force there, one of its duties is to just stay abreast of, of these issues and what's happening and so forth. So four leaders of mission driven organizations. I mentioned up front that they may run a gamut from this is the future to I'm not too curious about this. I don't think it's relevant to us and everything in between, given how much information there is out there. It seems suddenly about AI. How can folks who lead community based mission driven nonprofits, how can folks stay abreast of what's happening so that they can become informed and can refine their perspective when it comes to AI on behalf of themselves and their careers, but certainly and most essentially for their organizations and the mission of their organizations.
- [28:35]
Dmitry Charovic
Yeah, I think it is, it is a matter of taking, accepting the culture of learning and experimentation at this point. So I would just recommend everyone to, to get a copy of ChatGPT or access to ChatGPT or Claude or one of those wonderful chatbots and just starting using it as, as a tool initially not as a collaborator, but at least as a tool and just learning all these strong and weak points about this technology, but then also to create a culture of experimentation and sharing within the organization. So make the technology broadly available to everybody and then maybe have a weekly, weekly show and tell where everybody can be sharing their experiences with the technology.
- [29:20]
Ray Spadoni
You know, many of these tools are free for the general Public, you just mentioned ChatGPT. Is that likely to continue or are, is there going to be an effort once this becomes more widespread to monetize it and put some of the features behind paywalls and to subscription models? And are we going to be subjected to having to pay more for what we ultimately become pretty dependent upon?
- [29:46]
Dmitry Charovic
I don't think the current technology, the current capability will be expensive in the future because there is too much of it and there are too many vendors. It's capitalism. There is competition. So the prices will be, will be checked through the competition mechanism. But new capabilities, like if a company has a breakthrough, for example, OpenAI has a wonderful voice interface where you can communicate with a chat bot is just, you know, by speaking and listening. And so while they have advantage over other companies in that they will probably charge, they can charge extra for it. But then, so that's one. But then also once you start, like there is one thing about these models is that they're frozen in time. They don't know anything about you or your business or your customers. So there is a huge value in connecting large language models to your data systems and teaching them to follow your business processes and rules and guidelines. And that is a lot of work, expensive work. And I think there will be actually a new set of people, AI whisperers who will be training models to work in your business environment. And so that is where a considerable budget can be spent eventually.
- [31:07]
Ray Spadoni
So for an end user, say someone's running a moderate sized healthcare company or a community hospital, do they have to concern themselves with these advancements or can they expect their business partners and vendors and technology, you know, partners to just incorporate the features, as you just mentioned into their solutions and that they will evaluate them as part of the normal core, you know, course of business.
- [31:36]
Dmitry Charovic
I think that that is what will happen eventually is that every product that they currently use, Epic, you know, or Cerner, they will have advanced AI features built in. So that will certainly come at some point. But I think AI and that is, that is approach to AI as a tool. It's just as Apple is making, you know, picture editing available on their iPhone, Epic will make some AI feature available as part of their system. But if you really want to take advantage of this AI as a collaborator, AI as a team member, I think there will be a new crop of systems, so new crop of solutions emerging to that because it is, they need to combine information and data from multiple sources. They need to reason in a certain way. So I think there will be value in creating that new Class of intelligent systems.
- [32:28]
Ray Spadoni
Okay, interesting. Well, Dimitri, thank you. This has been a good primer. I felt as though, I mean, in some ways this is a little off the beaten path for what I normally cover on my podcast, but there's been so much conversation about artificial intelligence, and I think that folks are wondering sort of what to make of it. And I thought it was important that we have a sort of a primer for folks for whom this may not be normally something they spend a great deal of time thinking about. But I feel as though we sort of covered the landscape pretty well. Is there anything else we haven't talked about, maybe related to leaders of businesses, mission driven organizations, non profits, anything we haven't talked about? You think that's worth mentioning?
- [33:17]
Dmitry Charovic
I think we covered a lot of ground. I would just encourage everybody to start using it. It's very easy to start. It is not expensive. Make it part of your small process and again, experiment and expand from there.
- [33:31]
Ray Spadoni
Okay, good advice. I'll confess that I've been a little reluctant, but I'm feeling motivated from this conversation to dive in and experiment and play around with it and see what it's all about. So I, I thank you for that. If people are interested to learn a little bit more about you, a little bit more about what you do, the, you know, the, the company, the companies where you've worked, or, or Suffolk University, how best for folks to find you.
- [33:59]
Dmitry Charovic
Yeah, like the old fashioned way. My name is rather unique. They can just google me and they can find me on the faculty page. All my contact information is there or they can just send me a message on LinkedIn.
- [34:11]
Ray Spadoni
Perfect. Great. Well, thank you so much, Dmitri. I appreciate your being on the podcast and I hope you have a great day.
- [34:19]
Dmitry Charovic
Thank you. It's been my pleasure to be with you.
- [34:22]
Ray Spadoni
Thanks for listening. I hope you'll consider leaving a five star review on Apple Podcasts or your platform of choice that will help others find us here. My mission is to help empower organizations that matter by supporting those who lead them. Feel free to learn more about me and my work@redsailadvisors.com.