Subscribe
Share
Share
Embed
Join the the Artificial Intelligence Student Organization, RAISE, at the Crummer Graduate School of Business and Rollins College, to discuss the reality and implications of AI in the business and academic world.
AI Podcast EP 3 - Isar Interview
===
[00:00:00]
Intro Music
---
Casey: Welcome to the AI Across Campus and Beyond Podcast. Today we will be talking to you about how AI will affect our future. We have Isar Mateas, CEO of Multiply. He is an accomplished serial entrepreneur and AI educator.
Pristine: I'm Pristine, your co host and co president of the AI student organization RAISE. I'm self designing a major in artificial intelligence at Rollins.
Casey: And I'm Casey, your other co host and the graduate side co president of Raise. I work at the AI Edge Center under Dr. Liebowitz, and I'm currently pursuing my MBA.
Pristine: Isar, could you tell us a little bit about yourself and your background, and how you first became interested in artificial intelligence, specifically Gen AI for productivity and innovation?
Isar: Sure. Uh, so my [00:01:00] background, I run software companies for almost 20 years. That's most of what I did most of my career. I've been in several different startups. Uh, some of them, I joined early.
Some of them I founded and was the CEO of, I sold one of my businesses, to my business partner. So my half to my business partner at October of 2022 and then thinking what I want to do when I grow up and chat to PT came out and as a geek and somebody who's interested in tech and businesses, I started playing with it very early on, like just as it came out and I became obsessed with the Possibilities for businesses and I'm an investor in several businesses and I'm on the board of several businesses and I, uh, have a lot of friends who are CEOs in different businesses.
And so many of them started approaching me or me talking to the CEO of the companies that I'm an investor in and very quickly started doing a little bit of consulting on how to use these tools. One of the businesses that I'm an investor in is a cohort based training [00:02:00] platform. And I had a conversation in January of 23 saying, Hey, listen, we got to start offering an AI course on your platform.
He's like, Oh, that's a great idea. Do you know anybody that can teach it? And I said, I'll teach it said, you, you're, you know, you're on my board. You, you had two successful exits. Were you going to be a teacher? And I'm like, yeah, absolutely. I love teaching. It's like, okay, you, you got the job. And so in April of 2023, I launched the first AI business transformation course.
I've been teaching it at least once a month. Since in many months, I've done two of them. So hundreds of business executives have been through that course in the past a year and a half. I just finished one yesterday. So every Monday there's a course. And like I said, if there's two, then Mondays and Wednesdays, there's, there are courses, but I, this is most of what I do right now.
I do training and I do consulting AI successfully.
Casey: Very cool. What are some of the common challenges you see when [00:03:00] working with these different companies and organizations?
Isar: So there's, that's a tough question. Uh, there's a very wide range of, of issues that companies have when they come to implement, uh, AI in their businesses. And I work with companies as small as 20 people and as large, my largest clients is 45, 000 employees, every place around the planet, you know, X number of billions of dollars in revenue.
And so every company has different challenges, but if I have to. Kind of like put them into buckets, the, and I'll start with smaller companies because they have less red tape and issues. The biggest problem with smaller companies is knowledge, right? So most people don't know what they don't know. And they now, hopefully by now understand that they need to start implementing AI, but they don't really know how to start.
And they're struggling with how do we provide training? What tools do we need to bring in? What's even the process to do one or the other? And so that's the biggest challenge, because if you don't, you don't have a [00:04:00] plan, it's very hard to execute on it. And they don't even know how to start the plan because they know nothing about this.
And I'm talking about people who are CEOs of successful companies, hundreds of millions of dollars, employees all over the world, growing great business people. But it's just a completely new arena that they've never dealt with before, which leads to kind of like the other problem that people have day jobs, right?
So everybody regardless of whether they're in HR or marketing or data analytics or product development, et cetera, they have a job and they have fires to take care of every single day. And they're always behind on schedules and their departments are never big enough and so on and so forth. And so with that.
in mind, they don't have bandwidth. Even if somebody will task and say, okay, I want you to be, uh, in charge of AI implementation for our company. They just don't have the time to do that. And so they will keep on going back to the stuff that they know they have to do, because that's their [00:05:00] job and they get evaluated on it.
And their bonuses are tied to that. And their progress in the company is tied to that and so on. And so it just falls between everybody's shares.
Casey: Seems like a big lack of clarity, um, in terms of what to do. And I mean, you mentioned it's a brand new field. I'm, you know, in my MBA program, we have to look at internships and I've had two conversations so far where they say, we really want an AI internship. We just don't know what that is. We know we want one.
So it's, it's definitely a common theme that everyone wants it, but no one even knows what they want.
Isar: a hundred percent. And so I think knowledge is the number one thing. Number two is, it's a mix of two things, right? So in the larger companies is data security. So larger organizations, and specifically if you started getting into fields like finance or healthcare or legal, where there's a lot more regulation and red tape to begin with, they have an even bigger problem.
But in general, bigger [00:06:00] organizations have more concerns about data, and so their fear about their data leaking out and how to do this safely is a big concern for larger organizations. I, A, totally understand the, the fear that's involved in that, but B, I must admit that it's a little funny to me, especially if you do this properly, and I'll explain.
If you think about some of the main platforms that are running today as far as AI, so let's say you go with, uh, Microsoft and use. Copilot, as an example, where your information is already in Microsoft servers, you're most likely using Microsoft 365, you don't have a server in your back room that you host all your information, it's sitting in a cloud computing somewhere.
And so adding another layer on top of that makes it much easier. No big difference on your data. That being said, you need to verify, or if you're in a regulated environment, verify even deeper or more. Where does that data go? Is it a trusted [00:07:00] company? Uh, can we use this data versus that data with that tool?
Will it train on our data and so on? So there's questions that definitely need to be answered, but I think too many companies are, you know, Using that quote, unquote, as an excuse not to get started because they don't want the data to be jeopardized. The interesting thing about this, that the statistics is as of from several different surveys from the largest, you know, the McKinsey's, Ernest and Young's of the world, KPNG, these kinds of companies that about 70 percent of employees bring their own AI to work.
So if you don't provide employees, the tools and the AI frameworks and guardrails and systems to use, they're going to use it on their phone. And then you have zero control on what's actually being loaded and no knowledge where the data is going. So the right solution is not to tell employees, okay, we're going to block all the IPs for any AI tool, but actually the other way around.
Find the tools that are applicable for your use case, for your data, for your level of security. Then allow employees and train them on how [00:08:00] to use it safely versus tell them they can't use it because they will find a way to use it
Casey: I really like the point you bring up where all of the data is already being stored in cloud storage. I mean, it's essentially still stored through Microsoft, but I mean, do you see a use for small language models and improving that data security? Or do you think that these large language models again? I mean, they were already storing the data on Microsoft servers.
It's already a cloud based society with cloud based data lakes that live in.
Isar: I think where the data lives and what kind of AI model use on top of it are two completely independent questions, right? So if your data is in the cloud somewhere, it's in the cloud somewhere. If you trust that cloud environment to keep your data safe and you can put a large language model, small language model, open source, closed source, whatever it is that you need for that use case within that environment.
So it's confined to that environment, that environment stays safe. Now, to do that, in most cases, you will need an open source model and there's multiple really good open source model [00:09:00] right now that you can run within the environment where your data is stored and the three big providers. So Google, uh, with, with their cloud solution, Amazon with AWS and, uh, Microsoft with Azure, all three of them provides.
AI solutions within your cloud environment. So you don't have to, and you can even pick from depending on which one you go with from three, five, 10, 20 different models that you can use. So from a data security architecture perspective, it doesn't matter where you have your data hosted. You'll have ample choices to pick from to use AI on top of your data.
The type of model is I think. sometime in the next 12 to 24 months will become irrelevant because where this is going, it's going to a multi layer agent approach where there's going to be some kind of an orchestrator agent, which is the only thing you are going to [00:10:00] communicate with. And you're going to define your task.
You're going to define your goal. You're going to define what you're trying to do, and it will go on its own and pick the right tools, the right models, the right extensions, the right plugins, uh, the right data. to actually complete the task. And so right now we live in a universe where we all know chat GPT and you go to the top left corner and there's a drop down menu and you get to pick.
And most people like, I don't know what to pick. Like, how the hell do I know which one is better? And the reality is we don't even people like me and I use these tools every single day, both for myself and for many of my clients in the courses that I teach and preparing for lectures that I'm doing and all of these things.
And there's many cases I don't know. And that's chat GPT. Then you have Claude and Perplexity and Gemini and Meta and like there's Lama and all these other models that are out there. So what I tell people is just compare, just test it out. And there are two tools that I use regularly to test new use [00:11:00] cases across multiple large language models or small language model, it doesn't matter.
So the first one is called, ChatHub. It's a plugin for, Chrome. That allows you to open multiple large language models at once and have a chat with all of them at once.
So as a one place for you to put in your prompts, but then you get the answers . And you do the same use case two or three times. You can immediately see which one works better for you.
So that's one option. The other option is a tool that I actually really, really like. I found out about it about two weeks ago and I've been obsessed with it since. It's called Hunch and the URL is hunch. tools, , but what Hunch allows you to do, it gives you this canvas where you can attach and connect literally anything you want in the AI world.
So you can. manually connect with lines, four, five, six, 20 different language models that they all have already connected to. So you don't even know, you don't even need your own API keys and all of that goes away, you don't need to know anything. So you open one thing that looks like a document that is your prompt, and then you connect it with four lines to four different [00:12:00] canvases, each and every one is a different language model, and you can do everything that you can do with those language models through their APIs, just with a really simple and easy to use user interface, and then you can compare the outcomes.
Pristine: So based on your experience with working with companies and small language models, um, where do you, how do you see AI being implemented, like, in the next 10 years or so?
Isar: Wow. Uh, I don't think anybody has the answer to that. I'm willing to make bets on 12 months. That's more or less as far as I go. And I'll tell you why these This field is moving crazy fast and yes, there's rumors that the pace is slowing down and maybe it's not rumors, but it doesn't matter because even the level we are in right now, the vast majority of the world still doesn't know how to use it.
And so even the gap between our current capabilities and where most companies are using it and individuals are using it is immense.
I think the concept of agents is going to be something [00:13:00] that's going to be the biggest thing in AI in 2025. Now, the problem is there's no clear definition of what agents are, or to be fair, there is a definition, but because it's now a buzz thing and everybody wants to have agents, everybody calls everything agents, even when it's not.
So. An agent is an AI system that can make decisions on its own, that can use various tools on its own.
So we'll pick the right tools for different tasks that can orchestrate nonlinear operations. So it can have four different processes run on four different other AIs or agents or external tools that are not AI in parallel to put them all together. So think about it like a project manager. Right? So a main agent is like a project manager that understands the goals of the project and now can build everything it needs on its own.
Like it understands the task and it says, okay, this is what we need to do. And to do that, I need one, two, three, [00:14:00] four, five, six, and I need these capabilities and I need these kinds of employees and I need these kinds of tools. And if it has access to all of those, it can execute on all of that on its own without getting instructions from us on what it needs to do, just like a good project manager would do.
And he will. Evaluate the tasks of the smaller agents that will take the small segments, the subtasks, and we'll go back to them if it's not perfect, and we'll report back to you on steps and progress and whatever, just like a good project manager would do. So that's the concept of agents, and it really has three different layers.
One is the AI layer, that could be the orchestrator or some of the operators. It has data, so what data do we give it access to? And this could be company data, it could be third party data, this could be public data, whatever data you want to give it access to. And then it has tools and the tools could be AI tools, but it could be Excel.
It could be searching the web. It could be, uh, writing code. It could be. So [00:15:00] the, the idea behind agents is that they work just like people do and they have access to different capabilities and so on. So that's the dry definition. What's happening right now, because it's a big buzzwords. Everybody's deploying agents and many of these agents are not really agents, meaning yes, they do multiple steps.
Yes, they can use multiple tools. Yes, they have access to different kinds of data, but they're quote unquote scripted, meaning you tell them how to do a specific use case. And there could be what ifs. It's like a decision tree. It's a flow chart, right? It's not a straight line, but it's still something you as a user define very quickly.
In a very structured way and you will follow it and we'll do an amazing job and you will do it in 1 percent of the time of a human or humans trying to do the same kind of job. So it's still highly valuable, but it's not a true agent from the traditional way of the autonomy aspect of this. So I think now going back to what we're going to [00:16:00] see in 2025, I think what we're going to see is we're going to see a big variety of.
Quote unquote agents with varied, varying ranges of autonomy it's still a tool that does a complex process on its own.
But you have to tell it what to do. Those of you who have used, tools like Zapier or make or any 10 allows you to create a lot more complex capabilities and connected to even more external tools that are beyond the reach of a GPT, but it's still scripted and they're awesome. I have dozens of these running my business, helping me with day to day stuff, and I think we're going to see more and more and more of these deployed in many different aspects of businesses.
many different aspects of schools and universities, many different aspects of our personal lives, helping us with shopping, helping us with trip planning, helping us with finding activities for our kids, stuff like that. Uh, and we're going to see better and better [00:17:00] agentic behavior and autonomous behavior doing these kinds of things.
Casey: You're saying like, uh, you're going to see varying degrees of human interactivity with these, these models, right? So do you think that there's any chance that agents that are more human centered might outperform or that as these models are released, there's a sweet spot between autonomy and human interactivity that might be best for business environments?
Do you think that as we release these varying degrees of autonomous agents, that there might be a more human centered agents that outperform fully autonomous ones?
Isar: We, we don't really stand a chance. These tools in their current form are better than most humans on many tasks. Not everything, and they still hallucinate and there's still issues and there's data security questions, like there's a lot of loose ends, but every day that passes, there are less loose ends and more benefits.
And as I [00:18:00] said before, even if we don't make any progress, like this is the final stage we'll ever get to with, with AI capabilities, once we figure out how to deploy them properly. It will take away most of the stuff that we do. Now. What are the caveats? Data security is one. Am I willing to trust it on a personal level with my bank account information so it can go book travel for me? On a business environment, am I willing to give it access to communicate with my clients directly, not through a filter of a human that's monitoring what it's going to say or do or send them? Right now, the answer is probably no. For most people. In two to three years, the answer is going to be most likely yes for most interactions, including our personal stuff. Why? Because it will run through my bank under the same level of security that I access my bank today from my phone. And so if I'm allowing [00:19:00] my bank to communicate with me through an app and I'm trusting that that communication is secure and I have an LLM that is secure, why won't I allow it to use that information that I have access to in order to book travel?
And so data security is one, two is consistency and predictability. They're currently not a hundred percent predictable and not a hundred percent consistent in a business, in some cases, not a big deal in some cases, unacceptable. So it just depends on what kind of business and what function in the business it needs to do, right?
If you are reporting on a patient's condition to a healthcare environment, getting it 95 percent is not going to cut it. But if you're writing a marketing campaign and it's 95 percent great, that's better than most marketing people, probably better than 95 percent of employees worldwide, which [00:20:00] means if you're in a business and you have marketing people, it's going to be better than all of them, most likely. And so there are cases in which it makes perfect sense today. There are cases in which you need higher predictability and you need consistency in the outputs. And the other thing that's coming is once you do a multi layer thing, you can always have the second step, check the first step and reevaluate it to make sure, or you give it to three different agents to run.
And then you compare the three and then say, okay, these two say the same thing. The third one said something else. Why? Let's go back to the data and check. Maybe they're all wrong, but there's a higher. Now, if you give it to 20 and 19 are saying the same thing, that's a pretty high likelihood that the 19 are correct. And the cost of that is still zero. And so it's not like, Oh my God, I need to give it to the 19 people now. And no, it doesn't cost, it costs sets to do every one of those cycles. And so I think in the long run, and when I say a long run [00:21:00] is two to three years, we will see more and more and more tasks. So not necessarily professions, tasks within the business world and our personal lives being taken over by AI agents,
Casey: I can see that. And , I do hear a lot of things boosting productivity. , do you think that there's a different approach for leveraging AI for innovation? , are these tools just as helpful with creativity, with new ideas as they are with the raw tasks and sort of the grunt?
Isar: a hundred percent. So I, when I teach my courses and when I work with the companies I work with on consulting, I tell them that the number one, the biggest value that these two will give you is brainstorming. So yes, you can get efficiencies on a day to day, but if you're thinking on the bigger picture, if you're thinking about making decisions as a manager, as a CEO, as an owner, as whatever the case may be, it's an incredible tool to do.
And what people don't know is you can customize them [00:22:00] to your needs. You can build a GPT. And those of you who are listening and don't know what a GPT is, it's like a mini automation, uh, that runs within ChatGPT and Claude has the same thing. It's called projects and Jim and I has the same thing. It's called gems.
Uh, but you can build these meet automations that you give them very specific tasks and very specific instructions and very specific data to use to help you do something. So you can give it data about your industry, about your competitors, about your niche, about the current stage of the state of the economy, about your role in the business, about the people you communicate with, about the projects that you're running on.
Casey: , you mentioned earlier they're going to eliminate more and more tasks, but not as many professions. , that kind of gives me hope that, , you won't be replaced or that we won't be completely replaced, because at least in that creativity aspect, we will. Hopefully, at least for two to three years. , I would hope a little bit longer than that. Maybe five or ten. We're still using artificial intelligence in a collaborative way, in a human centered way, at least [00:23:00] for these strategic, creative, innovation based decisions. And I'm hoping that the fully autonomous, the agent based approach is replacing more automated tasks, more of work that's already being outsourced.
, do you see that as wishful thinking?
Isar: the answer depends on the timeline that you're, Yeah. looking through, right? So if you're looking in the next two to three years, I agree with you 100%. I think the more creative tasks and the more thinking tasks that the stuff that requires you to actually think versus do, uh, are collaborative. And I enjoy this tremendously.
Like I use AI for these kinds of things every single day and I'm literally enjoying it. It's fantastic. I'm now helping, , write a book about AI and writing that chapter has been an amazing collaborative approach. Because I have. Dozens or hundreds of hours of recordings of myself because I have the leveraging I podcast and I have recordings of me speaking on stages and I have recordings of me doing workshops.
So I have [00:24:00] access to all that data that I've already said in a very structured way that I thought about in order to how to present it. And just dumping that into AI and finding gold nuggets and then arranging them in a logical way that will tell a story that will be easy for people to follow while I can interject and say, ah, no, no, no, stop for a second.
And then I want to add this segment because I spoke about it there. Here's a transcription. Find where I talk about this and break this into bullet points. Let's put it under this section. It's phenomenal. It's such an amazing feeling when you have the ability to look at dozens of hours of recordings, And in one sentence, have a bullet point summary of a topic, you know, you talked about and thought about and have information about.
And the same thing in every aspect of business, right? It doesn't have to be, you have a podcast, you've done, let's take marketing, you've done X number of marketing campaigns. You can throw them all in there, show the results that happen and say, okay, which one worked better? Okay. That's what I want to mimic.
So I wanted to use the same framework and the same style and the [00:25:00] same format that we used in the one that worked. in this new thing that we're doing, but we're going to do it about this. Here's some information about it. , can you merge the two together, take the concepts and the thing that was successful before merge it with the new data that I'm giving you, the new campaign that we want to run, and it will do it in seconds.
In a very cohesive and easy to follow way. So that's collaborative environment is extremely powerful, especially that these tools still are not great when it comes to deep learning. expertise in specific topics. So if you know your stuff, and that could be in anything, right? It doesn't matter which field you're in.
It doesn't matter which industry where you know your stuff, you will be in your very particular niche better than most AIs today. Now you can give it information to make it as smart as you, and you can use that information to help it be very aligned with what you're trying to do, but you're still no better.
And so that back and forth, that collaborative effort is still very much required. Now, AI will [00:26:00] be smarter than 99 percent of humans within three to seven years, depending on who you believe at that point, can most people be more innovative, more, out of the box thinkers, more?
The answer is no. What happens then? I don't think anybody knows. It's a very interesting question that I think not enough people that matter, meaning people Governments, large companies, , big groups that can influence society. I don't think investing enough in figuring that out, but it's coming very, very quickly.
Pristine: So this might be another one of those questions that's kind of like too early to answer. , but we talked a lot about how artificial intelligence is like going to collaborate with humans, but how do you think it's going to collaborate with like other new technology that's coming out? Like for example, I've heard a lot about quantum computers.
Isar: AI right now is a big buzzword, but I think it's going to become an infrastructure for everything. [00:27:00] Just like electricity was, just like the internet was, just like mobile phones were, it's going to become the infrastructure for everything. Every new technology will have AI combined with it. Now, specifically quantum computing is actually very interesting. And the reason it's very interesting is because AI at large scale requires a lot of compute. And as you probably know, it's a very big problem right now. I think what we're going to see is that everything will have AI in it, or it will not be relevant anymore when it comes to providing the value that everything's providing. It's just like saying, Oh, I have a, uh, a calculator and it's good to help me with math.
It is. But would you rather have the calculator on your computer connected to the internet where you can give it more complex questions and it can actually figure out how to quote unquote use the calculator and help you with that step of the process as well. The answer is yes. Now, will we still use calculators?
In some places, but I think in [00:28:00] most cases, , AI will just be infused into everything and specifically from quantum computing, it has the opportunity to solve some of the compute and power needs that this new universe of AI requires.
Casey: reduce the environmental impact. That seems to be a big buzz concern with AI and with the large data needs.
Isar: So, I hope so. I think the, what we're going to see in the next few years is a lot of negative impacts to the environment when it comes to AI needs. I hope that in the long run, the promise that AI represents, and if you're asking people that are way smarter than me, that are driving this ship They're claiming that this will solve all the big problems for us, right, including global warming, including clean energy, including et cetera, et cetera, et cetera. And [00:29:00] so if in two to three years, we'll figure out fusion because AI can help us solve some of the issues that we can't solve right now.
Well, guess what? There is no energy problem in the world today, period, done. No environmental impact, no coal, no gas, no nothing. Pure clean energy that's created, through fusion. Now will it actually happen? I don't know. Can we develop better solar panels that are a hundred times more effective than the ones we have right now, potentially.
And so, yes, there are potential benefits that will. hopefully overcome the negative impacts that we are going to experience in the next two to three years. , I definitely see in the near future, especially with the new administration coming in, , seeing a very strong push to innovate, innovate, run forward, let the U.
S. be the number one global leader in AI and [00:30:00] everything else would be a means to that end, including anything you can imagine when it comes to adding, you know, , the wrong gases to the atmosphere and , contributing to global warming and other environmental issues of, I think I heard that, that a hundred words coming out of ChachiPT is one water bottle for cooling. 100 words. Nobody thinks about it when they're going to ChachiPT and say, Oh, you know, write this seven essays for me and now redo them. And then I wanted to summarize it all and blah, blah, blah, blah, blah. . And so there are many big negative impacts of this, which again, the promise from people who are way smarter than me are saying, we will get 10 folds in positive returns. Time will tell which one is right or wrong.
Casey: It kind of sounds like they're just saying, Oh, well, you know, we'll fix it on the backend. , the benefits we can come back here and now that we have this amazing AI, we'll solve all of these problems that we don't even know scientifically. If, if those solutions are [00:31:00] valid.
Isar: So to, to be fair, One of the people that I trust more than some of the other one is Demis Hassabis. Demis Hassabis is the CEO of DeepMind, which is the AI arm of Google. But he founded DeepMind way before Google acquired them, and he has committed his life to helping the human race and our planet with AI.
So when he's saying stuff like that, I have a much higher trust that that's what he's trying to do versus when Sam Altman is saying that. No offense to Sam Altman. But, you know, he just won a Nobel Prize for AlphaFold3, which allows us to do stuff with proteins that were not possible before, either cutting them or connecting the ways in ways that were literally impossible.
And they're creating new protein based materials every single day. That did not exist before that can do stuff that was considered magical before from solving diseases, cancer, [00:32:00] curing different things,
And so there's many when he's saying stuff like that, I have a much higher trust that there's people in this game that that's what they're trying to do.
Casey: that, uh, optimistic vein, I mean, when I was in high school they started, , talking about quantum computing and it seemed, , I, I didn't think it was really possible. It seemed like a far out kind of theoretical thing and, , it seems like the state of quantum computing back then is pretty similar to the state of fusion or some of these other high minded technologies today.
I mean, they say they've done it once and they've generated some power from it in an isolated incident. I think I was reading something pretty similar about quantum computing, you know, a decade ago. So, hopeful, , promising, and, , kind of scary though because, I mean, you're mentioning the physical infrastructure, I think about the social infrastructure.
People don't understand AI, and, , you're very familiar that businesses don't. , but we don't have, , necessarily the environmental infrastructure to support it either yet. And just in terms of water and, , other resources that. If it [00:33:00] does become a global infrastructure, AI will need and continue to train.
Isar: I think you touched on a big point that I think is important, and I don't want to, I don't want to end on a, on a sad note, but we don't have any infrastructure, not just physical infrastructure. We don't have the social means to deal with this. We don't have the economical frameworks to deal with this. We don't have the educational frameworks to deal with this.
Like on every aspect of society, this will be profound and nobody has answers, but we And everybody's hoping we will figure it out. The question is what's going to be the ramifications of the timeframe it's going to take us to figure this out.
And it comes with big fears and great promises, right? And what I think doing stuff like we're doing right now, meaning educating people on the good, the bad, the ugly, what's happening, what's not happening, what tools exist, where is this going, what might be the [00:34:00] impact is a good step in the right direction where people jointly will be able to think about it, will be able to make educated decision for themselves and for society as a whole.
Pristine: in your opinion, do you think AI will have more positive or negative impacts overall?
Isar: It depends on what time of the day you ask me and what day of the week you're asking. So I'm, I'm riding this roller coaster. Like there's days that I lose sleep literally, , because I have three kids, , my daughter is about to go to college and I don't have a clue what recommendations to give her because she will graduate.
I don't know if whatever profession she's going to study today will exist at all. And if it will exist, which most likely will one way or another, how it's going to look like in five years, like what kind of skills, but I can guarantee you that whatever they're going to teach you in year one and two at university, no offense to anybody listening in this, , at Rollins, , is not going to be what it's going to be in five years.
And so when I say we don't have the frameworks, we [00:35:00] don't have the frameworks. And in this particular point in time, in general, if you think about the way our education system works is more or less the same as it worked a hundred years ago. Yes, we now do this online versus with a chalk on a board, but it's still the same thing.
There's a professor, there's, or a teacher, there's students, there's lesson plans. It's one too many like this. This is how education has been taught for decades.
Casey: , Christine and I brought up the point with Dr. Myers last episode that, , The modern education system stemmed from the second or the first arguably industrial revolution And people have been calling AI the fourth industrial revolution the internet was the third So I can see needing an entirely new educational system or infrastructure to support it
Isar: So to end on a positive note, it's actually now we have the tools to create personalized learning that will adapt to the level, the speed, [00:36:00] the needs, the way people learn better. Some people learn better listening to lectures. Some people are visual learners and they need graphs and charts. Some people like playing games.
Some people like reading. Some people need some kind of a mix between these two, depending on where they are in their learning process. building an educational system that can adapt to the needs of every single individual while still bringing those people together into a virtual or physical classroom where they can collaborate with other people, where they can collaborate with the tools, where they have a professor that in now, instead of just giving them the knowledge, he is a facilitator of innovation, of conversation, of discussion, being more of an educator than a teacher is huge.
It's an incredible opportunity that we never had before because the ROI just didn't exist. Like even if you found the best teachers in the world, you can't give the best teacher in the world to every student to adapt to their needs. And now we [00:37:00] can do this. And so, yes, we will need to see a very dramatic shift in order to A, Help people move faster, but B, help people adapt faster because this thing is going to keep on changing every few months and we don't have any infrastructure, whether in business or in academia or in our personal lives to deal with that rate of change and using the tools to help us deal with what the tools are creating is, is the way forward.
Insert Music
---
Casey: Thank you so much for joining us, Isar. I really appreciate you taking the time out of your day to speak with us. Could you tell us a little bit more about your bi weekly podcast and where people can find you?
Isar: Sure. So, my podcast is called Leveraging AI. , if you look podcasting platform, you will, you will find me. The two episodes a week are, one is a how to. So, I find the best practitioners of [00:38:00] AI that I can find, and I host them, and they share everything they do and how they do it. So how to create automations, how to build GPTs, how to use it in HR, in finance, in marketing, in sales, how to use it for self promote, like everything you can imagine, just the best people from different aspects of the business, what AI tools they're using and exactly how they're using it, including prompts and everything.
And that happens every Tuesday and every Monday. Saturday comes out a news episode where I talk about what happened in AI this week, so it keeps everybody kind of like updated on what's happening. And we record most of these episodes live on Thursdays at noon. So if you go to my website to multiply, , you can find how to sign up for those live sessions as well as every Friday, I do an open session that is kind of like a office hours with me.
And every Friday at 1 p. m. Eastern, there is, we call it Friday AI Hangouts, and it's just a bunch of business [00:39:00] people. There's usually 10 to 15 people and we talk about AI and what people find out and what issues they have with their business. And we talk about how to solve it and stuff like that. So there's multiple of these free resources that people can use to learn more about AI if you're interested.
Casey: Awesome. Thank you again for being with us and a really great conversation. I appreciate your time.
Isar: Thank you. This
Casey: was awesome.
Pristine: Thank you for joining us on AI on campus and beyond. , this was our interview with Isar Mateus, CEO of Multiply AI. , be sure to check out the show notes for more information. , and we'll catch you next time. This has been Pristine.
Casey: And this has been Casey. Join us again.