Michael Todasco is a Visiting Fellow at the James Silberrad Brown Center for Artificial Intelligence at San Diego State University. He previously served as Senior Director of Innovation at PayPal and holds over 100 US utility patents. Michael earned his MBA from UC Berkeley's Haas School of Business and a Master of Fine Arts from Johns Hopkins University. His work focuses on the intersection of artificial intelligence, payments, identity, biometrics, and blockchain technologies.
Here’s a glimpse of what you’ll learn:
[1:53] Michael Todasco shares his background and career journey from PayPal to AI research
[7:51] Why Mike is passionate about educating people on AI and its impact on daily life
[9:13] The game-changing potential of ChatGPT and how it empowers non-coders
[11:54] Why business leaders must personally engage with AI to stay ahead
[18:07] AI-driven personal productivity hacks, from meal planning to vacation prep
[22:23] Mike talks about creating a custom AI sales consultant using ChatGPT and real-world data
[24:21] How AI can transform education through personalized learning experiences
[31:19] The importance of understanding AI limitations and avoiding over-reliance
In this episode…
AI is transforming how businesses operate, but many leaders have yet to embrace its full potential. While companies acknowledge the need to adopt AI, many executives delegate exploration to their teams rather than using the technology themselves. How can leaders integrate AI into their workflows and empower their teams to do the same?
According to AI expert Michael Todasco, business leaders must actively engage with AI to understand its capabilities and limitations. He recommends that executives personally experiment with tools like ChatGPT, provide their teams with access, and encourage knowledge sharing through dedicated discussions. By incorporating AI into everyday workflows such as sales strategy, performance reviews, and personal productivity, leaders can drive innovation while fostering a culture of curiosity and collaboration.
In this episode of The Customer Wins, Richard Walker interviews Michael Todasco, Visiting Fellow at San Diego State University, about AI’s impact on business and leadership. Michael shares how AI can optimize workflows, improve decision-making, and revolutionize education through personalized learning. He also dispels common misconceptions about AI, offering practical advice for leaders looking to stay ahead in an AI-driven world.
Resources Mentioned in this episode
"[AI Series] Streamlining Operations With Cutting-Edge AI Technology With Babu Sivadasan" on The Customer Wins
"[AI Series] Tailoring Client Acquisition in Finance With AI With Farbod Nowzad" on The Customer Wins
"[AI Series] Harnessing AI for Strategic Leadership With Geoff Woods" on The Customer Wins
Quotable Moments:
“AI isn’t just for coders — anyone can use it to boost productivity, creativity, and problem-solving.”
“Leaders must engage with AI personally; it’s the only way to understand its power and limitations.”
“The best way to learn AI is by experimenting; start small, make mistakes, and discover its potential.”
“Education will be transformed by AI, enabling personalized learning that meets every student’s unique needs.”
“AI isn’t perfect, but neither are humans; understanding its strengths and weaknesses is key to success.”
Action Steps:
Experiment with AI in daily tasks: Start using AI tools like ChatGPT for problem-solving, writing, and brainstorming to understand their capabilities.
Encourage team-wide AI adoption: Provide employees with AI tool access and create a dedicated space for sharing insights and successful use cases.
Make AI part of leadership discussions: Regularly discuss AI applications in meetings to ensure leaders and teams are actively integrating it into workflows.
Personalize learning with AI: Use AI-driven platforms to tailor education and training based on individual employee strengths, weaknesses, and learning styles.
Stay updated on AI advancements: Follow industry trends, experiment with new tools, and continuously refine AI strategies to maintain a competitive edge.
Sponsor for this episode...
This is brought to you by Quik!
At Quik!, we provide forms automation and management solutions for companies seeking to maximize their potential productivity.
Using our FormXtract API, you can submit your completed forms and get clean, context-rich data that is 99.9% accurate.
Our vision is to become the leading forms automation company by making paperwork the easiest part of every transaction.
Meanwhile, our mission is to help the top firms in the financial industry raise their bottom line by streamlining the customer experience with automated, convenient solutions.
Go to www.quickforms.com to learn more, or contact us with questions at support@quikforms.com.
Episode Transcript:
Intro: 00:02
Welcome to The Customer Wins podcast, where business leaders discuss their secrets and techniques for helping their customers succeed and in turn, grow their business.
Richard Walker: 00:16
Hi, I'm Rich Walker, the host of The Customer Wins, where I talk to business leaders about how they help their customers win and how they're focused on customer experience leads to growth. Today is a special episode in my series on artificial intelligence, and today's guest is Mike Todasco, a Visiting Fellow at San Diego State University. Some of our past guests in this series have included Babu Sivadasan of JIFFY.ai, Farbod Nowzad of Cashmere AI, and Geoff Woods of AI Leadership. And today's episode is brought to you by Quik!, the leader in enterprise forms processing. When your business relies upon processing forms, don't waste your team's valuable time manually reviewing the forms.
Instead, get Quik! using our Form Xtract API. Simply submit your completed forms and get back clean, context-rich data that reduces manual reviews to only one out of 1000 submissions. Visit. Quickforms.com to get started.
All right, I'm excited for today's guest. Mike Todasco is a Visiting Fellow at the James Brown Center for Artificial Intelligence at San Diego State University and former head of innovation at PayPal. Mike, welcome to The Customer Wins.
Michael Todasco: 01:30
Rich, thanks for having me. Or should I refer to you as bruh bruh?
Richard Walker: 01:34
Yeah, let's do that. All right. So for those of you who haven't heard this podcast before, I talk with business leaders about what they're doing to help their customers win, how they build and deliver a great customer experience and challenges to growing their own company. Mike, let's understand what you do a little bit better. How do you help people?
Michael Todasco: 01:53
Yeah. No thanks Rich. Yeah. And for folks like we were talking beforehand about how our kids often referred to each of us as bruh, so that's why. How do you help people look?
Honestly, part of it is I do podcast to talk to people about probably the most impactful technology we will ever see in our lifetime. I love doing this. I'm not here selling anything. I want to get the word out because I want people to embrace and understand and get into the conversation about and be informed when they get into the conversation about like, where is AI going to take us? Because it is already having meaningful impacts in how I work, how I write, how I interact, how I plan family vacations, how I cook, how I drive my car, how I do all of these things.
And it's just getting started. And so it's an amazing technology that like, I try to just have nice conversations where people so people aren't afraid of it, and people are just willing to just try it and experiment and learn about it for themselves.
Richard Walker: 03:08
You know, I love that mission, Mike, because, look, I'm 50 years old at this point, and I saw the mosaic browser in college and said, what the heck is this thing? I missed the boat. I really didn't understand it. And then blockchain came out and you're like, oh, blockchain is gonna be everything. I think it's important, but not like this.
AI is a fundamental change in how we do everything that's coming, coming towards us. So what brought you into this world of AI? Is this something you studied or are you a physicist, a PhD in this stuff? What's your background?
Michael Todasco: 03:39
I'm an accountant. So that's how I. This is true, actually. No, Rich, my background is very similar. Look, we're about the same age.
I started college in 95. I never used the internet before. I got to college, and I went to school at University of Illinois, which is where they were just happened to be building the mosaic browser and so forth. So I get to school and I'm like, oh my God, what is this thing? Like there's these blue links you're clicking on and it takes you to these magical places and all this.
I have never seen that in my life and I was just completely blown away by it. I'm like, My God. And then I come back and talk to my friends who aren't going to Illinois, like over Thanksgiving and, you know, Christmas break and they're not experiencing this. I'm like, oh, crap, I'm living in the future like I and it's just by, by chance that I happen to be going to school where Andreessen and Team built this amazing browser that changed all this stuff. So I was very lucky.
So look, I started I've always been into technology. I've always I one of the back in 2003, we had we were living in Ohio and we actually had a robotic lawnmower that we had on our 2003, 2003.
Richard Walker: 04:56
Wow. I didn't know that happened.
Michael Todasco: 04:57
And I know this. I know it was a company called Friendly Robotics. It was out of Israel. I'm sure they don't exist anymore today. It didn't necessarily do a good job.
So I was always trying to do work to try and figure this out and all that. But it was, I will say, if you're ever trying to sell a house, we sold our house in 2006. I think that's when we moved out to California. We actually included the robotic lawn mower in the house. It was the best line item you could possibly imagine when you're actually looking at like, well, which should I buy?
Wait, this one has a robot lawnmower. What is this thing? So yeah, but I've always been into tech and I've just been fascinated by it. Fast forward to when I'm at PayPal. Machine learning was a huge thing.
Just part of like, what teams at PayPal did because like, if we didn't have world-class machine learning teams, PayPal wouldn't survive as a company because of the.
Richard Walker: 05:57
Security and fraud detection, right?
Michael Todasco: 06:00
Fraud. Exactly, exactly. And that's where. And that's where it was. World-class within PayPal.
Not as much on like the consumer side, like targeting you with ads or something or emails like not as much on the fraud side. Huge. And so I got to know all these machine learning engineers and they would build these amazing things. And I was eventually running innovation at PayPal. We could talk about that whole path.
But you know, so but I was an accountant by trade, just somebody who loved tech and was interested in it, but I never was able to build stuff. And then fast forward to 2021, which was the first time I really got my hands on GPT three. And I found I could do all these amazing things in GPT three, and I don't actually need to code. I don't need to know how to code. I don't know how to code.
And I'm like doing this in like, just plain English and seeing these incredible outputs. And that was so eye-opening to me of like, oh my God, this is a paradigm shift. Whereas if I think if I was actually an engineer, it wouldn't have been as clear to me how different, how amazingly special this was.
Richard Walker: 07:04
Yeah. Yeah, it's eye-opening. I remember getting it. What was that? December 1st of 22.
And just blew my mind with what it was capable of doing. And here we are a few years later. So I admitted to somebody my energy vampires. ChatGPT because I can't put enough time into it. There are so many things I want to do with it.
Michael Todasco: 07:24
Your energy vampire.
Richard Walker: 07:26
It's just sucking the life out of me.
Michael Todasco: 07:27
It's all going into it sucks. It sucks your energy from you.
Richard Walker: 07:30
Is that not a positive way? Like, I just can't get enough. I want to do more. There's so many things I want to accomplish with it, because it's this great enabler of speed and thought and automation, and I'm still putting all the pieces together. What is it you do, you teach?
Do you have a course? You're teaching at San Diego State?
Michael Todasco: 07:51
On and off. I do a lot of, like, guest lecturing at the school. I am very much focus more on the application side, which is great for podcasts like this. So. So it's not as much about the architecture, how these things are built.
It's more like what you can do with these models. So that's most of what I write about for the school. And I will usually jump into other AI classes and things like that and kind of, you know, show the application layer side of it. We don't have a full application class. I don't know if that's really necessary because and one of the things I always tell people is like, you don't need this go to school to learn how to use these tools.
It's the you know, you don't even need to take a class online. You don't need to say like, oh, is there whatever. Like literally just play with them. Just play with these tools. It's the best way to do it.
Like find a real problem that you have in your life and see how this tool might be able to just help you think through that. Again, these are just tools and they are super simple for you to use. You just got to be comfortable with finding out what it's going to take to actually make it work.
Richard Walker: 08:55
Yeah. Okay, so you want to help get the word out there that people should have awareness of this. So what is it you want them to do? Do you want them to play with it, or do you want them to observe it and understand it or learn how to? I mean, because lots of software providers are building AI into their products, so many of us are going to experience AI whether we know it or not.
Michael Todasco: 09:13
Yeah, 100%. Look, we all have been for long before, you know, everybody even knew about ChatGPT or was worried about AI or anything like that. I mean, you know, if you're using Google Maps, if you are using a social media feed that has an algorithmically generated timeline that it's showing you there is AI running behind the scenes to do that. And it just wasn't up in front, you know, this was you know, if you're making a payment on PayPal or anything like that, like there's tons of algorithmic machine learning and AI models that we had on the back end that was doing this. But the difference now is it's in the front end and you can actually manipulate these world-class models today.
So yeah, my recommendation for people is honestly just make it so that it is your default. If you tried ChatGPT back in, you know, November 2022, right when it launched at the end of the month and all that, like you can and you're like, yeah, it didn't do this. Well, like, do you know what? It's totally different today. Like it was two years ago, for goodness sake.
Like, that was like that was an eternity in AI life. You just have to play with these models and find out what it is good at and what it isn't good at. If models stopped improving today, we would continue to see the benefits for a long, long time from these models because I'm finding out stuff constantly like, oh my gosh, I didn't know it could do that. And like, you know, like there's always just finding these little things, these little ways to help and so forth. And so what I would say to people is just make it your default start, you know, so if you have a Google as your search.
Go into your browser, switch it to ChatGPT, search or perplexity or whatever it might.
Richard Walker: 11:01
Be, perplexity, and.
Michael Todasco: 11:02
Start using these as a default and you'll start to realize what they're capable of and also what they're not capable of.
Richard Walker: 11:08
Yeah. So one of the things let's turn this towards leadership a little bit, because one of the things I've noticed I went to recently, I went to an AI roundtable, and it was leaders from my industry. And I discovered I went in there thinking, oh my gosh, I'm going to be behind everybody else. And what I discovered is I'm not actually behind everybody else. Most of the people in the room really are not touching it themselves.
They're oh yeah, my teams are looking at it. It's definitely something we want to do this year. ET cetera. So I think a lot of people are coming into this with we know we have to do it. There's a fear of missing out because we just talked about that as we started this podcast.
But there's also hands-on. So what should leaders of organizations do? What should be their first thing they try to do with these large language models or AI?
Michael Todasco: 11:54
Yeah. So if you're a leader of an org, first thing to do is use it yourself. Like, honestly, just find out ways to use it yourself and set an example for the rest of your organization. Because this is going to be the way that business is going to be done. Like if you're doing things on a computer, like you are going to be interacting with these tools going forward, it might be more integrated, it's going to be smarter or whatever.
But like do it today. Think, you know, this is the internet in 1998 or something like that. Like, you know, it's there, you know, it's on the horizon. But you kind of don't know what to do with it sometimes. And that's okay.
So, you know, thing one is just use it yourself. By everybody on your team a subscription. So do it in the right way. You know Claude. Google, OpenAI. They all have corporate subscriptions to these things and spend the 20 bucks a month per person on your team. And you could cancel it after a few months if you find out it's totally useless. But useless, but give it a fair shot and just spend for that corporate license. So everybody has access and everyone has access to do it in the right way because you don't want, you know, somebody putting a bunch of Social Security numbers into, like, their own personal ChatGPT account to do that.
Like they shouldn't be doing that anyway, but like, you know, make it as easy for your employees as possible. So use it, get that license, and then open up communications. And two ways to do that. So one is if you use Slack, if you use Teams or whatever, create a specific channel just for people to share. AI prompts AI and things that they find and discover with AI.
Just like have people share this because again, it's possible you are doing something in AI and you're finding something out that nobody in the world has ever discovered before. Rich, as you said, like we are so early in this game, like, like and this is one of the few times that people are building products, not even knowing what they're completely capable of. Yeah, they're constantly finding this out. So share that within your organization because that's going to spark and spur on other ideas within the group. And then finally, and this is where it gets a little bit more controversial.
But if you've got a staff here's what I do. And I did something very similar with the team I manage once. If you got a weekly meeting, you know what, start first week and say, hey, do you know what? Every week I'm going to cold call on one person at the beginning of our staff meeting. I'm going to ask them, what did you use?
What did you do with AI that was interesting over the past week? Just one, you know, simple answer. And it could be personal. It could be professional. It actually doesn't matter.
Like it could be either way. And you set the example yourself the first week and say like, hey, I used it to plan out what we were going to do with my family when they came over the holidays or something this year Something like that. And like. Great. And show them a little bit.
Just show the prompt and that's it. Then next week you actually cold call on someone to do that again and again. Again you do this for a few weeks. Look, your team might hate you a little bit at first, but here's the thing. They're going to start thinking about it.
And then within a couple of weeks, they're actually going to start to see like, oh, like wait, they're going to be constantly in the back of their mind thinking about how to use these tools. And then they're going to say like, oh, well, maybe I could use it for that. They're going to be looking for ways to use it during their week, because they don't want to be like called out by the boss and not have anything. So it's a little bit more of a stiff-arm way to do that, but a little bit of pressure on your team. But within a few weeks I think they'll move past that.
And I think they will all discover and like you'll actually by then have people raising their hands in your meeting and saying like, hey, actually, I did this really cool thing, even if you're not calling me, I really want to share it with everybody, and that's what you.
Richard Walker: 15:49
Want to get to. That is great advice because we have a weekly meeting and I talk about AI all the time, and periodically somebody will volunteer something cool they did. But I have not decided to make it a facet of the meeting every week and call on people. I'm going to start doing that. Mike.
Michael Todasco: 16:03
So, I did this. It wasn't about AI, so I was managing a team and it was a very it was a mixed team. It was called product launch and it sort of program management. But product management put together it was a weird hybrid and I was very keen on creativity. I was eventually running innovation at PayPal, so hence that was just how my brain worked.
So I actually did the same things, basically saying, hey, I want you to talk about something creative you've done and you did in the past week. And I went through this whole thing with my team. And yeah, they hated me at first. But like, I still remember the stories and I know they remember the stories about like one was talking. One person talked about how he, like, snuck through the airport, that is, by the papal office to get to work or something.
I'm like, what? How did you like, like you snuck through the airport to get here. Like he's like, yeah, it was way faster. This is how I did it. Like. And like it was just all of these things. And it led to so much great conversation, honestly. And yeah, they'll hate you at first, but like, within a month, they'll be loving these tools.
Richard Walker: 17:07
So I want to give some leaders who listen to this call the incentive to move forward with a ChatGPT or cloud, whatever. And you had actually given an example of this already. But do something personal. That's simple. Create a recipe like I have these four items in my fridge.
What can I make or plan a trip? We planned a trip last summer to go to Utah and from Austin, Texas to Salt Lake City. Utah is a very long way. So I put in parameters and I said, help me plan my trip. I don't want to drive any more than six hours in one day. I want three stops along the way. I have little kids. They need things to do at these cities. Where should I go? What should I do?
Do you know how much time that would take to do manually? Yeah. Oh, and ChatGPT is like, here's your trip. And I'm like, okay, route it backwards now, but take a different path through different states. Okay, here's your trip.
I was amazing, but I want people to try personal things like you've done because you get that feeling for yourself. You see, oh my gosh, this is powerful. And that sparks more ideas of what else you could do.
Michael Todasco: 18:07
Absolutely. I mean, let me give you one that I had done, which was really one of the more eye-opening ones for me. So I make I work from home, so I make most of the meals and stuff like that. I got two smaller kids and look, between the four of us and our family, everybody's got different dietary restrictions. It's ridiculous.
I basically don't want to eat carbs. My wife won't eat this, my kids won't eat that. But like the rule is in our house always, we're cooking one meal and like you're going to do whatever. I will try and accommodate as best as I can. But you know, we are not making multiple meals in the Todasco household.
So what I did was I created a basically, actually, it wasn't even before. It was just like one thread within ChatGPT. I'm like, hey, here are the requirements for the four people. Here's what we want. I want you to make this easy.
So I want it to be less than 45 minutes. I want you to, you know, I put a bunch of other requirements, like one pan maximum. So I don't want a lot of cleanup, all this other kind of stuff. And I want you to meal plan for me and then give me a shopping list of what I need to actually buy for the week. And it would do it.
And then this is called reinforcement learning with human feedback, or a version of that, I would then have the family rate, the what they ate after every meal, like five-point scale, like, hey, my daughter thought this was the three. My son thought this was a five. And then it actually starts to get better with that, so it'll tweak the other ones. But I would do this and like it got better and better and better and doing this and so like and frankly all this is in OpenAI now and in ChatGPT. So it knows my family pretty darn well with this.
But one of the best parts about it, especially if you are horrible at following directions like me, If you screw things up, which inevitably I'm like, wait a minute, I have a box of corn starch. I'm almost at the end. I didn't add it. I screwed something up here. I will often either take a picture or say like, oh geez, I did that.
And you know what it will do in the recipe? It'll say, oh yeah, well, you could do this instead, or you should do this. I've also like taken a picture. I'm like, hey, are these green beans done? And just put it in there and it'll actually respond to you with that.
Wow. I mean, that's the beauty of these tools that you can take visual inputs. They can adjust it. If and if you screw things up, it'll say, okay, here's how we're going to fix it. You can't do that if you're just doing a Google search. So like that's one way I love using these tools.
Richard Walker: 20:34
Yeah I love that. I'm going to share another example with you from a business standpoint. It's one of the more powerful ways I've used. I wanted my own sales consultant. So I asked ChatGPT, how could you become my sales consultant, given that you know, all this information, my type of type of company, my industry, etc. it said, here's 21 questions to answer.
So I recorded myself answering all these questions. Honestly, it took around four hours of talking to answer these questions. Took the transcript, fed it back in and asked me two more questions. Fed that back in. And then it said, let me write a prompt for you to create your own custom GPT so you can have your own template, if you will, of your prompt.
And then I fed it into that custom GPT, all that transcript. Again, everything we had talked about in the prior chat. I gave it, sorry, two interviews from podcast where I talked about sales, our philosophies, etc. and I gave it like 30 recordings that customers allowed me to record in sales demos. So now this custom GPT knows everything about my industry, my company, where my origins are, my competitors, all that kind of thing. It is so powerful.
Mike, the other day I had a customer say, hey, in the in the 11th hour, we're looking at one of your competitors. Tell me, give me a comparison of you versus them. I'd never heard of this company before? So suddenly I've got to figure out. How do I compare these two?
Hey! Perplexity. Tell me everything about this company. Now compare it to my company. I did the same thing with ChatGPT to see which one's better. I fed that information into my custom sales consultant and said, now write up a write-up from my perspective, using my voice because it heard me talk for four hours. Yeah, in a way that I would respond to this and position us really well. But be truthful, right? Show good and bad, but position us well to solve this problem. It spit out the perfect email response. I really didn't have to change a word because it speaks like me.
Michael Todasco: 22:23
I love that. Love that. Wow. I mean, that's what. So the voice mode in ChatGPT is highly, highly underrated. So I have actually used. So I was working on a children's book, and one of the things that I did was I actually I talked, I had voice mode, so I just put the phone in my pocket, I put in ChatGPT voice mode, I had a earbud in and I just went for like an hour walk and said, like, hey, I'm working on a book.
I'm going to just start talking. I need you to transcribe all this. Occasionally I'm going to ask you questions because it's a science-based book and I don't know the science on this. Well, and I and then at the end, summarize this, transcribe this like it's already transcribing everything and doing an amazing job of it. So like, you know, if I was just driving into work in the morning, I would be like banging out emails and to-do lists and all this other kind of stuff.
And then like when you get in there, you can just start sending stuff out or whatever, like there's so much you can do with that. We use it to, to quiz our kids in the car sometimes. So like, you know, if they got a test on US history coming up, we could say, hey, ChatGPT and just put it on the car speaker system. This, you know, through Bluetooth. Let's ask questions.
Ask us questions about US history at a fifth-grade level and get increasingly more complicated as they answer them correctly or something like that. And it'll do that. It's a brilliant idea to do with voice mode. And I love your example, Rich.
Richard Walker: 23:54
I love your example. I'm learning here, man. Like, we could talk about examples all day long, but there's something you and I talked about offline that I want to bring into this conversation. You said you had a really strong focus on education and how it's going to change and morph to be customized to the student, to their specific needs. And I want you to talk a little bit about how you see the future enrolling, because I think overall we're going to see mass customization of all sorts of things.
So what is your view on this? What do you think is going to happen.
Michael Todasco: 24:21
Yeah. Yeah I'm a child of the 90s. So just as you were talking there, I'm like, yeah, what do you call this thing? And I'm like, oh, extreme customization is really what this is with like a big X like that or like that was very 90s way of branding that. But look, I think you're going to get to the point where I hope three, four years ago this could happen.
Today it's a little bit of work, but it can really happen 3 or 4 years from now. If I were a math teacher for like third fourth graders, what I would probably do is at the beginning of the year, I would ask all the students in the class, hey, I want you to fill out this piece of paper. And on this piece of paper, give me your dog's name. Give me what TV shows you like. What are your favorite video games like?
Who are some of your favorite friends like? Are there family members you know? And you know, I probably have them like give a nickname or something. So it's anonymous. So like I can't tie it back at the end.
So like, you know, you're a jazz or whatever, like just some crazy nickname that you can't tie back. I would take all these things I would feed into a large language model, and then anytime I'm doing a math assignment, I would basically give each kid custom questions, you know, and maybe it's a 75, 25 split. So maybe like 25 I'm going to give everybody the same one. They're going to just those will be the ones we discuss in class, but the other ones are going to be about things you enjoy. Because here's the reality, like, Especially with kids.
And this is true with adults too. You got to meet them where they're at. Talk to them in the language that they understand. I learned math, and I learned to love math because as a kid, I loved baseball. And on the back of all these baseball cards that I collected were all these numbers.
I didn't know what the hell these numbers were. But, like, to me, it was like almost figuring out a puzzle. Like, what was this? And like, oh, batting average is this slugging percentage? What is that like?
You know, and just like figuring all these things out and like, just being able to do basic math from that and understanding these concepts and why it's important, if I didn't have that, I may not have got the same love of math I got otherwise. And like this is true for everyone. There is something that you are passionate about, and to be able to talk to 30 different kids in effectively, in their own language and the way that's going to emphasize their passions. You know, whenever I'm making problems for my daughter. I don't make them generic.
I'm like, hey, you and your best friends are going out and doing this, you know? But for a teacher in, you know, pre-2020, that would have been impossible to do that for 30 kids in a class. Yeah. With I that is totally that is not only possible. Like you know at some point it's going to almost be criminal not to do that for your kids to just think like, oh, I'm just going to give all the kids the exact same thing for their homework.
No. Because like, do you want these kids all have different interests. And so like that's where I think education is going. And I think like the biggest benefit is going to be elementary. But this could go all the way up.
Richard Walker: 27:25
I love that idea. But I also think that it's going to help challenge students who need more challenge.
Michael Todasco: 27:30
Yes.
Richard Walker: 27:30
And help students who are over-challenged to get the right level of challenge so they can grow effectively and not be left behind. Because, look, my oldest son, he taught himself third-grade level math in his head in first grade.
Michael Todasco: 27:45
That's great.
Richard Walker: 27:46
I had to ask this teacher. I'm like, are you teaching negative numbers? Are you teaching division? He's like, no. Like, how does he know these things?
Yeah. And so he was frustrated and bored and acting out.
Michael Todasco: 27:55
Yes.
Richard Walker: 27:56
So I think that being able to challenge kids at the level they're at and on the strengths they have, but also to augment their weaknesses by helping them learn those things in the way that you just described by making it relevant to them. I think that is so special.
Michael Todasco: 28:09
Yeah. And look, and there are things like Khan Academy does this right. There are software things out there that do this concept. So it isn't entirely new, but it is entirely new to do this in the classroom. And that's the difference now.
And look, I'm so this doesn't actually even mean that the teachers have 30 times more things to grade. Like they can have an answer key for every darn one of these things. You know, it could effectively just be a fine. I mean, there's many ways to do this that is not actually creating more work for the teachers. In many ways, it's creating less work for the teachers, but giving a much more personalized experience for the students.
Richard Walker: 28:47
Yeah, I wonder how. Well, I wonder how well these large language models will actually grade the work for the teachers.
Michael Todasco: 28:53
So for basic for basic math, I think it's going to do actually a really good job. Yeah. I did an experiment a couple months ago where I had all the language models generate a math problem. I had the other language models generate an answer, and then I put it back in the first one, the original one, and had it graded. And I will say like some of them like ChatGPT oh one, which is like more advanced.
It's damn good. Like it was actually having like the other ones do math problems. I couldn't even imagine. The old Gemini, which I think they've updated, it was historic, like comically bad. Like I mean like, so I mean, like, there are gaps in there, but you could see where all of this stuff is going.
And that's why, like, okay, if it's not there today, you know, wait a few months and we're going to get to the point where, you know, I think basic math at least like that, we're going to have no problem with, you know, supposedly you can actually do PhD level math in the latest, the $200 a month version of ChatGPT. I don't know, PhD level math, so I can't actually vouch for whether it is or not. But supposedly you can do like much higher level math than that.
Richard Walker: 30:09
You know, this brings up something else I want to relate. And then we're going to have to wrap this up, which is these large language models fail. They're not perfect. Yeah. And so when you're playing with this, you may not get the result you want.
And you have to accept that this is part of the learning. You'll find out what it's good at, what it's not good at. So as an example, I use ChatGPT oh one as a preview to write a very complex set of equations for an ROI calculator. Now, I don't mean calculus equations, just a lot of inputs and parameters and re-editions etc. and the problem I found was especially going back to 4.0 version of ChatGPT. It cannot actually calculate the math.
It's not a math machine. It writes it out, it understands the premise of it, but it can't keep all the variables in its memory to plug them all in. So it kept dropping things and having inconsistent answers. And then I go back to zero one and I'm like, you run the model. And it did a much, much better job.
Like 99% of the time it was right. Yeah, but not 100%. I can't use it if it's not 100. Excel does it super well, right? Yeah.
So I just want to reiterate, you have to kind of go through this learning curve problem and accept some of these failures to figure out what is the best use case for these models. They're not for everything.
Michael Todasco: 31:19
Absolutely. Look, and in some ways, hallucinations are a feature. If you're writing a fiction book like you want it to kind of get a little crazy, get a little bit out there. You know, the temperature when you're when you adjust these models, it's called temperature. So like like, you know, if you adjust the temperature higher it's going to be crazier.
If you just lower, it's going to be like more predictable in what it's putting out. So, you know, effectively you need to do these with these models or the models at least need to understand when they need to turn it up or down themselves based on the type of question you're being asked. But it is really important to know what the models are good at what the models are not good at, and which models are better for different things. You know, if I were to do, say, performance reviews using a model, I think you could do a pretty good job of having the models aggregate stuff. But, you know, the best tool for that might be something like Google's Notebook LM, because that is called Rag retrieval augmented generation.
So effectively what it's doing is it's prioritizing the information. You're feeding it above and beyond anything else that's in there. It will kind of go outside. But basically this is the it's limiting itself to the training set that you have. And, you know, if I'm taking if I have a big team, if I have all these, you know, all this feedback and I want to aggregate it and all that in there.
You know, you could do a pretty good job in notebook LM. And you don't want hallucinations when that's happened. You don't want to misquote somebody about somebody on your team or anything like that. But there's a lot ultimately you got to one. You got to just try these things out, play with the tools.
That's what we were talking about in the very beginning of this. And then, yeah, also know that they're not perfect. They're not these are not God machines or anything like that that are just going to be always right. All knowing I was somewhere and I was actually talking up at my alma mater, Berkeley, and there was a professor in the audience of the class that I was teaching. I was kind of guest lecturing.
And he said, like, yeah, I don't trust these models. I'm like, well, why? And he said, well, it doesn't know who I'm married to. And I was like, okay, like, say more about that. Like, I'm confused.
It's like, yeah, I put it in there. I put in my name. I'm like, hey, who am I married to? It doesn't know that. And I and I'm like, well, how would it know that?
Like, and I actually meant no offense. It turned out it was very offensive. I'm like, are you like in Wikipedia or something like, do you have an entry like or like? No, but and I'm like, well then how would it know? Like these models are not like these all-knowing, omniscient, reading-your-mind type of thing.
And if you're not a famous person who has a Wikipedia page, it's not going to know who your spouse is for something like that. And those are just the things that we all need to kind of get grounded on, like where it's good and where it's not good.
Richard Walker: 34:28
Yeah, for sure, man. I want to keep talking to you. But I got to wrap this up. And before I get to my last question, what is the best way for people to find you and connect with you if they want to reach out?
Michael Todasco: 34:39
I would say LinkedIn is probably the social network I'm most active on. I'm actually a little bit active on blue Sky. So just Todasco my last name on both of those. Best way. And I have a newsletter on LinkedIn.
It's totally free. It's more of these ramblings that I have about I stuff. I'm not selling anything. If you want to learn about that, you know, you just find me in my newsletter over there on LinkedIn.
Richard Walker: 35:04
And go find his LinkedIn because he's got a really, really cool profile image background header thing, I love it.
Michael Todasco: 35:10
Yeah, that was generated by an AI based off of a poll that my followers answered upon. Nice. Yes.
Richard Walker: 35:17
Nice. All right, so here's my last question. Who has had the biggest impact on your leadership style and how you approach your work in the role that you play?
Michael Todasco: 35:27
I have had some really great bosses and I've had some pretty bad bosses in my life. I think luckily, more, more good than bad. If I were to say who has had the biggest impact, I don't think he would know that I would say him, but it was a mac. Ramada is his name. He's now at Chase.
He was one of my bosses at PayPal. He put a lot of faith in me about like being able to have a team. I was so different and so weird, but he just kind of embraced that. And he was so good with his organization. Like I actually remember.
And one of the reasons I'm probably thinking about him is because he did this over the holidays. Matt had like a 50-person org. He wrote every damn person in his organization a customized, personalized Christmas card every year. And not just like signing his name like front and back specifics. All that for 50 people in his organization and then sent these out, you know, and I'm like, God, if I could be like a fraction of who Matt is, I'd be, you know, lucky to be a leader like that.
So that was a very and is still I don't work with them directly. He's still out there. If anybody wants a job at Chase, go work for Mac Ramada. He's an amazing leader.
Richard Walker: 36:51
Oh, that is awesome. I love hearing that. Thank you for sharing that. Yeah. All right.
Michael Todasco: 36:55
Thanks for asking that.
Richard Walker: 36:56
Yeah. Yeah I one of my favorite parts of the show. All right. I want to give a big thanks to Mike Tedasco, visiting fellow at San Diego State University, for being on this episode of The Customer Wins.
Like I said, go find Mike. And don't forget to check out quick at dot com where we make processing forms easy. I hope you enjoyed this discussion. We'll click the like button, share this with someone and subscribe to our channels for future episodes of The Customer Wins. Mike, thank you so much for joining me today.
Michael Todasco: 37:21
Rich. It's been a pleasure. Thanks for having me.
Outro: 37:25
Thanks for listening to The Customer Wins podcast. We'll see you again next time and be sure to click subscribe to get future episodes.
If you’re looking for an easy way to convert JPG images to SVG, there’s a handy online tool that does the job in seconds. It turns your images into scalable vector files, perfect for design work, crafting, or digital projects. No more losing quality when resizing!
You can try it link. 😊