62 | Strategies for Bridging The Gap Between Personal & Organization AI Adoption with Amanda Bickerstaff

Let's break down barriers between today’s potential and tomorrow’s adoption.

Personal AI adoption is through the roof while at the same time organizations are struggling with developing an implementation strategy.

Understanding the gap between individual and institutional adoption of AI has become crucial. This session is designed not just to enlighten but to offer practical solutions and frameworks for bridging that gap and achieving effective organizational AI adoption.

Our guest, Amanda Bickerstaff, an esteemed figure in AI education, will guide us through this journey. As a pioneer in the field, Amanda has made significant strides in responsible AI adoption within educational systems, advocating for AI literacy and ethical usage. 

Her insights will not only shed light on the challenges but also highlight the immense potential AI holds for transforming organizations and educational institutions alike.

In this webinar, we'll explore the nuances of AI adoption at the grass-roots level and how it contrasts sharply with institutional hesitance. Through Amanda's expert lens, we'll discuss:
↳ importance of change management in embracing AI
↳ developing strategic frameworks for safe and efficient use
↳ innovating within the realms of AI technology to stay ahead in this dynamic landscape.

Whether you're a C-suite executive, a business leader, or an educational pioneer, this webinar will equip you with the knowledge and tools to navigate the AI era confidently.

If you are ready to leap from knowledge to action and learn from a leader who's transforming the landscape, listen to this. 

About Leveraging AI

• The Ultimate AI Course for Business People: https://multiplai.ai/ai-course/
• YouTube Full Episodes: https://www.youtube.com/@Multiplai_AI/
• Connect with Isar Meitis: https://www.linkedin.com/in/isarmeitis/
• Free AI Consultation: https://multiplai.ai/book-a-call/

If you’ve enjoyed or benefited from some of the insights of this episode, leave us a five-star review on your favorite podcast platform, and let us know what you learned, found helpful, or liked most about this show!

Transcript

Isar: 0:00

Welcome everyone to a live episode of Leveraging ai, the podcast that shares practical, ethical ways to grow your business, advance your career. And we are here today to talk about AI implementation. And AI implementation today has a very interesting dissonance in it. On one hand, you have a lot of people who are using it daily across multiple things, whether in their businesses, in education, on their personal lives, but from an organizational perspective, there's a huge gap as far as how to actually. Implement AI in an efficient way through the organization in a way that's gonna be a effective and B safe. I literally had two conversations in the last twenty-four hours with two potential clients that told me, yeah, we got started. We set up this group that are like the AI experts of the company, but they know, really know what to do and they're not making any progress and everybody's really frustrated and this has happened. And they're actually a few steps ahead because they've actually set up a committee to do these kind of things. Our, guest today, Amanda Bickerstaff is an expert on exactly that topic. She is the founder and CEO of AI for education, and she's been helping educational organizations. Understand how this whole AI thing works and how to implement it in a way that everybody will benefit from, right? So the educators as well as the students and everybody in the organization. And the conversation is gonna be not just about education, even though we're gonna focus a lot about that, but on an organization in general, how should an organization approach the implementation of AI in a way that will be. As I said, effective and safe. So I'm really excited. I know a lot of people are struggling with this right now, and so Amanda, I wanna welcome you to Leveraging ai. Welcome to the show.

Amanda: 1:52

Hi everybody. I, yeah, I'm happy to be here today. like I said, I'm Amanda. I'm the co-founder and CEO of AI for Education. and this is our world. we are fully focused on the responsible adoption of general ai specifically, in educational institutions, whether that's from the school level, the district leadership, higher ed, but also working more and more with, Companies, so education companies, tech companies. and so it, it's been really interesting. Just had a good conversation today about thinking through, like how you even think about using generative AI to support content development and, how do you start to create some infrastructure for these tools. So happy to be here and looking forward to conversation.

Isar: 2:32

Awesome. I want to thank, first of all those who have joined us on the chat. So feel free to o on the zoom. feel free to write in the chat who you are, where you're from, what you're looking to get out of this. And for those of you who are hopefully watching us on LinkedIn, live, thank you as well for joining us. We're doing this almost every Thursday, so Thursday noon eastern time. We're doing these live with experts like Amanda talking about. AI practical implementation. So if you wanna join us, just follow me on LinkedIn. There's always like a place to sign up, and if you're in the Zoom, then you can absolutely ask questions and we will try to relate to those as we move forward. So Amanda, let's really dive in and get started. What is the problem, right? It doesn't sound like such a big deal on paper, right? It's a new technology. We had new technologies before we've been through change. Processes through organizations before, like it's not the first time this is happening, and yet literally across the board people are struggling to deal with this and to understand how to move forward in, in a way that makes sense. Can you help us understand why there's such a big of a problem talking to so many organizations and helping them through the process?

Amanda: 3:42

Sure. I think that while artificial intelligence has been around since the fifties and we have been, if you, eighty-four percent of everyone daily uses some form of artificial intelligence. If you have one of these seventy-seven percent of your apps, like your interactions will have something to do with artificial intelligence. If you've got your computer. if you're a big, social media fan, or you're actually working with an organization that already has your own machine learning models, so we have had artificial intelligence part of our lives for a really long time, but ways that are not consumer facing, like we don't, it doesn't say TikTok powered by AI algorithms like that isn't something that happened, even though it absolutely is. It's gonna get you to watch that cat video and learn about you every single time you interact with that technology. So it's been this kind of underlying technology that has become, a part of our lives and it has driven a lot of technology development. And what we see is that there was, the first like general bad that kind of hit the scene was IBM Watson. I don't know if anybody's a Jeopardy fan in the audience, that's almost a decade ago. And it did something pretty well. It beat Jeopardy, right? But it wasn't like. Making up new Jeopardy questions and answering and playing against three people. what we were saying is being more possible with generative ai and it took quite a bit of time and. What ended up happening is we had this like move towards more like deep learning and looking at how the brain works and creating these artificial neural networks like a transformer model, which is the underpinning technology for ChatGPT. And so there's a lot of development that happens between like 2018 and 2022 that allows us to see this really complex processing possible for the first time. And it shifts the way that we think about like technology and. Really, the reason why we're talking about this today is that OpenAI decided to put an experimental conversational AI tool into the world as an experiment and said, here you go. And November, 2022, ChatGPT 3.5 coming out like a couple months later. And then we had four coming out in March. What happened is the world became OpenAI's Guinea pig. And we, we say if it's free, you're the product. But that's what happened. And, but the thing is that this technology is like unlike anything that has come before it. If you've used a Siri and predictive text and a terrible chat bot online that you can never get to a representative, that is not a step change from ChatGPT. We're talking like this is a magnitude change in terms of capabilities, interaction, etc. And what happens is you have this acceleration that happened so quickly. and what we see is that, the, if you look at Facebook hitting a hundred million users, it takes about a decade. Even TikTok, everyone loves TikTok. It takes nine months. Even Threads, which was like just Instagram took months. ChatGPT took five weeks. And so suddenly you have this massive new technology that is consumer-facing in a way that has never happened before. it took a decades for the internet to impact our lives. It took the quantum of weeks, days for ChatGPT, and that's why we're in this position today at just about a year change later, is that this is transformative, but also consumer-facing in a way that's never been seen before.

Isar: 6:57

I love the way you framed it. I, when I talk to you and when I do this on stages, and I talk about this topic, there's two things that are very different about this than any other transformational change we had in history. And there've been a few, right? You had multiple steps in history. But these steps gets closer, like shorter and shorter in time. And as you mentioned now it's within weeks and things are changing regularly. But the other thing is how wide it is. So the number of use cases this solves. So if you think about the agriculture, change that we had, I dunno if you, okay. People built a tractor, so now a farmer could farm a lot more land with one person, but that's all it could do. It could go through his field. Then the internet came in the beginning it didn't do almost anything. Like the early days of the internet, you had a browser. It was very lame. There was like landing pages like Yahoo that you could get information. That's all it did. And it was like this for a few years. And now we have a tool that can do. So many things across what seems to be an endless number of use cases and new tools and new use cases are born every single day. And I, and the speed. And I think the combination of these two things, like how wide it is and how fast it's happening is what's overwhelming organizations. So now that we've framed the problem. What is the solution? Like how should I'm a manager, a leader, a CEO, a head of school, doesn't matter what I am, I'm in charge of the success of the organization that we're running. What do I need to do? Or maybe people underneath me, like what other people? I'm just a nobody in a organ, in a large organization. What can I do? what's the right approach in order to actually start having a structured implementation of this thing?

Amanda: 8:44

the best thing I can say is that let's not throw out. You know the things that work and there's a change management process If you're an organizational leader, and I see a couple here. Then please do not think that this is somehow something so special that like we can't go through a change management process and we think about change management. We're thinking about creating a couple things. One is creating a common understanding of what we are talking about and what actually needs to change. And I think that's the most important first step, is that because this technology is unlike anything that's come before it and is essentially you as an actor. Are going to be interacting directly with a computer interface, meaning like it's not, you are now a programmer for using generative AI and you're using ChatGPT or others, right? That it's really important to think about that, that part of this is just learning what this technology is and isn't. So for example, if I ask. a room of a thousand people, if you used it or not used it, there'll be a wide variance and we'll still in a very early adopter space, more and more people are using it for their own work, their productivity. Some of them are using it and not saying it because they don't know if it's appropriate or if guidelines will be changed, but we have a lot of people that have some deep and significant misconceptions. Around things like thinking versus computing. are these tools thinking, is this like Skynet or is this like data from, or Janet from, the good place, like a one I've recently heard. But I think that there's some big misconceptions, but some of this is just taking the time to learn. And so if you look at Slack, for example, slack took a week off. A week off to just focus on, they say ai, but really generative, right? And what does this mean for the organization, both internally and externally. So this is a moment where this cannot be another thing, like then we talk about change management. you commit to a process and you know it's big enough. I promise you right now, if you have not committed to a process around adopting generative ai. You need to today, or tomorrow or in a week. And the reason why I say that is because this is not only is it not going away, but it will be deeply embedded in almost every part of our lives. Over the next thirty-six months, Google Maps will have a button that tells you where to go to dinner because you're like, you're fighting with your, your friends about who, no one can decide. Google Maps will use gender to help you do that. So we're gonna see it in this very deep way. So I think that there is something to be said about the change management process. So building that base of understanding, spending some time, making it a priority, and then starting to identify use cases in which you have a problem that can be uniquely solved by generative ai. So it's not like this, like I love that you said something about is this idea that, it can do so many things and it absolutely can. And so this is where it's gonna get really interesting is that this does have multiple applications. But it is not an end-all solution. It is something that needs to have a lot of different thought and approaches, and it needs to be something that is, we need, you need to understand what your problem is and then look for a solution. And that's where a lot of people are living right now. And then it's time, like once you've done that, I'm sorry, identifying these use cases and I, and looking at the ways in which you can potentially increase productivity for your team. You can create better experiences for your, your customers through better interactions, whatever it may be. The next stage is like innovation. what are you actually gonna do to start bringing these technologies to drive real change in your organization? Because there is a technology that can do things that have never been done before. The only thing is like you can't start there because you don't know enough about it. To know what it's possible, most likely you have to build towards that space.

Isar: 12:12

I agree. I two points. one I wanna summarize two of the key things that you said. The first thing is awareness and education, right? You need, you said, that Slack stopped everything for a week. That's, think about it. It's a pretty big company, and they've stopped everybody to do this, so you need to make it clear for everyone that this process is happening, right? We are going to work on ai, we're going to work on understanding it. You need to lay out what your process is going to be, and we're gonna talk about this in a second. So this is step one. And something that I want to add to that actually Bob in the chat related to, it's not a one-time thing. I, when I teach my clients or when I teach on stages or my courses, it's. The number one thing that I tell people is continuous education, like this thing is constantly evolving. Now, an organization, especially large organization cannot change every day, every week. They just cannot. But you need somebody or a few people that are actually are paying attention, that are gonna filter the stuff that actually makes a difference. That going back to what you said as the second thing, figure out where can this provide value easily and with low risk. Then you can get quick wins. You can get people to start using it. You can get people to start understanding what this thing can do, and then, okay, then you go to the next thing and the next thing and the next thing. So having this mindset of this is gonna be continuously evolving and we need to create a process that will allow us to evolve with it so we can benefit from it is the right way to do this. What are the practical steps or mindset that you help your clients define in order to do this effectively?

Amanda: 13:50

Yeah, the first thing is let's create a space in where we have some time to, to learn and explore together. let's actually set aside with an expert. it doesn't have to be internal, someone that's good about it, like understanding the technology and can explain it in ways in which you can understand it. But let's experiment. Let's hang out. Let's try this. These tools are fascinating and they could be super weird and super funny and super oh my goodness, like this is something that I didn't know was possible. But the thing is this is not, if I go into my staff room and say, Hey, everybody. generative AI is gonna change the game and it can do all these things. Until you use it, until you see it, until you have the visceral reaction of doing it and seeing what's possible. It's very difficult to have that person's thing say that thing and then it be real to you because like how, you mean it could do an entire front end of a website with like in forty-five seconds now? No, that's you can talk about it, but until you show it and of course it's not gonna be perfect, but My goodness. what is that gonna change for us? Or it can gimme a pretty decent, like a GPT-4 class model. It can gimme a pretty decent starting place for a quick response to customer feedback or qualitative feedback. And it's not gonna be perfect, but maybe like it just helps me to bounce that idea off and does it in a minute or less. And I think that's where there's just a really strong opportunity to have some dedicated time with the right supports in a scaffolded way to just get in there and start using this. So the one of the ways that we do this with schools and is education institutions. So we has a, we have our prompt library. Our prompt library are prompts, so when you interact with ChatGPT or other tools, you're prompting you. That's the way that we call the interaction, the questions that you ask. And while it's not like a computer program, nutritional sense where you create a prompt, it's always gonna work. There are some best practices and some really good use cases that are commonly used like that email that from a customer that you don't wanna answer'cause you're afraid you might be a little bit snarky or not have the head space or whatever may have you. not using any personally identifiable information, you can use a prompt pretty reliably to help you with refining a response or at least getting some talking points. And so I think that's something that we see. So you can collect and create that for your own team. Like maybe there are 10 really good use cases for productivity that you feel confident in that will really help drive some, us usability, some productivity feel good. Do it in a safe and responsible way. So you're using it in the way that your people aren't creating enough unethical moments where they, or practices that they not realize potentially and just get in there. And I think that we skip over that as part with something new where if we even bring in a new technology, we don't have that time where you get a play around a demo. And sometimes, like something like a new CRM is not gonna be very fun to play around with. A GPT-4 class model right now, potentially Gemini Ultra and are advanced, and now GPT-4. It's a really good time. And so I think that's where we say is the first step is just giving that time to explore, have it be scaffolded, start to build some internal knowledge, and then when you go forward, you're talking from a space of shared understanding, of common experience, and that is gonna allow that next step to be a lot easier.

Isar: 17:07

So I love what you're saying on two aspects. One is the fact that if you don't experiment with this, you don't understand what it can do. And it's going back to what you said, it's the first time in history that a really highly capable software came with zero instructions. there is no user manual, there is no blog. It changes all the time. How to use it. okay. It's okay, go figure it out. I'm like, what do you mean? Go figure it out? tell me how to use this thing. no, it doesn't exist. And the way to know how to use this thing is to follow the right people, see what they're doing, experiment yourself and figure it out. And so that's one thing that you said that I really like. The other thing that I love that I actually don't say a lot, but I think it's a very good idea is give people a solid starting point. if you don't, if you give people a blank sheet of paper, it's very hard. But I say, okay, here are 10 prompts that you can use to do either this, or that. Now they're like, oh, okay. I. Now I understand how I can use this in the context of a specific use case, and I can see what it's doing and it's not a black box. I'm not clicking a button. I can actually see what's written and they're like, oh, what if I change this? What if I do that? Which is what you called, like Experimenting, like you can then play with it. So I think giving people, like you said, a safe space. A good starting point and a few use cases, they can start using that you spoon feed them, right? It's sort go figure it out. It's okay, here are stuff that we quote unquote the organization figure out that you can already start using. I have an interesting follow up question to that. Who in the organization creates these initial use cases and problems and based on what?

Amanda: 18:42

So I would say that there are probably some people in your organization already that are early adopters. These are the people that are like, it came out in the playground even before November, and they're like. Oh my goodness. Or the person that used it for the first time last spring and it changed their lives and they use it all the time and they're really encouraged by it. They're engaged, they read, they understand. I would tap those people with potentially someone within your organization that's focused on operations. Someone that's focused on those kind of day of the day tasks.'cause these productivity gains are that great low risk entry point, right? Like where are the places in which we can help our team? Do something that is going to increase the quality of their work and is going to allow for it to be done in a more efficient manner and or to extend their work in terms of a creative partner, a brainstorming partner, whatever may have you. And so I think that what you do is you already, I promise you, you might, it might not be someone you know, but there is someone within your organization or many people within your organization that would love the opportunity to do this in a productive way. But make sure that you're now combining them with someone that's going to know, first of all, they're gonna take into account the ethics and the protections of your organization. for example, the ChatGPT is considered to be a leaky. It's a leaky model, meaning that sometimes the things that you think are private will randomly show up in someone's chat box in China. Like it doesn't happen. Like it happens very small, but it can happen. There's also been experiences in which someone has entered something that is international intellectual property that has now been something that was shared with someone else, and Or things like that. There are also questions around your privacy of you're working with different organizations, you'll have different privacy components about identifiable information, etc. So you wanna make sure that you're having that ethical consideration as well as knowing that these models do not think, they are probabilistic models and they make stuff up. Literally that's how they work, is they are always making stuff up and sometimes those made up things are wrong and that's called hallucination. So you wanna make sure that you have those kind of considerations and then do it then. Then that's a great place though, because you're not asking your operations or your. Your compliance people, your cyber security people to do that on their own, but you're now partnering with those really excited people that are gonna bring the energy that are gonna be able to help you do that. And that would be my suggestion. Another suggestion that we see that's really interesting is that especially you have a risk or a change of verse group that you work with in one area. bring someone in that's an expert for that time too, to help you start figuring out those use cases. Because what'll also happen is we see this all the time, sometimes like you do have those early doctors, but just doesn't, they're just not being listened to because how does Terry from, HR suddenly an a expert, right? And so I think that there's an opportunity also to think about if you do know this is a big barrier or you don't feel that confident, then bringing in an organization or a partner or to do that work with you. A great way to get started and accelerate that too. You don't always have to do it on at your, on your own.

Isar: 21:50

No, I agree with you a hundred percent. I tell people, and that's what I do with my clients is, first of all, let's build a committee. And the committee is aligned a lot with what you're saying. You preferably want a person from each department. So in schools it would be different than in a company. But in a company it would be somebody from hr, somebody from finance, somebody from operations, somebody from sales, somebody from marketing. Why? Because A, they will have different needs. B, then you'll have a champion in each of the departments that will help you do it. But the other thing that I tell people is you want the people who are geeks. You want the people that will actually spend. Nine P.M. to midnight playing with this thing because they find it fun and cool and exciting because then they will give you solutions that otherwise will be work for people. And this is oh my God, this is like the coolest thing ever. I can do this and I can do that, and they're gonna experiment and they're gonna do the thing. But the other thing that you touched, another thing is very important, and that's to me the number one thing that committee needs to do as far as education is really tell people what are the boundaries, the do's and don'ts of when you shouldn't. What you shouldn't, what you can do, what you cannot do.'cause people don't know, like a lot of people don't know that the data you uploaded to it can be used for training of the model in some scenarios. people don't know that. It makes stuff up completely in the same, by the way, in the same level of confidence as the correct answers. And sometimes we'll mix the two together and some of it'll be correct and some of it'll be completely made up. And so the education and if I'll make it even more extreme to a sales organization. Salespeople are driven by results because they get paid commissions as a big part of their salaries. They might push the limits of what's acceptable or not acceptable from your organization's core values. Like I'm not willing to do this and that in order to get additional sales. But some people now have tools to do stuff that they were not able to do before. You can literally. Create stuff out of thin air that looks completely real, whether it's reports, images, videos, testimonials, anything you want. And to most organizations, that would be unacceptable to manipulate your clients in order to buy from you. But these things have to be defined by a group of people that are in charge of doing this. The other thing that it does, going back to your point, is now Terry from hr. Everybody knows she's a part of the HR committee, so by definition, they're gonna listen to her and follow what she says and so on, because there's gonna be a bind from the leadership team that has defined this committee. So it's a great way to do, what you suggested.

Amanda: 24:19

and I think that, to Matt's point in the channel, at your school or your, like an early doctor somewhere, and no one's listening to you, don't feel alone that's happening. we often think about these moments in time that it's it's my industry, my company's being slow or too risk averse, or think this is gonna go away. It's not your industry. It's not your company. It is happening all across. This is a unifying moment in time where we have the same level of confusion, concern, excitement, fear, like completely and totally no idea how to get started happening across everything from medicine. To sales to tech itself. I promise you that is true to schools and I think that this is an opportunity though, to recognize to everyone here is that this is so early still and there's this, like this axiom or this, you had a thing that like you're either too early or too late for technology. It might feel like you're already too late, but I promise you, you are not, because these models, not only are they incredibly unreliable in some ways, especially to be used in a consistent manner, but they're also extremely expensive. The processing and compute costs of these tools are can get incredibly expensive, especially if you're using the highest class model. And so while it might sound great, right now, you'll see sales earning calls at Microsoft. At Google and at OpenAI, the big players in the US on these pieces that are gonna say they are losing money on AI and it's hand over a fist like Microsoft,$30 a month for copilot it. It's$20 extra just a month of what they're losing on you if you're using it in a way. So I think that's something to consider. Is that this is still really early, the technology is the worst it's ever gonna be today, yesterday, and tomorrow. It's moving at a rapid change, but there's no need to like just go ham and go as fast as possible. But also if you are on that risk-averse place, these incremental changes, this openness to trying and experimenting are there so that you aren't too late. When it, when we get to that place where these models and these use cases and these enterprise tools are in a better position to help you meet your client's needs,

Isar: 26:41

Great point. I wanna refer again to what Matt said, and I'll read to you what he's saying. He's saying, I'm trying to spread the word, but I'm being met with major resistance. And that works again in school districts and that works in organizations. And the way to address this is to show a success on a specific use case. Take a use case that is not risky to the organization, whatever that may be. Develop it, show research. Here's what I've done. I've used it for three weeks in this scenario with this group of people to achieve this goal. Here's the results that we are seeing. Allow me to spread that. Now. It's not, oh, I want to use AI across the organization or across the school or across teaching. It's this use case or two use cases. If you pick the right use case, there's no reason people will say no. If you're getting 30, 40% benefits of efficiency on something that is not putting anything at risk, people will say, yes. Once this is implemented, then you got more people to buy in are interested. Okay, now come up with two more use cases. And that's like the crack in the door where you've. You can slowly open the door and get in and put it into more and more, places in your organization. I wanna go back to your framework. So we touched on two aspects of it. You said learn, and then you said define guidelines. What is the next step that you recommend to your organization that they would do?

Amanda: 27:56

Yeah, so the first of all, also like we, you're gonna see more like organizations, especially if you're working in like you're doing RFPs or others that are, look for those AI guidelines and always be clear that they're generative AI guidelines within an AI guideline like policy. So specificity language, very important. So yeah, so define those guidelines and use, do those pieces. And then I think it's like time to start. Experimenting. And so like we talked about some use cases, but like this is like the time to actually you talked about skunk, like the idea of the geeks. What are your skunk work projects in which you could actually pull up some ideas about what this can really do. And I think this is something that gets really fascinating pretty quickly, is that why I talked about the expense of these tools? That's the expense of these tools at scale. The expense of these tools of even taking an in. Since on Microsoft Azure, another platform like I can use whatever your poison is, your Amazon using Bedrock, whatever you wanna use, right? and throw up something and build internal tools or build some stuff and start to play around, because that's going to be the special stuff. And what I, and I see that's not happening so much, and I, and that this idea that our imagination hasn't quite caught up yet with what the technology can do. And so I think that this is an opportunity to have some time to, to have those people like you've built the base, right? People are now building their productivity, they're using it responsibly. But now what is the next version of. Your technology, your work, because that is going to be, in my mind, the defensible place to use generative AI or AI writ large because, and that is what is missing right now. You've got a lot of big players that are focused on frontier models, right? The underpinning, right? But you have a lot of small players or individual players that are building these point solutions or these very what do we need to do better today? But very few of them are thinking about what could we do for the spaceship world, like the crazy world that's coming. And I think that's an opportunity to do that and to activate that because that's going to be the difference.'cause you're gonna start to see like OpenAI is gonna release a tool that is going to, like this has happened multiple times. It's gonna take out entire categories of different types of companies. It happened, it's happening and a lot of these right now. Not just generally bad companies or AI companies, these are companies that like copywriting, graphic design. You see this happening. So if you guys, you, I would really highly suggest at that next stage, where are you investing in the experimentation, animation, and creating space for that to happen? Because I. It's going to be an existential question pretty soon, especially in high-knowledge skilled worker type of places. The IMF put out a report that said about 40% of all jobs will be impacted by generative ai, but that is 70% in the U.S, 60% in the U.K, but 70% in the U.S is the expectation.

Isar: 30:40

great points. I think the first step for people who haven't played with it is you started with the basic right. Have a prompt library that really gives some kind of standard to people to start with. The second stage of that is you can create your own GPTs right now on OpenAI's platform. And for those of you who don't know what a GPT is, the problem with ChatGPT is that it's so broad, right? oh, it can do anything because it can do anything. It's so overwhelming. You don't really know what to do. A GPT is like a mini version of ChatGPT that lives within ChatGPT, but it's a mini version of ChatGPT that's geared to do something very specific. If we'll take you to the school world, how do I write a math test for the first semester of, fourth grade? You can create a GPT that will do that. You can create a GPT that helps you answer customer service emails. You can create a GPT that will help you create marketing content. You can create a GPT that will help you innovate and think about the next version of your product like, but each and every one of them is something you can do with people who are special with their specialty. And like you're saying, it brings the organization secret sauce, right? You can take documents. Agendas, concepts, ideas, and bring this into those GPTs that now only your organization can use to gain a benefit that other people cannot use because you're taking the capability of the model, combining it with know-how that only your people have to create something that is unique to you, that will allow you to excel beyond your competition.

Amanda: 32:09

And I like, if you think about GPT, there's a lot of different ways to do, there's GPT trainer and others like Chatbase. There are all kinds of playgrounds. It doesn't have to be with, it's actually probably the most expensive to use as a large organization, the enterprise solution from OpenAI. but if you use a GPT for teams with them, twenty-five dollars a month, if you're paying for the yearly piece, you can create it. They data's not trained on. You can create your own GPTs that are shareable, but also they're like really unreliable everybody. If you create a GPT, I saw that Matt's creating it. If you're creating a GPT, it is so easy for me as someone else to have that GPT tell me all its training data. Yeah, all its directions and its system prompt, so like these are not safe systems and just be aware of that. Again, I will always advocate for the outside of the experiment, like out outside of the getting people involved. Think about pulling these tools internally and finding your sandbox and going off of OpenAI. And going into a Microsoft, like other places that can give you some security around not training your data building solutions. You can also do it in whatever, if you're, have a very technically advanced team, throw up whatever sandbox you wanna use and then use an API into the best models, right? So I'm still suggesting that. But there are also a lot of open source models that are improving every day. You've got Mistral and Mixtral, the model, the coming out, smaller language models as well. More point solutions that are the places to try. But the fascinating thing is that still today, even fine-tune models like that are focused on a specific thing with good prompting. Do not, are not more, are not better reliably than a GPT-4 with good prompting. Even without fine-tuning. And so there's something to be said too about like where are the places in which you have your internal people, maybe having them be able to continue to grow their ability to prompt and use these tools for the foreseeable future. But then if you're building something, make sure, like I really highly suggest it'll be cheaper for you and it'll be safer for you if you're throwing up your own instance with API's.

Isar: 34:21

Correct. And to touch on two points, that you mentioned as far as safety in these models. First of all, Hugging face, literally just this week, announced that they have a GPT kind of solution. You can still not upload your own files to it and so on, but I'm sure that's coming as the next step. So over there, it's an open source. You can run it on your own servers. Nobody's training on the data. Nobody has access on the data. But as far as the GPT side, yes, it's very easy to quote unquote hack GPTs, but you don't have to publish them to the world. So you can just keep them within your team and then nobody else has access to them. So people within the organization can figure out how they work. So I always, I still think from a user friendliness perspective, without a more advanced team, or at least a few people in your organization that can figure out how to build a sandbox with an open source model and then run on top of that, the easiest thing. Get a teams license for ChatGPT. Have a few people have access to it. Run your own GPTs only internally, so nobody has access to it. if you are using the teams model, they're not training on your data either. So it's a relatively safe starting point. I agree with you a hundred percent that if you want scale this to, oh, now I want every student in the district to use this. You probably want a different solution, but as an experimentation. Easy, fast, quick and dirty thing to see if this thing does what it needs to do. That's probably the easiest way to do it.

Amanda: 35:44

Yeah, I totally agree. Like path of least resistance as well. But I think that if we move towards that experimentation phase, like that's where I'd like again, but I'm also very privacy minded and very safety minded. I think that's just something like, like I was talking to a district, a large district yesterday about like safety and some of it's if you can try your best to teach people how to use these tools correctly. Or you can create a sandbox and a place in which they can be used like ethically and, some of these cases. So I'm always gonna err on that side. but I totally agree that the path of least resistance, the best way to get started to me is at foundational stage. A GPT teams is a great place to do it. And 300 bucks a month for, a group of, a like our size business. That's a no-brainer.

Isar: 36:32

I agree a hundred percent. Let's take a few examples. I'm just curious to see, in your field, what do you see as far as innovative things that some of the people you're working with are doing with this within schools or companies and so on that are taking really the extra step beyond, oh, we figured out how to prompt this thing.

Amanda: 36:51

Yeah, I think the fascinating thing is that the innovation right now is not what I would call innovation. The innovation right now is adoption. But it's innovation because it's so new and so different. I don't see significant like moments where, and I don't think I'm alone on this. I spoke at Stanford last week and there were the closing with Sal Khan and, Christopher Gatch, she's a computer science professor, and they both said we, I haven't seen anything that like, has like. Really blown me away in terms of what will education become? And I think that we are not, we're not moonshotting right now. And so like true innovation, I haven't seen as much, but I've seen interesting and more innovative adoption, which is where we. Start to see like real commitment to building knowledge, to identifying use cases, to creating spaces to work with students with disabilities in ways of in that have never happened before, to ways to truly differentiate and update content based on student need. Skill level and interests that like that are pretty basic. we're not talking go out and create 15 bots, but like, how do you take this one research guide and make 25 in the same amount of time? It'd take you to make one. there's some really fascinating things here and making it consistent. Trying to drive like. Cohorts of people that get that space to learn and bring it back and then start to build those embedded practices. But what, and I think it's happening in pockets and some that are adopting and trying, like New York City and LA which are the two largest districts in the country, are building essentially their own generative AI tools. like our sandboxes. Instead of saying, let's open this up, and they're doing it in a way in which they're able to control. Like this is one thing that's really interesting about this moment in time. Because at schools, for example, I've always given away A lot of data. A lot of data in safe, hopefully safe ways, but this is time where you don't have to give it away. Like you can create, if you have that much money behind you and you have that bitch buying power and you have some teens like. Keep that data for yourself and build that knowledge of what it is. And I think that's really interesting. I think we're a little bit like, I think in education we're still in that early doctor land so much that, so much about it is just, I. Basic adoption, removing the stigma about use of these tools from students and potentially teachers and to create the space for that to happen. And that's where I see the majority of these early schools really and districts focusing on. And I know I'm saying that there are tons of people individually doing amazing things where they're having kids. Like submit prompts and learn how to prompt engineer and create, not go for an essay, but you now you create a whole business plan. there are some really beautiful opportunities, but they're happening in pockets, in silos, and there hasn't yet been a cohesion of like best practice in or innovation. Yeah,

Isar: 39:49

I, and by the way, it's the same thing in organizations, right? You have people who are early adapters, like in businesses who are early adapters, who find ways to use it in the organization. And but there's no coherent strategy on how to use it across a thing. And in the school world it's even bigger, right? Because it's not just, okay, we figured it out in the school. Then you have the county, the district, the state, the federal, there's so many other layers to the education system that needs to eventually, preferably sooner than later, figure it out. People like, unless you have anything to add, I wanna let you share with people how they can, work with you, follow you, learn more about this topic. But if you have anything to add, then start with that.

Amanda: 40:31

Yeah, no, I think this is, I think that to everyone here, this is just a really unique opportunity to build and create like never before. And I think that the way I think about it is that you're limited right now by your creativity. Your resilience, meaning like you don't always get what you want out of a bot, and your ability to prompt right now is the limits of what you can do as long as you're not trying to do fifth grade math because these large English models are absolutely trash at math. But I think horrible math. Yeah. And that is going to be the move to, the move towards like artificial general intelligence will be where we start to see like this reasoning, this math reasoning and, will be part of that process. But what I think is really cool though is this is again, the limits are really the ones you put on yourself right now. And so if you're an organization and this is exciting and you're thinking about this, like there are opportunities to start to build this. This, take this assignment and start to try to build it within your organization. we do a lot of work around, like strategy and adoption. We do a lot of free webinars. I, we have a women AI and education group that Marta, who is here with us today, and the audience is a part of, we have a lot of ways to interact with us. I'm also super on LinkedIn. I think that's how I'm as our, as a non-social media person, it's quite. Funny to be this on LinkedIn, but we find it a really great way to be able to share resources. And a lot of, there are education, but there's so much more. I think we see that over and over again that the approaches that we're taking in the education field are ones in which, if you took out education, would work across any organization. So feel free to connect that way and just really appreciate the time and the ability to be with you all today and hope that you're enjoying your journey too, and that it's helping you out on your own productivity and creativity.

Isar: 42:13

Amanda, thank you so much. It's been a great conversation. I think we are, we discussed a lot of topics that so many people are struggling with right now. And I, if I summarize the key points that we talked about, one is you want to experiment in a safe way. And as the leader of our organization, you want to create that sandbox that people can experiment in a safe way. Two, you wanna start bringing in your own information into this, again, in a closed box, not risking it so you can build stuff that is yours and that you can innovate. We touched on the point, if you're a quote unquote, a nobody, a small cog in a large machine, you can still be proactive and find specific use cases and through that grow within your organization. And in many cases, you may become. The person that's in charge of that innovation, which will give you a very exciting new role that you can do. So there's, we touched on a lot of great points. we also touched on, the world between closed. Companies like OpenAI and Microsoft, but also that just the whole open source universe that is incredible right now and it's thriving and it's easy to use. You don't have to be like a crazy tech person to be able to use some of these tools. So lots and lots of great information. I really appreciate you joining and sharing with us today.

Amanda: 43:32

Thank you. And hope everyone has a good day. Thank you so much. we appreciate it and enjoy your day.