Please note : This transcript is automatically generated and provided for your convenience.

[00:00:00]

Georgie Peake: It's like having an expert in your pocket.


I've been surprised at how realistic you can get your AI counterpart


Everybody having their own AI personal tutor


Rob D. Willis: welcome to superpowered with me, Rob d Willis. Each week I talk to leaders about their superpowers, how they got them, and how you can get a little bit of them as well. If you are new here, you are awesome. Please make sure to subscribe wherever you listen to podcasts.

We've got some amazing guests coming every single week, and today is no different. I'm talking to Georgie Peak about how to use AI for learning in a world where we are. Constantly bombarded with content courses, learning resources. I would say the big challenge isn't actually accessing information anymore.

It is making knowledge stick and making it affect our behavior and how we approach [00:01:00] our work and our life. And Georgie has spent a good part of her career solving this problem first as a data strategist, and now as C of rehearsal.ai, helping people turn knowledge into practical skills through AI powered practice.

Can't wait to hear what she has to say. Georgie,

welcome to the show.

Georgie Peake: Hey, Rob,

it's

Rob D. Willis: Great to be here.

Georgie Peake: yeah, thank you so much for

a kind

introduction.

Rob D. Willis: Thank you for coming on. For any listeners who haven't heard about you or rehearsal.ai yet, could you just tell, tell us in your own words who you are and what you do?

Georgie Peake: Sure. Yeah. So hi everyone. I'm Georgie. So I've been working in the product and AI space for about the last 10 years now, and that's been across consulting, my first job at IBM and then media and Entertainment at Channel four. And in the last four years it's been in education and learning and development.

So that's at a company called Filtered at memorize [00:02:00] language learning company. And then most recently I have co-founded a company called Rehearsal. Which is a SaaS business where we help communication experts create interactive experiences or tools for their learners to practice with ai. Such their training really, really sticks. So to make that a little bit more concrete, we work with an expert on difficult conversations, who goes into companies and trains teams on how to have these important but difficult conversations. However, when the learners leave training room, quite often, week later, a month later, trying to put those skills into practice have a real difficult conversation, training's all gone a bit blurry.

Can't really remember what you were taught, however good the training was in the room. So we're trying to basically bridge that gap putting skills learned into practice by helping the experts create these interactive experiences to allow their learners to [00:03:00] practice their skills over time and improve their recall.

Rob D. Willis: It's such a great asset to have because of course, you know, you could do this just by having the experts stay with the people and practice with them all the time. But that's not really gonna be financially viable for almost everyone in the entire world. So you are helping people get the kind of support that only the elite would've had not that long ago.

So I'm gonna be really excited to, to talk about that. But let's just start by talking on a very, very basic level

Georgie Peake: Mm

Rob D. Willis: about AI and learning. What do you feel AI can do, in particular AI

Georgie Peake: mm-hmm.

Rob D. Willis: us learn more effectively?

Georgie Peake: So you touched upon availability there, like always on after you leave training. Versus not having the expert there with you physically. It's like having an expert in your pocket. So that, that's a really cool one. But to think through some others, one of [00:04:00] the big ones for me that it's safe and judgment free. So you can practice these quite high stakes conversations in a low stakes environment, which just really hasn't been possible before. So we can practice with humans, and if you're lucky, you'll get a, a sort of a safe and, and judgment free human.

But on the whole, we've got, you know, relationships with humans, everything's kind of imbued with how we feel in our relationship with the person that we're practicing with. And that will shape how we internalize the feedback they give. AI removes a huge amount of that. So you can practice with ai. It yeah, won't pass judgment. It removes that sort of fear of making mistakes. And then another huge benefit is the personalization of the feedback. So this ties in with sort of the availability angle. In a workshop you might get some expert feedback, but perhaps [00:05:00] the the trainer won't have time to give you really individual one-on-one feedback, 20 people in a room.

It's just quite hard to scale with, with one trainer versus after the workshop. If you can practice on your own with the ai. Using the expert's content and expertise, you can get this really sort of personalized feedback on exactly where you rob as an individual could, what you did well and what you could improve, which just I don't think has really been possible at any kind of scale previously. So I think those are two yeah. Really amazing things that AI opens up many possibilities for the future of learning and practice.


Rob D. Willis: Do you think the reason that it is safe is just because it's a machine and people are not worried about. The machine essentially, or is it because of where they're doing it? Why do you think the, the safety comes from and could that ever change, do you think?

Georgie Peake: So I think it's a [00:06:00] combination of factors, but I do think, yeah, removing the, so there's positive and negatives, removing the relationship element and your fear. Fear of making a fool

of

Rob D. Willis: Mm-hmm.

Georgie Peake: of your manager or your

peer. I think that's of. And then, yeah, the on any time availability, sort of being able to practice where you are comfortable versus having to, yeah, practice in a, in a meeting room if you ask your manager for sort of half an hour to help you practice before a presentation, for example. so it's kind of practicing on your terms, which I think, yeah, you can create the environment that you feel safe

in.

Rob D. Willis: Yeah.

Georgie Peake: With that sort of

judgment free aspect of, of removing the relationship. Of course there are huge benefits to practicing with humans. I don't think these are you know, perfect substitutes. but I think it's, yeah, creating an amazing compliment that you can practice

Rob D. Willis: Mm-hmm.

Georgie Peake: and get the the benefits of their expert feedback, but also [00:07:00] I. Of being able to practice with AI

on your

Rob D. Willis: Mm-hmm.

Georgie Peake: without somebody else being able to see is, is really awesome.

Rob D. Willis: Because I'm, I'm thinking that occasionally there are moments where the discomfort is kind of exactly what people need to confront.

Is there a way, do you reckon, or a possibility that AI can be used to practice that discomfort or to at least get you to a position where you are closer to it? The real life scenario.

Georgie Peake: I've been surprised at how realistic you can get your AI counterpart to some of the sort of emotional realism of a real life conversation. So I'll take an example for one of the scenarios I call them that I've built for a customer. So practicing a difficult conversation, we've purposely created a character, it's called Resistant and Defensive.

So it's, quite [00:08:00] rude, responds very, feedback is really quite resistant. And you, and you talk to it and they're like, I know it's an ai, I know it's a machine. I really do have quite an emotional reaction when it responds being very resistant. And it kind of surprised me the first time it happened. sort of, I, I would, my mind wouldn't sort of let me have that emotional reaction. And this was, this goes back to sort of even some of the earlier models and they're getting better and better all the time. And, you know, you've got these sort of emotional realism in the voice and the tone of some of the more recent speech to text models. So yeah, I think I've been, I've been surprised at how

you can

Rob D. Willis: Mm.

Georgie Peake: close to emulating some of that discomfort than I, than I expected.

Rob D. Willis: What do you see as some possibilities in this area or the future or where things might go next in terms of how we're able to interact with machines

in these sort of practice environments?

Georgie Peake: So I mean, I think what [00:09:00] one of the major themes is going to be the coming years is AI personal tutors. So everybody having their own AI personal tutor exactly how that happens. Who builds it. I,

I'm not

Rob D. Willis: sure

Georgie Peake: And there's various sort of

challenges to overcome, but I think that's just gonna be an amazing democratization of yeah, personal

tutoring,

if you've heard of the Two Sigma

problem.

Rob D. Willis: Mm-hmm.

Georgie Peake: research back in the 1980s by Benjamin Bloom and showing sort of a two sig, a two standard deviations difference in performance for students who had one-to-one tutoring compared to a classroom.

Rob D. Willis: Yeah.

Georgie Peake: as I think you said at the, at the top of this episode,

that, you know, that's incredibly expensive and just not available for the, the vast majority of us.

But AI is, I think in the coming years going to make that possible so you can have one-to-one really personalized to how you learn, [00:10:00]

Rob D. Willis: Mm-hmm.

Just right difficulty. that, yeah, will empower us to sort of in that one-to-one setting.

Let's say someone is trying to develop a skill with rudimentary AI tools. What are the elements that you need to set in place in order for it to be productive? Accurate, you know, because hallucinations, you don't want to be learning the wrong thing for sure.

Rob D. Willis: And just generally a good, very, very basic element. Nowhere near what you have with rehearsal, but the very basic process that someone could put in place. What are the essential things they need to have?

Georgie Peake: Ah, that's a great question. And so I will start by saying, and this I'm taking inspiration from Ethan Molik, who I, I follow closely. He's an amazing commentator on AI and its use first sort of work and life and education. He's a professor at Wharton. He's described in [00:11:00] his book Co Intelligence, how we need experts at this point in time to be able to tell where AI has those those gaps.

You mentioned hallucinations, which are, I think, a, it significantly decreased over time, but they still for sure exist. able to spot those has become, you know, quite subtle or spot mistakes has become really quite subtle with how advanced

Rob D. Willis: Mm-hmm.

Georgie Peake: are becoming. So I think one of the first steps I'd suggest for anybody sort of wanting to. up to speed with how to use AI for learning is to, at first use it for something that you are already an expert at,

Rob D. Willis: Okay.

Georgie Peake: sort of develop an intuition of, you know, where it's working really well and where it's not quite there yet and, and needs extra inputs, extra prompting or maybe just extra sort of research development over time. I think that's one of the sort of main first steps I would really highly [00:12:00] recommend use it something that you are an expert in, such that you can tell or not it's, it's it's doing a good job for your use case.

Rob D. Willis: That's something that our friend Greg said is this is gonna be a skill that we need to develop is understanding the relationship between the input and the output, like what I need to put in. And so you get a predict, predictable thing coming out of it.

Georgie Peake: Yep.

Rob D. Willis: Yeah. So.

Georgie Peake: For sure. And that's only gonna, I

think was also the challenge. And I, I consider myself a, a power user of ai and I'm still quite overwhelmed the amount of like, change I can't keep up. I'm sure that's a familiar skill to a familiar feeling, to lots of your, to your listeners. I think what I was just gonna say is that relationship with inputs and outputs, that's only gonna change over time as well.

So whilst you might get familiar with, if I put this type of input, this prompt into the AI and it reliably produces this kind of

output,

Rob D. Willis: Mm-hmm.

Georgie Peake: that relationship

might change with the next version of these models. [00:13:00] So I think we're gonna have to sort of learn how to adapt to that kind of

Rob D. Willis: Mm-hmm.

Georgie Peake: well. But I, yeah,

that for sure, there'll be some sort of meta level principles of how we communicate with these with these systems. Definitely.

Rob D. Willis: So let's, like, let's put it in those terms then to build this kind of set of tools for learning. Let's say we're trying to, we've got this sort of being,

Georgie Peake: Hmm.

Rob D. Willis: we're the first step in using it for learning is kind of getting to know them a little bit.

Georgie Peake: Mm-hmm.

Rob D. Willis: is kind of how they work, what they react well to, what else do I need in order to create.

A learning set, let's say that I want to learn, I mean, maybe this is cheating from my perspective. I want to learn how to do storytelling more effectively. 'cause that's literally why I teach people. But let's say I want to make myself redundant and create the ultimate tutor in that. I've got used to work understanding how the AI works with inputs and outputs.

What [00:14:00] else would I need with this quite like high level skill? What else do you think I should put into it? Or what resources should I find?

Georgie Peake: I think one of the next steps would be to think through and the AI can be a thought partner in this, your learning objectives. So what do you want to achieve? 'cause storytelling can mean a lot of things for, to people at different levels of their storytelling journey. giving the AI a sense of what you as an individual want to get out of an experience

Rob D. Willis: Mm-hmm.

Georgie Peake: that it can adapt to your, to experience level and to where you ultimately want to be able to, to get, I think it will do. A really good job of understanding

Rob D. Willis: Mm-hmm.

Georgie Peake: can then sort of suggest a

path to, get you to achieve the objectives that you described to it and it would be able to help you a step, take a step before that, [00:15:00] think through potential, objectives and how you might want to think about developing the skill if those aren't already sort of known or aware to you.

Rob D. Willis: So we've given this being, we know how they communicate and how they, we, they, we can work with them. We've given them the sort of North star. That's where we want to go to. What else do we need to give them?

Georgie Peake: An idea of your learning preferences. So for example I think I learn quite well from reading and then having to try and explain a concept to a non-expert

Rob D. Willis: Mm-hmm.

Georgie Peake: see whether or not they understand.

almost I can't quite remember the the learning design term for it.

Not quite a sort of flipped classroom, but you could act like. The AI is the student and you are the

teacher.

Try explaining a concept

to the AI [00:16:00] and it will evaluate how well you did at explaining the

concept.

That's one example of like

how one individual might learn. And I think that works quite well for me, but for somebody else a different kind of sort of practice exercise might work and you don't know what your own sort of personal learning

styles

And so asking the AI to help you understand that it could ask you some, some questions to understand your own learning style and preferences

Or indeed you, you can

describe them to it, such that it will be able to cater to that in any learning journey or experiences that it designs for you.

Rob D. Willis: And what else could we need? We've got the learning preferences now. Is there anything else that it requires?

Georgie Peake: Depending on the skillset itself, it needs the expertise. So it's been trained as I'm, I'm sure you'll be aware on a huge amount of the internet,

[00:17:00] books, yeah, a huge amount of of the English language internet. if it's a particularly nuanced or or niche skill, you might need to provide some additional content. So when, when we work with experts from rehearsal, for example, we will teach the AI about that expert skillset. So teach it about a particular form of negotiation.

Chat, GPT

or, or one of the AI providers will have a great sense of what negotiation is and some of the classic methods, but for a particular style as taught by one of my expert customers, you need to be able to teach it

what the,

what those specific skills are such that it

can then, translate those into all sorts of experiences for the

learner.

Rob D. Willis: Yeah, yeah, because that's the. The thing about getting to a point of excellence, I feel [00:18:00] like the, the internet is without, it's kind of an average of everything. So it's, it's always gonna be a kind of diluted version of everything. Having particular academic papers or research that you know, is good, high quality content, I would call it,

is such an important, important aspect.

I mean, there are tools like Notebook, LM and so on where you can put all these sources in. Is that the kind of thing you would do if you are using as sort of amateur set of tools? Yeah.

Georgie Peake: For sure. And I think some experts will give access to their versions of these bodies of knowledge where you can interact, with the AI version of their expertise and content, which is very exciting. Indeed, I think many are, are using hel,

Rob D. Willis: And

Georgie Peake: example of that.

Rob D. Willis: Let's go move on to just hearing a bit about rehearsal. .

what inspired you to create

the company?

Georgie Peake: So the way I think [00:19:00] about it is I break it down sort of why conversations and communication, because that's what our, our sort of real focus is on helping experts in those areas. practice and why ai? So with why and communication. I don't think I thought about this too, sort of. Actively or intentionally when I was younger. and then I just sort of started noticing things both at work and in my general life and realized like I'm not particularly good at them myself. For example, I'd be having a conversation with someone. I'd be, you know, thinking about what am I gonna say next or be my next question, or yeah, what can I say next without, you know, fully listening to what they're saying. sort of habits like that. I don't mean it in a hugely weighted way, but I just yeah, develop this real interest in, in communication and conversations. And whilst we have all these sort of little habits, conversation is communicated, so they're just [00:20:00] incredibly important. I think it's quite often cited, sort of one of the number one sort of regrets later in life is spending not enough time on connection and communication.

So it's just an area that sort of really fascinates me, something that I want to get better at myself. so that really sort of sparked my initial interest. And again, I am, I'm not an expert in these. I'm very much on the, on the learner side, but because it's something that I am passionate about and want to learn myself, I will read books about it.

I will listen to podcasts about it and when the opportunity arises at work, attend workshops. yeah, I even recognized in myself not necessarily always being able to take those skills and put them into practice. So that sort of takes me onto the why practice, and as I said at the start, this gap between having these amazing teachers on these courses and books and [00:21:00] podcasts, but not. Having an opportunity to, yeah. Practice in this safe and judgment free environment with expert feedback. So that sort of, yeah, really interest me in the importance of

And then why ai, that's sort of where my, my expertise comes in. It's been a sort of a large part of my career, sort of back to sort of some traditional machine learning before the more recent developments in, in generator of ai. But I think it's really sort of, as I've said, sort of uniquely and counterintuitively placed to help us learn human skills. And as, as we've sort of chatted through in quite a lot of detail, create these practice environments where you can learn in these safe judgment free environments with expert feedback.

So yeah, the combination of those three sort of is where rehearsal came from, me sort of. Struggling with some difficult conversations in, in the past in my career, and [00:22:00] wishing I had an opportunity to practice. And now I

think

Sort of enabler for us to be able to, to practice and get the feedback that we need.

Rob D. Willis: If you were to pick which of availability, safe and judgment free and personalization, which one was really lacking for you, that made you want to solve this problem?

Georgie Peake: I've not been asked that before. I think my instinct is availability and maybe some subtle sort of the safe and judgment free aspects there. I am not great at asking for

help.

Rob D. Willis: Mm-hmm.

Georgie Peake: So maybe somebody that is good at asking

for help, the availability benefit might, be less sort of prevalent for them, but for me, and it ties in with a safe and judgment free, but having a resource that I can go to at any point and ask for ask for help, ask for practice, and ask for feedback, is a, is a huge benefit. [00:23:00] Yeah, I can see, I can imagine different people's having, having different answers to that, but that's my, my instinct for myself.

Rob D. Willis: Let's just go back to you for a little bit I'd like to go a little bit further back. You, I'd love you to just tell me a little bit about the afterschool learning program that you were involved in.

Georgie Peake: Yeah. So my first job was at which is a, an after school. Learning company for learning maths and English. Think it started in Japan and its basic principle is sort of short, incremental worksheets, I think sort of 30 minutes a day. And you complete these at home and then twice a week you go into QM on sent to and you get support from instructors. And so I was, I wasn't a, a sort of qualified instructor, but I was there helping sort of with the huge amounts of marking these worksheets. But I was sort of sat amongst the students who were a yeah, a wide range of ages, skills, and they would inevitably ask for some help with their, with their maths or their [00:24:00] English worksheets. think that was one of the first times where, up until then, so I must have been, my time is a bit funny on this, but sort of 14, 15, 16 maybe. And up until then you've even been in school yourself, even learning and on the learning journey, I've, I don't think I'd previously had to try to teach or try to explain. So that was a really interesting and sort of famously only really sort of understood something when you can effectively explain and and teach it to someone else. So that was funny, like sort of an interesting experience for me, like how, how do I somebody four plus four equals eight or some sort of literary technique that was in their worksheet, which, you know, had been several years previous in, in my education went up by the time I was that age.

But really having to think, how can I explain. concepts.

And I guess I'm sort of

thinking there of that sort of being analogous to translation

and

Rob D. Willis: Yeah, absolutely.

Georgie Peake: I've had of that.

Rob D. Willis: And it's also [00:25:00] explaining to someone who doesn't understand it yet, which is one of the biggest problems I saw a lecture, I can't remember, I think it was a a, a writing lecture for academic writers. And the criticism he made of education is that we basically write articles to show that we understood the thing.

To someone who already understands it and they only read it because they're being paid to read it. They're not being reading it because they don't understand it or they're interested in it.

They're just reading it to check that you understand.

Georgie Peake: hmm.

Rob D. Willis: And so it's very different where you have to either help someone who doesn't understand something

Georgie Peake: Yep.

Rob D. Willis: and maybe don't even care about the thing as well.

And that's something which we are not really taught, which you were thrown into doing when you were 14.

So it must have been a really important skill. Yeah.

Georgie Peake: Yeah. To be honest, that's only kind of a recent reflection of mine. So I'm gonna, I'm gonna Yeah. Think about that more. But when we sort of started thinking about [00:26:00] translation that did come up for me. I'm fascinating. That example you just share. I feel in the early days when sort of AI was, it was coming up, one of the quite popular ways to interact with it, it'd be to like, explain this to me.

Like I am, like I'm

five or

Rob D. Willis: Yeah.

Georgie Peake: or that was sort of one of the quite

common and, and and effective ways, which, that kind of you've given there of, you know, content for people who already have an understanding.

Rob D. Willis: Yeah.

Georgie Peake: would be able to sort of help

translate, translate

that

Rob D. Willis: it.

Georgie Peake: Hmm. To some rapid fire. So just need short answers here. First idea is best idea. I always find, have you got a favorite book about learning or skill development?

Co Intelligence by Ethan Molik. so that covers learning and development as well as, as other topics for work and life in learning in the, in the world of ai. So that's a really recommended

Rob D. Willis: Have you got a habit which you use to learn [00:27:00] yourself?

Georgie Peake: So I think I'd use a meta habit and recommend trying out time boxing. Indeed I've got a recommendation of a book from a good friend of mine, mark Sal Sanders, who's written a book about time boxing, but he'd encourage develop a, a learning habit, sort of putting and I, I say here sort of could be 15 minutes or, or 30 minutes time boxes in your calendar to develop a habit.

And I'd apply that to learning. So if I'm trying to learn something, gonna put these short into my calendar, which the spacing effect and being able to practice space skill over time rather than just in sort of cramming

Rob D. Willis: Mm-hmm.

Georgie Peake: sort of one bulk has been,

Proven to increase recall and learning effectiveness.

So I'd really yeah, that would be sort of the habit I'd think about is sort of time boxing my learning,

Rob D. Willis: Yeah.

Georgie Peake: I I stick to it.

Rob D. Willis: Awesome. Let's move on to the listener challenge, and in this part of the pod, we give listeners an exercise or a ritual, [00:28:00] just a little thing they can try out over the next week to get a little bit of your superpower.

what

have you got for us?

Georgie Peake: So what I'm going to suggest, and again, taking inspiration, I've mentioned 'em both already. So lic um, recommends just getting started with AI if you've not already spending sort of about 10 hours or so with it to sort of develop your intuitions about how it works, what it's good at, what it's not necessarily so good at. I'm gonna take that and just encourage you to give it a go and I'll, I'll give you a specific exercise in a second. And also good friend Mark and and time boxing, just put, and, and I encourage you to get your, get your phones or your laptop right now and put sort of. Can be five up to 15 minutes time box in your calendar every day for the next seven days to, to practice and think about a recent topic or a recent skill that you've learned that you want to, yeah, make sure you can remember and put into practice. And then, yeah, when you get to each time box, you would log into one of the main AI [00:29:00] providers and say something like, I've recently learned about x, let's say negotiation, for example. For the next seven days, I want you to act as my personal tutor and practice partner to improve my recall and deepen my understanding about x negotiation in, in my example. And maybe you'd say something like, please suggest seven short activities we can work on together over the next seven days. And let's get started on day one. Just seeing how that goes. Might have sort of some mixed responses. Some days might be more effective than others, but it'll develop your intuition about how AI can be used to support your practice in your learning.

Rob D. Willis: Where can people go to find out more about you, Georgie.

Georgie Peake: So you can find out about my business at rehearsal ai and then I'm on Georgie on.

Rob D. Willis: Georgie, thank you so much for coming on.

Georgie Peake: Thank you so much. It's been such a pleasure. Thank you so. [00:30:00] [00:31:00]