Subscribe
Share
Share
Embed
Hosted by the Ivey Academy at Ivey Business School, Learning in Action explores current topics in leadership and organizations. In this podcasting series, we invite our world-class faculty and a variety of industry experts to deliver insights from the latest research in leadership, examine areas of disruption and growth, and discuss how leaders can shape their organizations for success.
To learn more about the Ivey Academy and the services we offer, visit us at IveyAcademy.com
[MUSIC PLAYING]
SEAN ACKLIN GRANT: Welcome to The Ivey Academy Presents: Leadership in Practice, where we discuss critical issues in business, unpack new research, and talk to industry leaders about the latest trends. The Ivey Academy and Ivey Business School are located on the traditional lands of the Anishinabek, Haudenosaunee, Lunaapéewak, and Chonnonton nations. This land continues to be home to diverse Indigenous peoples whom we recognize as contemporary stewards of the land and vital contributors of our society.
Over the past few months, ChatGPT has put a name to the business disruption we all suspected would come along with artificial intelligence. Seemingly overnight, it's given professionals in knowledge industries a highly valuable tool to accelerate and streamline their work. But it also raised more questions than it can answer, what tasks could potentially be automated and in which industries? How will it change the role of humans in organizations, and how quickly will that change come?
In this episode of Leadership in Practice, we are joined by Fredrik Odegaard, Assistant Professor of Management Science at Ivey Business School. We explore the impact of ChatGPT, discussing how it's being used in business, and what the future of AI-assisted work is likely to bring. This episode is hosted by Bryan Benjamin, Executive Director of the Ivey Academy.
BRYAN BENJAMIN: So Fredrik, it's great to have you here. I must admit, I'm going to learn something, I guarantee, through our conversation here today. There's a lot of fascination right now around ChatGPT. I want to start out with, what is it really, and how does it differ from previous AI applications?
FREDRIK ODEGAARD: Yeah. So I think if you go to just the interface right now, it looks so, what I would call, unimpressive compared to what the buzz is about. So it sort of looks like a boring Google version because of both the colors and stuff like that. So most of us have-- I think we're anchored on what Google looks like. And you go in-- and Google is amazing. You type in anything and it'll just give you this fact, for instance. But then Google has been able to do other sort of applications like finding the best route and so on if you want to drive, and so Google has all these cool applications that you can find.
Now, the difference is it has this creative aspect of it. So it's not looking up a fact on the internet. So it's not like you're going to say, OK, what's the highest mountain in the world? Which, I mean, it will answer that as well, but that's not the main point of it. The main point is that it can create new content that doesn't exist. So we can create a response to any question that you might have.
So you might have trivial questions, such as, come up with a slogan for my company. I'm selling shoes, come up with a good slogan for my shoes, and it will creatively come up with like 10 different ideas for you, for instance. And so the main point is that it generates creatively new content, it doesn't look up existing facts on the internet. I think that would be the big distinction right now between what it is.
BRYAN BENJAMIN: So for the layperson, how does it do that? Like, this feels like a huge step from where we were.
FREDRIK ODEGAARD: I would agree. I would agree. This is definitely, what I call, an industrial revolution. We are having an order of magnitude improvement over what previous things we're doing. I mean, you can call it whatever you want. Like, AI 1.0 was the old versus this is AI 2.0. This idea that it can creatively come up with new responses.
Now, that movement has been gradual, right? So we're already a little bit used to if you type in an email or a letter that a lot the software will guess what you are going to write. So you end, for instance, an email with best, then it'll come, best wishes of, and then it will fill in, and you can just hit Tab and it will auto populate that. So we've been a little bit moving in that, but this one is completely new.
So this one has that it has been trained on existing data, and I think up to 2021, I think it was, or it might've been trained more. So it has scanned like Shakespeares, and documents, and Twitters, and blogs, and everything, and come up with, oh, this is how people interact. This is how people respond to certain things. And then now when you ask it a question like, come up with a slogan for my shoe shop, it will then say, OK, well, I have some inspiration from how that has been done before, and then it will, based on this algorithm, give you that response.
But it is truly amazing that it's creatively being able to do that rather than just a lookup table. So based on the old models, old AI, or old software, it's a lookup table. It will just look up, OK, what should I respond? Here, let me search it up, and then give you that response. This one is able to, as a human, synthesize different ideas and come up with a new and creative response.
BRYAN BENJAMIN: That is amazing.
FREDRIK ODEGAARD: Yeah. If I may just-- so there's algorithms underlying this. So it's all algorithmic of how it actually does it, but for this conversation, we'll keep it at sort of a nontechnical level of what those specific mathematical algorithms look like. And just like at a high level, there was existing information out there, as I said, Shakespeare, and blogs, and Twitters, and it can then just mimic its understanding of how that has happened in order to respond to a query or a question you might have for it.
BRYAN BENJAMIN: Perfect. I appreciate that. So everyone's talking about it. I'm really curious to hear your perspective on implications for the workplace and organizations. Like, how is it showing up already?
FREDRIK ODEGAARD: Exactly. I think this is the fascinating part, and I think it's sort of got released without any big announcement, and it all just exploded and took off. And I think we would be surprised how many actually use it at sort of just on the back end, like, hidden away and not sort of confessing how much they're actually using it. So what it can do is it can-- from the simple things, it can create a letter for you, an email response if you got and you need to come up with something creative to say to a client, or some employee that you have, or your boss, and it can generate a creative response to.
BRYAN BENJAMIN: Amazing. So I'm wondering, if it's doing this and we probably are seeing it used more than people are willing to admit, what potential limitations are we going to be facing?
FREDRIK ODEGAARD: At present, there is this concept that it has a hard time to distinguish between fact and fiction, and so they call this-- it hallucinates. It makes up responses, so it'll make up sources. So you might ask, and hey, can you give me a response to this question from a client, and I want some substantiated papers or references that support this argument. And what they now do, this ChatGPT will just invent sources. So we call this a hallucination rate, and I think they talk that it's around-- that it will make up 15% to 20% of it.
So those are some of the things you have to keep in mind that suppose it generates something and it says, oh, Professor Huff and Puff from a certain University has already argued the point here A, B, and C, well, that professor probably didn't exist, and so on. So you have to have some knowledge about, OK, in this domain that I have business expertise or area expertise, this doesn't exist. This is made up, but oh, this is actually a legit source that it's responding with.
BRYAN BENJAMIN: Fair enough. You mentioned earlier sort of about the training that had gone on to get it to the point where it was ready for release. Is there ongoing training happening now?
FREDRIK ODEGAARD: There's new versions coming out, I'm not going to say all the time, but there are generations of ChatGPT. And I think currently we're at number 3, and we just keep getting better and better. And I think one of the sort of ironic part is that there's a lot of people who are critical of ChatGPT. For instance, this thing about the hallucination rate, that it makes up stories and sources. And so they say, oh, it can't do this and this. And I'm thinking, give it a break. I mean, this is in a third generation. This is an amazing piece of technology, and it's just going to get better and better. And when I look at other people make up things even though they might not believe it-- like, we have this whole thing about fake news and stuff like that.
I think that those versions are just going to get better and better, and we're going to get improvement on stuff like the hallucination rates and so on. And I would say that's even a further testament to how close we're able to mimic human behavior, right? And that's the part that people are so astonished about including myself, that the response is so human-like. And we're thinking, wow, it doesn't come up as just a pure fact that a computer spitted out, that it obvious seems to be some sort of thought process behind it, and it's all algorithmic, and it's quite remarkable.
[MUSIC PLAYING]
BRYAN BENJAMIN: Your comment around reference letters as an example. What else do you see this being able to either replicate or maybe completely automate or replace?
FREDRIK ODEGAARD: So I think that's only the limit of our imaginations. I think we can easily think of, for instance, if you're in a marketing department, that it can creatively come up with 10 slogans that you can test or come up with the different types of campaign that you wanted. If you're in HR or the a legal department, it can just automatically create new legal documents for you or new contracts.
I mean, if you have an existing-- if there's something that you do on a routine basis, like, if you're on the legal department and you're selling houses or you're setting up wills, yeah, you can have a template, and you can just fill those in. But if you have the legal department, then you're constantly having new types of contracts that you're working with. OK, this client wants this, and I have to adapt. I have to change these contract. I can't have a standard template. Hey, ChatGPT, you just plug-in, here's the primitives of this contract, so here's the deals I want to make, and it will like draft up that contract immediately.
On the technical side, we have coders. This ChatGPT can even create code for you. And so you can just tell them, hey, give me the JavaScript for an interface that does this and this, or give me the Python script that analyzes the sales in Southeast Asia for 2022. And so it can just create the code from scratch and say, OK, here you go. Here's the code for this application. So I think the applications are just our imagination, like, what it can actually do.
BRYAN BENJAMIN: Are there certain organizations or even industries or sectors that seem to be more prime for benefiting from this or early adopter?
FREDRIK ODEGAARD: I think the obvious ones are marketing because we think about creative aspect, coming up with creative things. But as I said, also, I think HR and legal and the coding. And I think about that, like, all these data scientists that we are training and the data analytics stream which is my background or-- I'm just wondering, is this going to be the shortest lived career there is? Is data scientists because we will just be able to replicate that the quickest, right?
And the answer, I hope, is no. I just think that the data scientists are not going to be using-- are not going to be the ones sitting there doing the code. They're going to come up with the ideas. OK, here's the code I need, and then I can just give it to ChatGPT, give me the code that does this, and then we have a good base, and then they can work from that. So I think this is where it's going to amplify the value added in these industries or these examples.
BRYAN BENJAMIN: Yeah. The amplification, so it's not necessarily replaced, it's freeing up time to do something else.
FREDRIK ODEGAARD: Exactly. Exactly.
BRYAN BENJAMIN: A natural concern whenever there's a new disruptive technology is job security and job elimination or substantive changes. We think about fairly recently the introduction of self-checkout and what that's done for the role of the cashier. Do people need to be concerned about this moving forward?
FREDRIK ODEGAARD: I mean, I guess, yes, and no. So I think, certainly, some jobs will be completely automated, and we won't be hiring for those positions or job activities, I should say. But I don't think-- it's not going to be that we're going to have an exorbitant amount of unemployed executives or leaders or managers out there and all these organizations are just going to be self-driving AI algorithms. I still believe that the human aspect of just society in general and business as part of that society is that we add value to the services and products that we provide.
And so in terms of job security, no, I think you're just going to have to, again, elevate your value added to the organization. So some of these tasks will be automated, but as a whole, I don't see it as being about, oh, we're going to all of a sudden decimate the workforce that we only need 25% of the workforce. No, I think we'll still be needing more and more people to do other aspects of the jobs when we provide these products and services and so on. So it's more that it will replace tasks, yes, it's not going to replace roles as these roles are-- as an employee, these are going to evolve to different things that you want to provide to the companies.
BRYAN BENJAMIN: So if we go down that same path and we think about the role of leaders and individuals within organizations that may be using this or will use this, what are some of the risks, what are some of the obligations? It's one thing for this tool to help me, it's another thing for me maybe to rely 100% on it.
FREDRIK ODEGAARD: That's right. I mean, at present, I think there's two drawbacks or two issues you need to have a big reservation about. One is the thing about the hallucination. You have to be able to audit it. You have to have enough expertise to understand, this is true or this is just completely made up. So I think that one is going to be very important. But I think the most important part as we go forward and more organizations use it is we need to come to assessment of who has the accountability.
And so if you have someone who has now drafted a response to a client and sent that off and 90% of it was generated by an AI and then 10% was just like some polishing the signature and so on by this person, who is supposed to be held accountable for this response that went out to a client or an employee or anything like that. So that's one very important implication for leaders to think about, how do we assign responsibility across these tasks now?
[MUSIC PLAYING]
BRYAN BENJAMIN: So for many, including myself, this kind of took us by storm, right? And so a lot's transpired even in a few short months, thoughts on kind of what we can expect next.
FREDRIK ODEGAARD: I think one of the things that they discuss is that at some iterations, these algorithms are going to be self-improving. So they're going to be able to just train themselves to become better. So learn on their own mistakes, learn on their own hallucinations, and so on, and it will just improve on itself. So just like a human person, as you grow and you absorb different experiences and knowledge, you have now better.
And I think that's what they say these algorithms are also going to be able to do. It's just sort of self-teach themselves, which sounds almost like sci-fi, if you ask me. I mean, it goes so quickly. I mean, just in those two, three months that we've seen. Now it goes really quickly it seems, but before that, it was-- we had a gradual little improvement, as I said, the auto filling on Google, or the autofill in your email, or Word document, and so on.
BRYAN BENJAMIN: Yeah. That notion of it training itself, and sort of building momentum from there. So we talked a little earlier about this, and it has kind of got me back to thinking around sort of the trainings. Help me understand, how does the training work right now? Like, the human component of the training. Like, is it one person? Is it an army of people?
FREDRIK ODEGAARD: On the one side-- so these organizations, they have, of course, a team of data scientists who then tweak these algorithms. And the algorithms are then that they will read, take in input, and the input is everything that's basically on the internet. So everything from, as we said, the Shakespeare's sonnets to code. So it takes that as an input, and then there is these algorithms, these mathematical algorithms which can then encode all that information until what would be an appropriate response, what would be the best response to this type of questions.
And so this learning is just that as more input comes in, as it picks up different types of discussions online, or different types of code provided and so on, it can then say, oh, OK, here's a new question that I got asked and now I will respond with this one. So it's very similar to how humans work, it's just from experience. Just as more input comes into you, and you sometimes make mistakes, sometimes you don't, and then you move on from there. And that's what these algorithms can do is, like, figure out, OK, this was a good response. This was a bad response. Let's move in this direction.
BRYAN BENJAMIN: Fascinating. So leaders today have so much coming at them from all directions. The reality is this is here, and it's here to stay, and it's going to only evolve. How do leaders need to evolve themselves in terms of their thinking and appreciation? And I'm thinking of the leaders, which are probably the vast majority that are maybe a little apprehensive of a tool like this and don't understand how it works. What's your advice for them?
FREDRIK ODEGAARD: Well, first of all, I think, yeah, you definitely-- don't try to fight it because that's going to be an end battle you won't win. You do need to in some ways embrace it. And I think just start by experimenting with it, and then let your own creativity figure out, what are the value-added activities I do in a daily basis in my work, and how can I now elevate that if I can sort of outsource writing these letters, coming up with these slogans, writing this piece of code, or whatever?
And so as leaders, again, there are certain activities you do in a daily basis that are probably what we could call routine and that you don't really-- would be great if you could sort of free up your time and then you can focus more on the value-added side, and how is it that this company can thrive, and what's the competitive advantage of my organization versus the market, and can I even try and enhance those. So that would be sort of my broad general, but, I mean, each-- it will be context dependent, and I don't think it's-- it's not one right answer for each leader, so.
BRYAN BENJAMIN: So how could this technology drive changes in organizations? I'm thinking from a structure standpoint, how organizations sort of organize themselves, talent development, training, leadership.
FREDRIK ODEGAARD: Yeah. I think in one sense that I see it now is that I think it's going to make organizations a lot more flat. So if you have silos of different departments right now, I think what this ChatGPT and AI algorithms can do is, for instance, let an accounting department come up with marketing slogans, or, let an accounting department do some data analytics on its own, rather than having a data analytics department team come in and do that.
And I think what you will be able to see is each department will be able to do more cross-departmental activities and not necessarily have to be exactly in your little silo. That you will be much more fluent across different things just enabled by the fact that you can ask these AI machines to take over particular tasks, such as marketing, legal contract, coding, operational-- ordering inventory, and so on. I see it as that it will enable all the departments to lean on and use the other areas expertise within their own department. That would be one application-- or implication I see of ChatGPT.
BRYAN BENJAMIN: Interesting. So that is a big structural implication. And I even think about talent development in terms of sometimes talent thinking about, well, I'm part of the marketing department, or the sales department. Probably needing to think, well, I'm part of the organization, and we work more closely together.
[MUSIC PLAYING]
Something we've been trying to do for years and years, and probably always will do is, what are going to be the in-demand skills? Like, where are things going? So if we think about AI and as organizations look to integrate AI like ChatGPT a little bit more into their operations, are there certain skills that you see increasing in demand as a result?
FREDRIK ODEGAARD: Well, I think basically, for each type of functional department, whether it's accounting, or marketing, or operations, or finance, or legal, or HR, or my favorite one, analytics, of course, I think we're just going to-- everybody is going to elevate their understanding and their knowledge and their value added in that specific field. So I don't think that it will remove sort of the need to do these routine tasks, and each organization will instead be able to just enhance, what is it that we in the marketing department add value to, and so on. I think that's going to be, in my sense, the most biggest benefit of this innovation here.
BRYAN BENJAMIN: I really like that sort of perspective on it. It's not necessarily deep skills residing with the select few people, it's actually capability that could rest with everyone, to some extent, in the organization.
FREDRIK ODEGAARD: Exactly. And I think it's not just like-- I mean, there's a big push, of course, in my favorite the analytics, and coding, and so on, but no, I think even having a-- if you're in the marketing, to understand more how does the markets work, how does our clients respond to our products, what does the competitive landscape look like in terms of what products and services we're launching. If we're in the legal department, OK, what are the legal ramifications of these new contracts that I have to do. And so you focus on your expertise, and this is a tool to help you add more value in that expertise, so.
BRYAN BENJAMIN: That's a great way of thinking about it, tasks versus roles. And so maybe my role is made up of 10 tasks today, three tasks get to sort of come off because AI can support them, and it gives me an opportunity to take on three maybe higher value added.
FREDRIK ODEGAARD: Yeah. And probably not 3, probably 5, or maybe 10 more. And that's the thing, that it will open up that window of what is it-- what's the value added that I provide to this organization as a business, so.
BRYAN BENJAMIN: All right. Bring it on.
SEAN ACKLIN GRANT: Thank you for tuning in to Leadership in Practice. We'd like to thank our guest, Fredrik Odegaard. Leadership in Practice is produced by Melissa Welsh, Joanna Shepherd, and me, Sean Acklin Grant. Editing and audio mix by Carol Eugene Park. If you like this episode, make sure to subscribe. You can also find more information by visiting iveyacademy.com, or follow us on social media at Ivey Academy for more content, upcoming events, and programs. We hope you'll join us again soon.
[MUSIC PLAYING]