How AI is Evolving from Assistant to Co-Worker

0:00Any leader, any organization must think about reskilling and upskilling workers to adapt to this new future. That means reimagining roles. We need to get started now. It's not too late, but get started now at reimagining what that future organization is going to look like, because the next three to five years, put your seatbelts on. It's going to be a bumpy ride. Welcome to the Talking AI Podcast, where we talk AI with both experts in the field and early adopters. I'm your host, Matt Page, and we're here to demystify AI for you so you can get some value from it. Let's talk some AI. We're doing something special today.

0:35We have not one, but two guests on the pod today. To start, we're joined by Patrick Lynch, PhD, AI faculty lead at Holt International Business School and TEDx keynote speaker. We will drop a link to that in the show notes. It's really good. Recommend checking it out. And if that wasn't enough, I also have Omar Shanti, Patrick's AI CTO, to join us in the chat as we talk some AI. Welcome to the show, Patrick and Omar. Thanks. Great to be here, Matt. Thanks, Lois. Great to share a platform with you both. This is going to be fun. We got multiple guests. First time we're doing it on the pod. I'm excited for this.

1:12And Patrick, you're contributing to a book on AI. The chapter you're talking about, Bridging Culture, is how human AI collaboration is redefining teams. And some really interesting stuff in there. You hit on some thought-provoking, kind of mind-bending concepts related to AI becoming not just a tool, but a teammate. And that's evolving into this new era of human AI collaboration. And you even go so far to say we're in this rise of a new AI industrial revolution. And maybe we start there. What leads you to believe what we're seeing right now is on the level of an industrial revolution? Well, those are big, big concepts.

1:51And I'm certainly not the first to point out that we're in a new industrial revolution era. I think that we'd been on the march here for Industry 4.0, really since the burgeoning of the internet starting. But we've clearly taken an inflection point now, building on the past almost two decades of big data analytics and understanding where it is that information into this new era where tools are just maturing. I think we can reflect on some examples in the past where these things weren't really working very well for us, but we are on the verge of seeing a future where we're going to be able to interact with them pretty much like a teammate.

2:32And I think that that poses new questions to leaders and managers on what this shape of the organization is going to take now that these tools have certainly more capability and depth than ever before. You mentioned some questions, some important questions we'll be asking. What is top of mind, do you think, Patrick? And then Omar, I'd be curious to get your take after. Well, I think we're just on this arc of seeing the way that these tools can help leaders make better decisions. Certainly the routine kinds of stuff, the mechanical and the simple analytical abilities of these tools have been clear and apparent.

3:10We're already seeing how we can rely on them to help us make sense of data, for example, in our databases. And I think that that's pretty much old news for those of us that have been in the industry. But this new idea that the tools will have almost an intuitive or empathetic kinds of abilities that we're about to get to. When we think about, I don't know, there's a person out there that's been happy with any kind of a phone decision tree that they've been able to navigate before getting to a real human and customer service. We're seeing firms are going to be able to adapt now these tools into automating those kinds of tasks.

3:47And I think we're all going to be better for that. And that means that there's going to be new roles for the human workers to really engage into more strategic, creative kinds of contributions. And I think that's what marks that we're in kind of a new phase for the way that the tools interact in the kinds of industrial revolution impact that it's going to have. Yeah, you mentioned the decision tree concept of where we're evolving from. We had a past guest who did conversational design, and she was talking just exactly about that. In the past, you had to define every potential edge case option that may happen, and it's completely shifted how they approach how they work.

4:30If anybody's interested, that's last season with Amber Pross, but a really interesting chat there. Quick break in the pod. If you're listening to this podcast, chances are you've been thinking about how to actually use AI inside your business. And that's exactly why we built the AI Opportunity Finder. It's a free tool that helps you uncover high impact, tailored AI use cases based on your business, your goals, your pain points, and your industry. No fluff, no generic use cases, just real ideas that fit your business and the ranked by ROI potential. It takes about three minutes to run and it's like having your own personal AI strategist for free.

5:06If you want to try it for free, check out the link in the show notes or go to hatchworks.com backslash AI dash opportunity dash finder. Omar, from like a CTO perspective, like what, what questions are on your mind as it relates to this shift? How should folks be thinking about this? I come from the perspective that generative AI in the, in the realm of sort of companion to developer provides a set of tools akin to what something like grammar detection and spellcheck detection did. all those many years ago. And gradually, those are going to become so ingrained in our day-to-day lives that we eventually stop kind of recognizing the fact that they were novel.

5:45They will just be the sort of calculator equivalent to the middle school math class, only to developers building projects that potentially have multiple millions of dollars in impact. Some of the questions that I've got are, what is exactly the right formula? Because the formulas will change, I suppose. The question is more accurately, what are the right parameters to think of when unlocking this companion's value? A lot of devs sort of have this cultural artifact of this rubber ducky. You put it on your desk and rubber ducky debugging is just talking to this rubber ducky, explaining your problems.

6:18And then in the process of explaining it, your rubber ducky will sort of solve it for you. At least it'll show you that by virtue of explaining this to the rubber ducky, you've been able to find the problem yourself. But what if the rubber ducky could talk back? And I think that's what we're seeing now. We're seeing this advent of a embedded generative AI model that is sort of present where you're coding, present where you're designing user stories, present on the edge. And that edge deployment, similar to the rubber ducky at your desk, just allows you to solve problems a lot quicker. So figuring out the right parameters, one of them being location and availability, another being multimodal input, another being situational awareness, figuring these out allows us to build solutions that are most effective in enabling our devs and our SDLC lifecycle members to contribute to the software in a differentiated way.

7:10So, Patrick, have you heard of the rubber ducky? I've never heard of that. It's such a cool concept. I think that calls for a whole other chapter. But what Omar and I have been discussing about in this rubber ducky, I think it's a great example. In this, imagining that scenario, the rubber ducky starts to be personified naturally by workers. The rubber ducky now becomes so much more than a tool. It is in fact a teammate. I think that's what we're seeing is what are the possibilities then when it does move beyond something that we saw in the past, like a simple calculator, to something that we're really collaborating with.

7:51That's a good example because the rubber ducky is talking back. And we just have a natural tendency then to imbue expectations for what the rubber ducky can do for us, if we can imagine that. And I think that that's going to spark new creativity and new options for workers across the board. So I have this vision in my head of, I don't know if you're familiar with people that have Jeeps. If they like the Jeep, they give them a rubber ducky. So I'm thinking of like the split personalities of all the rubber duckies talking to you at once. Maybe it evolves to that where you have, you know, we talk about agents personifying the agents there with us.

8:27They have a specific task and whatnot. But I want to get into in your chapter, you talk about the four faces of AI and artificial intelligence and the shift to kind of this, you know, era of AI co-workers. Patrick, can you take us through these? Because I think it's a really kind of interesting way to think about and approach AI as we start to work with it on a more collaborative plane. I think this can be helpful for any manager out there or leader just trying to decipher this because it's a lot to take in. But researchers, not my own work, but this is Wong and Rust as far back as it leads since 2018.

9:08they outlined four faces of AI, or what I call, Omar and I call four faces. Mechanical, these are going to be the routine kinds of things that we see AI taking over. Analytical, it can make sense of patterns. Intuitive, it can start to project and do some kind of predictive analytics for us. And then empathetic is being the highest level. Those four faces of AI are a progression of maturity that we should expect and can start to see in some of these tools. and as they move up into that higher echelon, that's really where the opportunity to start to see them more as teammates and collaborators take shape.

9:46Nice. Okay, I want to try something because I was looking at the empathetic intelligence one. Have you heard of the tool Hume AI, either of you? It looks like Patrick has. So for the company that creating the pin, for example, that a wearable has been, well, it's certainly ahead of its time. Yeah. Well, Hume, so I believe this one's separate from the pin. I'm going to try to show this off in the, well, hopefully the audio comes through. If it doesn't, we'll just cut this out and it will be forever forgotten. But they based an LLM where you can actually talk to it and it registers your emotions and comes back with the inflection relative to how you're talking to it.

10:34So let's give this a try. This may be fun. There we go. Share sound. So this should work. Okay. So for those on YouTube, you can actually see this, but Hume AI, foundational voice models for any interface. And we're going to talk to a personality. Any preference. We have Aurora, Kara, Whimsy, Patrick, who do you want? Stella? Stella sounds great. Let's do Stella. Okay, we're going to start a call. Hey, how's it going? I'm doing marvelously well. Thank you for asking. How may I be of service to you? Yeah, I'm kind of excited. I'm going to the beach this weekend with some friends and just excited to get down there.

11:17The beach! What a delightful plan. Remember, it's not just about the destination, but the company you keep. Do try to savor each moment. It's a rare and precious gift. Yeah, it really is. I'm a little bit sad about it because I usually go to the beach with my dog and play with him, but he recently passed away. Oh, I'm so sorry to hear about your loss. Losing a pet is never easy and the beach can certainly bring back memories. Perhaps this weekend can be a time for reflection and joy too. May you find comfort in the waves and peace in your heart. Okay. So I just muted it. So she's not listening anymore, but the thing that got me with this and what triggered from what you were talking about, the kind of emotional side, those on YouTube can see it's actually picking up in the beginning that I had excitement.

12:08I had interest when I was talking about going to the beach, but then when I talked about it, my dog is alive. My dog's sitting right over there. So no issue there. But then it got this notion of kind of sympathy and you heard the tone changing in the voice, which is pretty compelling. And you talk about kind of giving it human personification, this is leading to some of that. But curious Yael's reaction to this tool. I think it's really fascinating the level to which sentiment analysis has sort of gotten to. Historically, sentiment analysis has sort of been a classification sort of problem where given a utterance or a series of utterance, we would classify that with some degree of probability.

12:53but now we're using that as sort of the basic building block because with an app like Hume what we're seeing is not only are we just classifying but then we are building on top of and when we talk about building a companion building AI imagine that that AI is able to sort of identify your attitude when it comes to risk or the stress level that you have and then offer solutions that vary based on that so I'm sure we've all been in a situation where we've got a friend, we've got a plan, that plan might fall through. We say, well, how about we try something adventurous? Depending on the mood of your friend, they might say yes, or they might say, no, let's just stay at home.

13:33Let's stick to something basic. So imagine the equivalent of a AI program saying, you know, there's a library for that. We can just use an off the shelf solution. Or actually, I think if you were to build this kind of AI, there's a gap in the marketplace, and that maybe there's a way to sort of productize this or contribute it to open source or so on. So I think being able to identify an individual's emotional state is a core part of being a companion and can help sort of unlock better outcomes by knowing what to ask without overwhelming. Yeah, that's great. And that's why we got the CTO on here for that nuanced perspective.

14:09But Patrick, I'm curious, have you seen any solutions like this? And where does this start to weave in? You talk about this empathetic intelligence. it starts to kind of imbue some of those characteristics to a certain extent. I mean, the low-hanging fruit, it would be customer service applications that take advantage of this. And I think that's a good example. However, however much further that needs to go, it's way better than the press one, if this, press two, if that on decision trees. And already financial service companies have been taking advantage of this. even in earlier days. Companies have been able to effectively adopt these kinds of solutions, which have been getting much better reviews by their own customers, and elevate the human customer service agents to really deal with more strategic kinds of difficult problems.

14:57But you can imagine applications like these really helping customers to get to a solution quicker. I'm also familiar with some of these applications that are, in fact, helping with therapy. It's not appropriate for maybe all, but many get benefits from this. And we already know that companion bots have been one of the most developed apps as these kinds of customized bots have become available in the marketplace. So there's a real need for this. And I think that it just shows you they're only going to get better from here. Sorry, Matt, but maybe we could take that as an opportunity to shine a quick light on an individual end user use case where this kind of tool could offer differentiated value.

15:39Take the persona of a quick service restaurant, right, with the drive-through. Very common use case these days for using generative AI or using, you know, natural language processing in general. A lot of these kind of quick service restaurants are kind of characterized by some of the same pain points, which are an employee workforce, which is not gotten the most out of. They're sort of performing the same tasks day in, day out, and potential large lines, potential error mismatches and errors in sort of border fulfillment and so on. So it seems to be an area that has been rife for digital augmentation.

16:16And a lot of people have been leading with that hypothesis. But I use the word digital augmentation there because at no point really will the AI, at least in the foreseeable future, be able to fully accommodate all of the sensitive requests that a user might have. We've already seen AIs be able to adopt multilinguality, which is, you know, a really fantastic thing. But if it's something like a voice-based drive-through, then how are we supposed to accommodate individuals who have a hard time hearing or a hard time speaking? So being able to sort of identify based on sentiment, the sort of use cases that are sensitive and that require that catered differentiated touch can really help unlock repeated value for clients who are adopting these in their end user experiences can turn a potentially painful experience into a differentiated memorable one.

17:03And that can really create champions within the clientele. Yeah, while we're on the topic of use cases, another one I like, so I have an eight-year-old daughter struggling a little bit with math and I'm trying to kind of help her along. I feel like there's so much opportunity there where you can kind of have somebody that knows what level you're at to the point where just demonstrating can be kind of sympathetic, knowing when you're struggling, knowing when you're excelling, kind of understanding what drives you. I think that's another real, really cool opportunity that there's a lot of potential for.

17:39Patrick, any others off the top of mind for you? Well, I think that you hit on it. I think education is a huge opportunity for these tools. There are already companies that are effectively integrating this into adaptive learning. Look, we've been stuck in old models, whether we're talking about educating students or even training our workforces, which is a one-size-fits-all, here's the curriculum, these are the learning objectives, and it's very canned. By integrating these kinds of solutions, and maybe even compelling with these kinds of voices and intonation and sense of change, we may have adaptive learning that can speed the way that individuals learn the kinds of things that they need to perform even better.

18:18And I think this is really a glasses half full kind of a story where we're going to be able to get better performance, both out of our systems and out of our employees or out of our students learning through these kinds of applications. So I'm definitely bullish on that future. Yeah, that's right in your wheelhouse on the education side of things. And you mentioned the example earlier with kind of the therapy aspect of it and not to keep episode dropping, but we had Catherine Von John, previous chief strategy officer of Salesforce Innovation. she started a new company called tough day. And it basically is that it's, if somebody needs to talk to somebody, they're looking to leverage AI.

19:00The interesting thing they found though, is people are opening up more to the AI, having longer conversations, more in-depth conversations than with a human. I thought that was really interesting. Just kind of a nugget of information. You know, I don't know if it's just the aspect, it's not a real human, so you can open up more of what it is, but it was an interesting nugget they found. I think what's interesting about that, Matt, is people should understand. And I understand there's some skepticism out there about how real or effective these systems can be. But we are hardwired as humans to effectively interact with these machines as other social beings.

19:39And I mean, you don't have to look very far into science fiction for what that might look like. You know, I keep telling my own kids to go back and look at the movie Her. and I understand that it can get far-fetched because obviously you're trying to be entertained in that kind of format, but you can see the glimpses of this. And the truth is we don't need something even that elaborate. Tamagotchi toys is one of the things that we talk about in our chapter. And people have just a real emotive effect, even with simple kinds of digital interfaces. So you can only imagine what's going to happen as these interfaces become more sophisticated and more natural for us to be interacting with, it will bring about emotions for ourselves.

20:22So just natural ways of having a social reciprocity between yourself and these entities. I'm not trying to be, you know, I'm not trying to go over the edge here. I'm not saying that they even need to be sentient, but it's just a natural human response. It really is more a reflection of what it does for us. Yeah, and so I think the next logical, spot. We're kind of painting this rosy future of all the great stuff. But actually getting to that point where you have like AI as co-workers and they're fulfilling that function, it's going to be a progression. It's going to have to evolve a bit. I thought one interesting thing back to the chapter you're contributing with, you introduced this idea of the AI quadrant.

21:09So orchestrating human AI collaboration. And so just to visualize this for folks listening, you have the Y axis, which is measuring the AI's charisma. You have the X axis, which is measuring AI capability. And I'm going to go from bottom left and around. So Castaway is kind of low charisma, low capability. The Charmer is high charisma, low capability. The Champion is high in both. and then the cog is low charisma, high capability. But take us through this way of thinking. And do you see it as an evolution of going from castaway to champion? Does each kind of serve a role? What does that look like in the future organization?

21:51We call this the AI quartet. We understand the kinds of intelligences that AI has from the work that was done previously. But I think more attention needs to be put on how we're going to be perceiving them as teammates. And so this is a concept that we've developed to think about these different quadrants as ways that people will be interacting with AI. Unfortunately, right now, most of the tools that we have have been castaways. They don't have very good capabilities, nor are we really relating to them. And so they become often a sense of frustration. We've had high expectations of what these things can do, but it doesn't take very long to see that we've pressed the limits and we're asking a question that it really can't help us with.

22:34That goes on to even the Charmer. It's like putting lipstick on a pig. We have this expectation that it could do more, but in fact, the capability is really low. Most voice assistants right now lead to this where people want to smash the device because it's polite, but it's not really giving me an answer. I think the idea is that we're going to see a lot more COGS and eventually Charmers. So yes, it's a progression. I think by design or by capability that we have these champions that are truly collaborative. I don't know if we have a lot of actual examples of that today, but I think that we're getting there.

23:13I think we're going to see a lot more COGS, some greater capability. And then I think that they'll be imbued with these kinds of features, maybe to our detriment. Sometimes we might think that the tool can do more than it does when it solves that problem for us with the monthly numbers. So it's not a one size fits all. I think it's more a model that talks about the way that we might perceive these tools. And there's a caution to managers that whether or not they believe it's the case today, very soon workers will start to be aligning with these agents or bots and related to them as teammates. but you can't really get away from a teammate that's so entwined into the operation.

23:54I mean, I think that there's both benefits and risks as these tools take on one of these personas as we discuss them. Omar, what's your thought here? Personifying the AI in this kind of quartet manner. How do you think that jives? And also, where do you think we are? Have you seen anything that's kind of fitting into one of those champion type categories or what areas do you think we have right now? When Patrick was first explaining to me this idea of the quartet, I immediately thought of the y-axis, which sort of represents charisma as representing a form of appearance and the x-axis of capability as representing a form of substance.

24:39and in a way that's an oversimplification because there is substance in charisma it's not just a layer of paint on a car with no engine but to actually have a really aerodynamic body and a car that's optimized for you know performing well on the turn and so on is a feature of engineering so even if we do fall back on this appearance versus substance mechanism I think it would be wrong to dismiss the effective value, effective within A, the emotional value that the appearance will have on the end users. And I think what Patrick is doing is really centering in that personification and the associated psychology of personification that will sort of be a differentiator for the space moving forward.

25:22Where my mind goes is, I wonder whether the go-to-market of a lot of companies has fallen more on prioritizing and privileging the side of charisma than capability for a number of reasons. It's easier to go viral as a charismatic substance than to go viral as a substance that maybe outperforms on the latest benchmarks or that, you know, is able to boast a Turing complete operating, you know, framework or any other sort of technical feat. It's easier to captivate the world's imagination by telling fairy tales as a pirate than it is to, you know, code a very efficient Fibonacci algorithm that optimizes for memory, for example.

26:03And what I think we will gradually see, and correct me if I'm wrong, Patrick, but my sort of perception is the over-representation that we currently see in Charmer will be met by an equivalent sort of move to the right in a drive towards champions as people look to develop more of that AI capability. And so we'll end up with a landscape where Cog would be the least represented. Charmer, slightly more. Champion is where everyone aspires to get to, but likely won't get there. And Castaway is where everyone sort of ends up trying to achieve that sort of arc. I might be, and I'll call myself out for this, I might be underestimating the difficulty in sort of managing the charismatic side and encoding charisma into a bot.

26:54So please keep me honest there, Patrick. and I'll also kick it back to you Patrick and ask you know how you see charisma as a cultural phenomenon based on the dev culture that you're part of your demographic even your subculture within the broader society how do we problematize charisma we've already seen a variety of models most famously X's Grok model being charismatic in being a rogue I believe that's like how Elon Elon Musk describes it. So I suppose I'd be curious, do you see charisma as a pluralistic idea or is there sort of a standard set of charismatic features or a single charismatic persona that we are driving to in the market?