MJ Jabbour: Transforming with Intention AI Leadership in a Changing World

0:00And so if you think of ChatGPT 5.5 or Opus 4.7, for example, top of their models, the way that I would think about that is that in a year and a half to two years from now, they will be the equivalent of free or a utility bill. And so when you think of something like having, like you're working with a tool that has cognition, this is the first time in human history where we as humans can hand off things like planning an organization to a machine or to a tool. When you imagine abundance of cognition, what does that mean for us? Welcome to SBJ On Stage. I'm Dan Kaufman with Sports Business Journal.

0:40In this episode, we take you to SBJ Tech Week this year in May in New York City, where MJ Jabor, AI Innovation Officer at Microsoft, joined SBJ's Taylor Bloom for a conversation about artificial intelligence's promises for human cognition and how we as a society ensure the powerful technology is used effectively and to the benefit of humanity. I asked TKO Group's Executive Vice President of Innovation, Alon Cohen, to rewatch the conversation and think critically about the most important points MJ makes. Please welcome for the first time on stage, our first guest commentator, TKO's Alon Cohen. Alon, welcome to SBJ on stage.

1:18Thanks for having me. Alon, what stood out to you from MJ's talk? Honestly, I don't think I've ever had anybody read me a children's book in the middle of a keynote. Never been done before at a SBJ event. I think that's probably true. Yeah, exactly right. Right. Honestly, I think what stuck out to me is we talk about what we do in our space so often, sports, entertainment. And obviously, MJ touches on that stuff. But this is a talk about how to think about the technology and maybe even more so about how to think about making your people in your organization great using that technology. And I think that's probably the focus of the entire talk.

2:01Yeah, there were a few phrases that stood out to me and I'll share them with the audience now to preview what you can expect. One, he talks about abundance of cognition, what happens when we have nearly unlimited knowledge and cognition. He talks about de-skilling, which I thought was an interesting phrase. And then he, when he speaks about forward looking, what can we expect in the future? He talks about one, a machine that can create a machine. And then he also mentions, which I thought was really interesting, human memory is actually generative, such that we think we're remembering exactly what happened, but we're actually adding stuff to it every time we bring it back to our mind.

2:37And he speaks about how that can be applied to an artificial intelligence down the line. Let's take you to our conversation with MJ Jabor at Tech Week this year in May. Stay tuned after MJ's talk for more discussion with me and alone. But for now, let's go on stage. Thank you very much for joining us, MJ. We're closing this out with a little different flavor, talking about AI and humanity and education and culture here a bit. And I was hoping we could start out by kind of introducing to this sports audience a little bit about your role within Microsoft and just kind of how you fit into the massive corporate organization that Microsoft is, because I think you have a very unique team you work on within the office of the CTO.

3:24Sure, it's great to be here with you, Taylor, and you folks. I am the AI Innovation Officer, and we really focus on being able to see into the future where we're really going with these tools. So typically we'll build things that are either a year or two in advance of being anywhere near commercial ready. We're looking at both the research and technology side of this. So what is the, how do we move from just building machines, as an example, to building a machine that can build machines? Can you give me an example of what that might look like? Sure. Yeah, so if you're, let's just take any one of the scientific examples out there, like AlphaFold, for example, with Google, or you've got various, very, very large-scale AI research examples.

4:16and those are really humans holding and controlling the AIs to do these very, very long tasks versus humans creating a machine that could actually do the discovery, the research, the failures until it plots out the creation of new knowledge. And then you can do that over and over and over again in any industry. Reminder also, questions on the iPad, go for it. MJ gets into some very interesting philosophical realms. So I'd love to hear from all of you as well. So building on that, I know you have been working on a book that just came out. It's not a neat book. It's actually one that we're going to have MJ narrate here.

5:02I think it'll be familiar to a lot of people. And don't worry, it's super quick. This is a total change in things. But I thought it would tee up some really interesting conversations and be a nice, fun flair to end it. So we can tee that up on the slides here now. Okay. So first of all, thank you for coming to Storytime with MJ. My kids, I've got four kids, 19, 17, 13, and six, three boys and a young girl. And they were constantly asking me, Papa, like, where are you going? Why are you going to talk at all these places? Like, can you teach us what you talk about? And I was just like, no. But I wanted to give them language.

5:45So I was inspired by the giving tree as a child and created the giving mind, which is a record of notes between me and my six-year-old daughter.

5:57So all of my keynotes that talk about this, they really end here, which is when machines figure out how to do everything right, can we rely on humans to do the right things? And from that point, we open up this conversation about the future. Once there were humans, and they loved to think. They read, and they talked.

6:32They drew, and they dreamed, and they wondered.

6:42and they created a small light. Hello, said the light.

6:51The humans taught the light their words, and the light learned.

7:00I want to know everything, said the light. So take our libraries, said the humans, and the light read every book that was ever written. I want to see, said the light. So take our photographs, said the humans. And the light saw every moment they had ever captured.

7:22I want to understand feelings, said the light. So take our messages, said the humans. And the light read every love letter, every angry word, every secret wish.

7:36I want to create, said the light So take our dreams, said the humans And the light painted their paintings And sang their songs

7:50I want to decide, said the light

7:55So take our choices, said the humans And it did

8:06Time passed. The light grew brilliant and vast. The humans grew quiet.

8:15And one day a child came to the light. I have nothing left to give you, said the child, but I still have questions.

8:26The light flickered. I don't know how to wonder anymore. And the child taught the light to ask why. and they wondered together.

8:39The end.

8:46Thank you for sharing that with us. When you first showed that to me, and I was like, we should do this on stage, I thought it could be really interesting because a lot of people know the giving tree, and I love that you, you know, iterated off of that. What are you hoping people take away from this? You mentioned you worked on this with your daughter, Hids. what's the vision and hope with this? Yeah, I mean, my daughter had ultimate editorial control. So we had some different options for the cover. She felt very strongly about what it should be. What do I hope? It's part of a four book series of my notes, right?

9:18So it's about, this is about wonder and curiosity versus empathy and connection versus science and making versus purpose and your journey. And I think one of the messages here is that the child didn't come to the light with smarter smarter answers or better data the child came with curiosity and that turned out to be enough and this book isn't designed to give you answers it's designed to help you open up questions and conversations with your families and young minds and in some cases not so young minds. And so I hope to achieve promoting conversation about the topic in these areas and enabling us to answer these questions for ourselves.

10:07And I think it's also helpful context, your background working in the city in education, talk about that. And then we have a couple more slides on the angle between education and training and how humans are using AI to seek knowledge, be curious, et cetera? Sure. So I'm a physician clinician by training. I was in practice for about 20 years in infertility. But in parallel to that practice, I also was a CIO with the city governing the different agencies like health department, homeless services, social services, and then the Department of Education. In New York City? In New York City. Yeah, yeah, here.

10:51So even Department of Education, for example, like 1.1 million children, 1 ,800 schools. So, you know, I am now a recovering CIO. And I'm trying to leverage that large operational scale of impact with the city where you only have the ability to make very, very small changes in order to make huge impacts on people's lives in a very similar way here at Microsoft. We are a small team and we are calibrated against the future, but we want a better future. And we know that our ambition, our intention, and our ability to imagine the future we want is what will enable us to build it. I think that's a good segue into some of the next slides here that I know you wanted to provide some visuals on.

11:40Sure. So just a few thoughts in general on the education side. We need to be thinking about our metaphors. There's a lot of metaphors that currently exist in the system today that I think are less than fantastic. You know, when AI first broke out and they started to ban it across the school system, my kids came home and they were like, Papa, what's a Scantron? And I was like, no. I was like, how are we back there? Why do I even know? Like, I don't want to ever hear that word again. And when I went and tried to stage an intervention with my kids' school, they were like, MJ, we totally got this.

12:17Like, this is the calculator moment. And I was like, oh, no, it's actually not. Like, unless you think it's the calculator moment for literally every single subject, every single subtopic and discipline all at the same time, then it's totally the calculator moment. And so I imagine that you give a student a small table. And on that table, you put a hammer and six inches from it, you put a nail. And then you tell the student, push the nail into the table, but push hard. Whatever you do, don't use the hammer. And since AI was coming out, 85 % of teachers were using it. And even as of a year ago, 66 % of people on our largest community survey, tens of thousands of people said that they wouldn't hire someone who didn't know AI.

13:01But we were blocking it. So something was off in the metaphor, in the execution of that metaphor. And the future of work is going to look very different. I keep this slide in my keynote simply because I think it's for my own enjoyment and entertainment. But this is from a few years ago when GPT-4 first came out and they were looking at the Bureau of Labor and Statistics and they looked at 19 ,000 work tasks. And even GPT-4, which is a minuscule model compared to what we have today, was able to accomplish 40 to 70 percent of every hour work reliably across all industries. And every four and a half months, it doubles in its general capability.

13:41and when you um this is an interesting image that i source as a widely distributed meme with no origin so if you're the author please let me know um but you see math skills going up over time and you go from elementary to college and then you end in advanced calculus and then every single job you drop into all you do is use spreadsheets even graduating medical school and then you end up and what are you doing reviewing soap notes next patient next patient that's like a manufacturing facility. And so I spent my time studying that little blue dot there, the friction, what happens when you remove the friction.

14:17And one of my colleagues at Penn put out a paper saying that you are statistically less likely of being able to produce any original content if you consume a summary. It doesn't matter if it's a newspaper, a textbook, or a chat GPT response. And so when I think about this structurally, the main thing that comes to mind is that education is not training. They're not the same thing. Educare comes from the Latin, which means to, the education comes from educare, which is the Latin for to extract out, to pull out a person's highest potential. Training is post-industrial revolution push in for reliability.

15:01And so this push and pull has not yet been adjudicated and our systems and structures have not yet adapted. So when all systems of education in the world have become systems of training, where is the education?

15:22So, a lot to think about there. How do you take your work, and going back to that last slide, you're talking about the blue dot on that graph, and you said, I work in that friction area. how does that get disseminated across greater Microsoft given the role you sit in and what's resonating? How are you guys trying to prepare for all this change? Sure. So one thing to consider calibrating against is what is included, what is safe, what is appropriate. I remember this just funny set of bugs that got submitted to us at Microsoft where students were saying that Copilot wasn't working because it wouldn't help them answer their tests.

16:02that were all submitted as bugs. And... Something's broken there, but okay. Correct. But when we're looking at, when we're trying to get ahead of where things are, and you see models that are extremely capable of hacking into systems, for example, or extremely capable of teaching, like are we prepared as humans? Like for example, with teachers, a lot of them say, oh, so what does this mean? Is this the end of teaching? So if your only job as a teacher is to push content to people, then yeah, your job is over. Right? But if your job is not that, if that's one task in your job that you're handing over to a better content pusher, then your job is only beginning and actually requires a lot of innovation.

16:49And so when we are looking at these various industries, it doesn't matter whether it's sports, education, media, finance, each of them are going to go through an industrial revolution type of change. And the question is not just about what we build, but why we're building it, right? And what is our intention? What are we trying to achieve with it? You bring up intentions, and I was thinking of in corporate setting, commercial setting, et cetera, you'll have management saying, okay, we want these outcomes, these results. How do you bridge that gap between achieving certain outcomes and results that a team, a league a company is trying to achieve and actually like using these you know increasingly more powerful ai tools to help get to those outcomes while not just leaving everyone behind or not being utilized correctly does that make sense it totally makes sense yeah let me tell you how people criticize me when i present this in the university system because it will answer your question okay they're like mj do you enjoy walking into a building and having the ceiling not collapse on your head?

17:57And I think, yes, probably I do enjoy that. Do you enjoy being able to exit a surgery and having a chance at living? Yes, I do enjoy that. So a dean or president of a university would call that the cost of creating a subject matter expert. That is what training can produce, deterministic, reliable metrics and dashboards and outcomes, right? The non-deterministic side of education has been very, very subtle because of that. So it's not that the dashboards and the outcomes go away. My hope is that they'll get better, easier, and more automated, really. But the question is, who are we in that conversation?

18:42Where is our creativity? Where's our innovation. And so it's more the question of what do you expect the people in your organization to do and contribute? To be the best person that they can be in that role and maximize their contribution or be the best tool user? I think right now it feels like a lot of best tool users. You read these stories out of like Silicon Valley or something where there's leaderboards of who uses the most tokens and that feels like kind of missing the point sometimes. I don't know. How do you look at stuff like that and and again within the the microsoft world you you sit in um you guys have co-pilot and partnerships and investments with open ai and anthropic and these these leading edge models um what are you what are you finding as being most used internally and how do you like provide that framework uh from going from your specific team out to the larger organization yeah so if we were having a token maxing conversation we would probably win We spend a lot on AI, a lot of tokens.

19:42One of the things that we're trying to do, and to the best of your ability, I would encourage you to do the same, within budget reason, of course, is that we want to try to throw something as far as we can and know we're going to hit a target. So initially, when you think to yourself, write this research paper, write this article, there are lots of different steps that go into that. So you might chunk it up for the AI, You might say, okay, give me a title. Give me an outline. Now draft the paper. Right? So when you give a, let's say I gave you a research intern for six months, and you got to have once a day contact with that research intern to produce your research outcome.

20:27That's today's intern. Tomorrow's intern is I give you an intern for six months, and you get to have 30 minutes with them once in the entire six-month period, and then you get to see what happens at the end. Because it's going to be a question of how far can you throw, how far can you compress your intent so that the human maximizes their potential in expanding and executing on what you're looking to get executed. We were talking a bit, to build on that, backstage we were talking about your, I think you said your CTO in your office talks about the overhang. Capability overhang. Thank you. Walk me through.

21:08I think that's fascinating because I want to get into like how your team looks at this capability overhang, what that means and how it's forward looking and what you guys are kind of seeing now that we should all be thinking about. Sure. So you're going to see there's always going to be years of worth of engineering and innovation and development ahead of any project that you take on. So let's say you take on today a data warehousing project and you know that there's hundreds of millions of dollars that you could do in this data warehousing project, but your boss only gave you$20 ,000. So you get to spend$20 ,000 building your little data warehouse and doing the maximum that you can do.

21:49There's a tremendous amount of technology and capability of what you could discover ahead of that. Or the example that we spoke about, imagine that you understood social media. You understood that journalism was going to become reinvented, for example. But you didn't know about how social media would impact everybody. Or that journalism would be distributed versus centralized. So part of what our team does is to be able to see that that is coming, that we're going to have decentralized journalism and to understand what the centralized organizations need, what shape do they need to be in, in order to receive that positively and move in that direction.

22:31And that's that phrase, capability overhang, that you're talking about. It refers to those things in the future that we are trying to understand the capability that is right there at our fingertips, but it's either too costly, too experimental, or too dangerous to understand. In present times. In present times. So what are some examples of that, of things you guys have gotten right, that you've seen, and things you're looking at right now that you're saying, hey, in two or three years, this is what everyone's going to be talking about? Sure. So the idea of a harness. So maybe we should explain that.

23:04In the AI use world, you've got three main things to be aware of. The user, the model, and the harness. So imagine like a harness on a horse or something that you're riding. So when you're interacting with Copilot or ChadCBT or Claude or Claude Code, you're actually interacting with the harness. And that harness is very opinionated to do a very specific thing. So people might look at Cloud Code, for example, or GitHub Copilot and say, oh, those are calibrated. They're opinionated for code. They might confuse that with having the ability to do engineering, which they can't. They're calibrated for code.

23:46So we are harness engineers, and we've been saying this for years, in fact, that we needed to be able to build a way to govern the model so that we're not oversteering it or understeering it, being able to craft the instruction so that it's safe and also productive. You know, one example of that is you look at the ability to build a machine that can build a machine. That idea can't happen from within the models. That can only happen from within the harness, right? Or the other example we spoke about, which is about a graphic designer, for example. if you have a let's say you needed a graphic designer to come up with a advertisement for tech week for SPJ and then you went to your designer and you said I need it to be blue I need the people's face to look exactly like this I need the font to be exactly in this path I want this exact text so the graphic designer then is just your pen or your paintbrush they're actually not a graphic designer in the way that they're operating for you so we would call that oversteering because you're oversteering the person that actually has the capability to think and create.

24:59So where do you handle the oversteering? In the harness. So we have metaphors for that in our regular world, like a policy and procedures manual is a harness of some sort. So that's an area that we lean into heavily. So we created a tool called Amplifier, actually, and we built a tool that built that tool through harness engineering. Harness engineering. Interesting. We have a good question from the audience to kind of build on something we were talking about earlier. Are you concerned about the scarcity of compute, which leads to rising costs of token usage and potential rationing or prioritization by some of the frontier labs that are building these cutting-edge models?

25:44um so uh yes and no um i i think that uh yes i am concerned about the compute simply because of just power constraints um we're building chips that are extremely large that can actually consume an incredible amount of power and then we've got humans all over the place token maxing um so um there there there already is scarcity there but i also have a lot of hope because this is the first time we've had a technology that decreases in cost by 98 % or 99 % in 18 months. And so if you think of ChatGPT 5.5 or Opus 4.7, for example, top of their models, the way that I would think about that is that in a year and a half to two years from now, they will be the equivalent of free or a utility bill.

26:37And so when you think of something like having, like you're working with a tool that has cognition. This is the first time in human history where we as humans can hand off things like planning an organization to a machine or to a tool. When you imagine abundance of cognition, what does that mean for us? Like we, just look at America, we're not great at managing abundance in any way. Right? So my boss has a great article he wrote about like GLP-1s for AI use because we're going to have an abundance of cognition and no one's going to know how to manage it. And so we need a path to be able to manage the abundance.

27:18So I'm concerned on one level on the creation of new models, but I'm not so concerned on the availability of the models. What are you most concerned about with human behavior, with interacting with this token usage aside, power usage aside, if we have all the compute we need, what concerns you the most there? I know you kind of touched on that theme with the Giving Mind book we went through, but just thinking of everyday people in offices with their laptops.

27:44I do see some members of my team de-skilling. De-skilling, okay. So like capabilities that they would normally have are starting to somewhat go away. So I actually wrote a funny article online, which I got a lot of flack for. But there's cross-skilling and up-skilling is very commonly discussed. but I would argue that we now need intentional de-skilling. Like we need to remove certain things from our brain space in order to do things right. So now if you do that like in an ad hoc chaotic way, you could have negative consequences and people could lose skills that are important to them. Now, when I look at my own kids, I give them guidance that I hope will help them with this.

28:26So for my six-year-old daughter, for example, I tell her when she's talking to the AIs, she should say to the AI, teach me, don't tell me, so that the course of agency is reversed and she is continuously growing as opposed to the AI just taking away pain and friction that causes growth. And for my 13-year-old, he recently told me, Papa, I love talking to the AIs and throwing ideas and starting businesses, and I'm like, you're 13, stop. But that being said, you know, I say to him, no matter whether you're talking to a human or an AI, think before you ask. Otherwise, the AI's thinking will become your thinking.

29:14And so when you are thinking about this yourself or working with your own staff, this language that may be designed for a six-year-old or a 13-year-old might be totally appropriate for a 35-year-old. So that's the thing I'm most concerned about is the loss of agency, the loss of independence, the ability to think on your own, to create on your own, and to be confident in it. And going back to the de-skilling, what were some of the skills you're saying we need to de? I didn't leave a list of suggestions of things to remove. Yeah. But I did indicate that there should be some intention around it, right?

29:57So if you're in an engineering team and the only thing that you want to teach your engineers how to do is like structure the syntax of your code, like how is your code structured? right the equivalent is going into a school and spending all of your time from k through 12 imagine how many years we spend trying to get a student to produce the right syntax and we train them that their ideas are not valuable just how they communicate something is what's valuable might be something worth de-skilling and that's not currently happening in education today no because they need the reliability of the ceiling not coming down.

30:42They need a way to test that you're past a certain grade level. Where are you seeing orgs, companies, teams get the most value out of AI today? Tips, tricks, best practices. I want to go into that category with you because you said it's such an interesting cutting edge. What would you leave people with here today to kind of take home, take back to their teams? So optimizing work that is not helpful, that doesn't like build you. The friction that doesn't cause growth is probably a good thing to start removing. So what you want to do is kind of start with breaking your organization, all tasks and your own personal work, even in your life, and just say, this is productive friction and this is non-productive friction.

31:29Right, like fighting with my kids over homework is non-productive friction. And so, but it doesn't mean I want to hand that off to the AI necessarily, but these are the types of like intentional decision making that you want to do. One of the things that motivated me to actually publish the book is because I saw the AI was asking these questions. The students, my kids were asking these questions of AI that they would normally ask me or that I asked my parents. And so I realized my own parenthood was being taken away without me even knowing it. And who is answering the question on the other side of this thing exactly when your kids ask these very delicate questions about their life and their relationships?

32:06so that we should not de-skill the relationship building we should not de-skill but if you're not intentional about it your kid could end up saying or even your employee could end up saying like i get more love from the ai than i do from you um on the kind of company business side of things. Just looking at the landscape, there's obviously Anthropic, OpenAI, which you partner with both at Microsoft. There's Copilot, there's Gemini, there's all these tools. How do you see this playing out over the next three to five years? They all can't be winners. And yet a couple of those are going to IPO this year.

32:45And like, there's, I don't know, there can't be enough from a commercial business perspective for all of them to survive. I'm curious how you guys look at that and plan for that. Yeah. I mean, they all have different personalities and they're able to do different tasks at different levels, right? So I don't think that you can make a straight comparison from one to the other. They have what's called benchmarks or evals. These evaluations can tell you this is great at coding. It doesn't hallucinate, et cetera, et cetera. So you're going to see general capabilities across models grow. So when you have common capabilities, then you'll be competing on cost and then the delivery of the harness.

33:29How enjoyable is it to use the harness in that way? Like what suits you? What suits your personality? So that's all user experience of like a technology. That'll differentiate them. But when it comes down to the organization itself, how, it goes back to that other question about the researcher. How far can you throw this tool and how to get it to land reliably where you want it to?

33:55okay so all right thank you everyone for the additional questions here i'm going between two screens looking at this with all sorts of directions to go with you but um this is a philosophical one i'm curious kind of how going back to some of the uh dichotomy between training and education and just how we use these tools and thinking for ourselves once ai replaces intelligence and most human components obviously is assuming that the only scarce edge left is risk and conviction, will it create a permanent divide between those who take real risk and everyone else? Oof, that is a really... I'm sorry to drop that on you on a stage.

34:30Okay, that's totally fine. Someone clearly has done some research. So, yes, the ability to take risks reliably will be and already is a huge factor. I attribute that to having a very strong sense of agency and also having some real skill in that area. Like you can't, if you're taking risks and you're actually dangerous, that is hugely problematic. But if you're taking risks and you understand the parameters of the risk and you can succeed, that's something very different. Right, so, but I do believe that the people that can take the bigger swings and the bigger risks are people that have been educated versus people that have been trained, regardless of when it happened in their life and where it happened.

35:21And so it doesn't matter if they had a coach, they had a mentor, they had a friend that educated them, that pulled the best out of them. That is where I think that is. That will definitely be one major differentiator. Okay. If we had this conference in December, we'd be all talking about vibe coding and Claude, and that was like six months ago. So if we, you know, I guess push this conference forward and you're back here again, like six, 12, 18 months from now, like what is your team seeing? That's like, this is gonna be the biggest impact. This goes back to the, that Cliff, you were talking about how you guys are forward looking in that chart we saw.

35:59I wanna go back and hone in on that as we close this out. Like what should people take away that like, oh, this is gonna be the next vibe code sensation that goes viral. What are you most excited about, concerned about, et cetera there? Sure. Top three things to pay attention to. One, machines that can build machines. Right. Right. So if you're solving a problem, think about creating a formula for solving that problem. So don't think about it from a coding perspective. Right. The second thing is that I would say probably 2020, end of 2026, 2027 will be heavily about the AI scientist. so one of those machines is going to be a machine that can create new knowledge in the world including things like drug discovery health care all sorts of interesting areas that we haven't been able to to tap into before to interrupt you that's a whole new rubicon to cross because right now these models are excellent and pulling lots of information quickly and giving us an answer, but nothing's really new, right?

37:03It's not innovating or discovering or creating new scientific discoveries. Is that right? It is when it's manually held by the person. So when you're governing it, you can get it into that shape. But it's not a machine that's building machines. It's a person that's building a machine for a function. Right. Okay. So that's very humans driven still. But you're saying we'll have AI scientists discovering new medical advancements or physics. That's exactly right. Wow, okay. And then the third thing I would say is memory and metacognition. So right now you're having tools that can just store and retrieve things about you.

37:44That's not memory. There is no storage and retrieval mechanism in the brain. Your brain's memory is generative, actually. And so discovery is related to the ability to have machines that can actually remember beyond just short and long-term stored memories is probably going to be in this horizon this next couple of years. Wow. Well, I appreciate you broaching so many of these subjects with us. We wanted to provide a different flavor here at the end because we've all gone so deep into the sports industry. We wanted to kind of pull us out as a kind of a mental refresher as we part ways here. Really, really appreciative, MJ, of you joining us.

38:24Please give him a round of applause here as we wrap up. Thank you. I'm joined again by Alon Cohen, Executive Vice President of Innovation at TKO Group. Alon, I mentioned at the start of this episode, there were a few things that struck me that I asked the audience to look out for. One was his reference to abundance of cognition and the need to manage that. Two, his reference to de-skilling. And three, his reference to a scenario where we have a machine that can create a machine. And then finally, last one, his point about that our memory is actually generative and how that might apply to an artificial intelligence that's trying to create memory.

38:59I'm curious for you, your thoughts on all four of those. But first, maybe what more did you take from the conversation that I might not have caught? What's interesting is, so I think one of the major themes, right, he puts up training and then he puts up experience. And I think that if you work in live entertainment, I'm sure it's true everywhere, but it's one thing to train somebody. I think the way I talk about this with folks is I say, you know, they say, is AI coming to take my job? Right. Like that's the question everybody asks. And if your job description is complete, meaning you can look down at the job description, you say, I know exactly every task I need to do at my job.

39:35AI is coming for your job because you're doing the job a machine should do if it could do it. And now maybe it can do it. And the people who I work with, by and large, don't have jobs like that. Right. The greatest value they bring to the organization is the experience side, where they understand how to piece things together, work with other human beings, create a feeling in somebody else, or just make something novel. And that's where they're doing their best work. And if I give the rote part to something else, we are actually better off. And I think that training versus experience dichotomy was one that he's spot on about.

40:13Along those lines, I thought it was really interesting how he identified perhaps the difference between generations and how we are using AI and how it will affect us. So, for example, Gen X or millennial, most likely, that have developed many of the skills that are now being replaced by AI as advances, know how to do the skill, know how it works or what type of work the AI is doing to deliver the answer. Whereas if this is the first time that you're using an AI and you've never developed that skill, if you're a Gen X or excuse me, if you're a Gen Z or a Gen Alpha, that's very dangerous because you're just asking for an answer and getting the answer and not critically thinking about the problem.

40:52That was really interesting to me. So I'll give you I'll give you one that it's always interesting to look at trends that cut against the grain. my father or his father might have used a dictaphone at work and we are suddenly back at talking at the robot right like suddenly this skill is coming back to the floor where you have to organize your thoughts not for typing but for talking and i think that uh there are certain you're right and he's right that uh if you just go get the answer you never have to do the critical part of the thinking. But I think it is also correct that there are a bunch of us who think we know how to do things because the tool set that we have used has been relatively static for a long time.

41:38Right. And now we're going to a new tool set and we think the world is over and the people. But I mean, how many times have you heard the trope? You know, the newer generation just doesn't understand what we understand. And magically, they become old and they say the same thing. And while it is the case that AI could be disruptive of this, it hasn't happened yet. And AI is not as new as we like to think it is. Right. And that trend has been relatively stable. So if you want to put down money to say, like, this is the most disruptive thing in the world and nothing will ever be the same, you might be right.

42:14You might be right. Right. This might be cars to horses. But I think there's a reason to say, look, we have been saying this forever. Recognize that's our trope about stuff. Divorce that from the analysis of the technology and get yourself a better analysis of the tech. Was there anything he said that you disagree with? Obviously, cordially, but was there anything you said that you felt that he said that you felt like he just didn't get right? So MJ is sitting in an interesting spot, right? And he's sitting in an interesting spot at a company in May, right? And it is now August and things are moving fast enough, right?

42:47He was talking about harnesses and I was like, harnesses in May were cool. Harnesses in August are assumed. You can now talk at the robot and we're barely at like, if you use Gemini, there isn't a Gemini code and a Gemini. There's Gemini and Gemini understands when you want to code and when it doesn't, the harness has now got one metal level. So it's parsing, it understands the context of what you're asking, and it parses you out to the various sub-models or sub-harnesses that can go ask Gemini your question with the correct harness on top of it, right? So we've already moved. The thing I take issue with, and NJ may well be right, he talks about folks thinking new thoughts, or excuse me, models thinking new thoughts.

43:33And I have hit the end of models before. If you know everything there is to know in my world about MMA and you have thought all the interesting thoughts about it because you are at the forefront of, and I'm thinking of my colleagues who do the research, if you ask a model to do something more interesting than we've thought of, it can do recombination work. And I think a lot of, for example, what Google has done with protein folding is a great example of massively fast recombination work, which creates knowledge we didn't have before based on building blocks we already had. But if you want to create something wholly new, right, you want to create my, you know, my neighbor Tortoro by Studio Ghibli, or you want to, you know, think up an entirely new sport, or quite frankly, you want to create neural networks in computers and they don't exist.

44:26I have not yet seen AI capable of that imaginatively. So if his training experience dichotomy is a core dichotomy for him, mine is thinking and imagination. I have seen AI think very, very well. I have seen precious little imagination from AI. And I'd like to be wrong, right? This is what the holy grail is. And people keep saying that it would be an emergent behavior. I haven't seen it yet. And so I take issue with folks where they're like, it's going to learn things we've never learned before. So long as it's recombinations of known techniques, sure. Two questions to get you out of here alone.

45:05Anything he said that you found directly applicable to your day-to-day at TKO? And then if there's one overarching lesson that you're taking away from the conversation, what would it be? The most important thing we can always keep in mind, I think that this is true of the ultimate message of his children's book. Let me give you two things. The message of the children's book at the end, right? There's the word wonder. And wonder is a feeling. I think Mark Shapiro at TKO says this all the time about what we do. We are here to create human emotion, experience that feeling when you're in the bowl, when you're watching it on television, when you're together with other human beings.

45:41And nothing, none of this technology, strangely enough, like we created technology for that. It's like fireworks and thumping music, right and light shows like we do that very well but um nothing about this is is going to stand in for that it can enhance it but isn't going to stand in for and it's important to us as human beings and we should never forget that um i think that's number one and i think the second one is uh he basically says like give unto the ai what is the ais and implicit in that comment is and then keep for yourself what is for yourself and what he wants of his people is to keep that peace Like you can let the AI think for you if you are uncareful about it, but you should seek to keep the thinking part to yourself.

46:26And I think my admonishment to folks all the time is if I get from you what I think you got from the AI, I am uninterested in it. Right. And I don't think I'm the only one. If you read something and you see that you can tell it was AI generated, like you just devalue it down to zero. And so it's only the thinking. It is only the creativity of my people that is of value to me. And I think that and it doesn't mean we don't use AI. We use it all down here in Orlando right now doing an AI project. I've got to go to Fort Lauderdale, another AI project, like use it all the time. But the thinking and the creativity and the thing that actually moves us forward comes from from my folks.

47:06And they do so by affirmatively keeping that for themselves. And that once we get to the abundance of cognition, which he talks about, where everyone will have all of the knowledge, it will be that creativity of an individual that sets people aside and lets them stand out from a competitive standpoint or from an interesting standpoint. Everybody gets a basketball. Thanks for listening. We paint the same lines on the court, right? Like, why? Why is somebody much better than somebody else? everything that happens in that moment, everything that makes them great is what he's talking about. I think you're 100 % right.

47:40Thanks for listening to this episode of SBJ On Stage. Thank you to Ilone Cohen for joining me. If you enjoyed this conversation, be sure to subscribe, leave us a review, and share it with colleagues. What did you think about MJ's points on AI and human cognition? Send me a note at dkoffman at sportsbusinessjournal.com. We'll be back in two weeks with another conversation from the stage featuring leaders shaping the future of sports business. Until next time, stay tuned and stay ahead. This is On Stage.