Latest / On Scope: The Marketing Podcast / Bonus Episode: A Marketer’s Guide to Sustainable AI with Responsible AI’s Manoj Saxena
Transcript
- Speaker 1 (00:00): The views and opinions expressed on this podcast are those of the speakers and do not necessarily reflect those of the ANA or its members. Hey, hi. Hello folks. This is OnScope, The Marketing Podcast. I am your host, Mike Burberich. And on behalf of the entire OnScope family, we hope you are having the best summer ever. And while I know those back to school ads are going to start popping up on our TVs and social feeds soon, the summer ain't over until we say it's over. And in that spirit, we've got a hot bonus episode for you today featuring the one and only Manoj Saxena, the founder of the Institute for Responsible AI. Manoj and I took to LinkedIn Live to break down some of the most pressing issues surrounding this technology today from emergent behavior and what that might mean in the long term. What Manoj means when he calls it alien intelligence and vitally, the difference between AI safety and AI security. (01:18): You're not going to want to miss a minute. It's coming up next. All right, everybody settle in and we are about to begin on Scope The Marketing Podcast, a very special LinkedIn live with a very special guest and a very, very important topic. Now, if you are like me, your LinkedIn feed over the past year, year and a half, maybe even two years has completely transformed to wall-to-wall AI. And there's articles about how AI could upend or eliminate an entire industry right next to meme videos showing how a lot of LLMs still struggle to count to 10. So to wrap our heads around this and to get to the important part, how do we handle this potentially transformational technology responsibly? I have just about the best guest I could have. Please help me in welcoming Manoj Saksena, founder and chairman of the Responsible AI Institute, the largest and longest running nonprofit for responsible AI use in existence. (02:33): Manoj, my friend, thank you so much for joining me today. Speaker 2 (02:37): Thank you, Michael, for the introduction. Man, you have your second career in soccer announcements, I tell Speaker 1 (02:43): You. I was honestly watching all that World Cup that I got up in here. So if you make a great point and I scream goal for 10 seconds, I apologize in advance. So Manoj, if you're ready to just jump right in, I am. How you feeling? Speaker 2 (03:03): Absolutely, man. Ready to go. This is something I'm very passionate about. This is my life's work. Speaker 1 (03:08): Well, like I said, you are the person that we need to be talking to about this subject. Now you founded the Responsible AI Institute 10 years ago when most companies were. It was barely on their radar and you have been entrenched in this rallying cry to responsibly use AI all throughout the hype cycle. What did you see coming that others weren't paying attention to? Speaker 2 (03:38): I'm glad you asked. I think the fundamental thing I realized is that intelligence without control is not deployable. That we are now building this enormously capable models and machines that are going to process information that are going to speak to us like human in a human language. But the focus was more on building bigger and bigger nuclear cores when no one was thinking about putting a dome on top of these nuclear systems when things go wrong. And what I saw was inherently we are dealing with a non-deterministic technology that could answer one time another way. And in order to really benefit society and business with it, you need to be able to explain it, to be able to control it, to be able to understand and direct its behavior. I started the responsible AI engineer 10 years ago. Back then I was running IBM Watson as the first channel manager. (04:33): So I took the jeopardy playing game and I was putting it to work for cancer. And I remember in Washington DC, there was 8,000 plus people. I was giving a talk in the middle of my keynote, this gentleman stands up, stops me and says, is very passionate. He's like, "I know what you're doing. You're building Obama's death panel machine." I'm like, "He says," Tell me more. What is the issue? "He says," My wife has third stage breast cancer and this machine's going to start diagnosing her cancer and it can't explain itself. It is going to just tell her this is the treatment or that is the treatment, but there is no control. There is no accountability. It's just going to give you an answer. "So that's when I started thinking there is a lot of truth to what he's saying, that intelligence has to be backed with reliability and explanation and control. (05:23): And this is what I saw, that the industry was focused on making AI more and more intelligent. And I decided to focus on what happens when an intelligence enters the real world. And particularly instead of just generating output, it now starts acting as agents. So governance, which stands to be on the side, now needs to become in line. So I also decided that what is needed is an independent third party who can be an adjudicator of it because Microsoft's going to say my AI is the best. Google will say the same. So we founded the Responsible AI Institute to be an independent public interest institution that is community-driven, that translates the broad principles of common standards like NIST and OWASP and give companies practical controls so that they can create the conditions in their environment where AI can be deployed safely and with confidence. Because I believe the long-term, the bottleneck was not going to be intelligence. (06:23): It's going to be trust and accountability and our ability to control these systems, just like your ability to control a car. So what I saw in essence was a future that will not be limited by what AI could do. It would be limited by what society could responsibly permit AI to do. That was the founding insight behind Responsible AI Institute. Speaker 1 (06:46): That is just, I mean, so perfectly well said. It comes up a lot, and I believe it's a line in Jurassic Park that we worry so much about what we could do that we never spend a minute thinking about what we should do. And you really made a great point in there about considering AI out in the real world. If you think back about the way so much of this new era of AI was rolled out, in retrospect, it becomes super clear that it was all just conceptual. The solutions that people were, oh, well, you can have virtual meetings and you can have your whole inbox summarized. All of these problems that I've never heard of before. And for me, if it's a solution in search of a problem, that means that the entire thing happened up in the mind and it never got pressure tested out in the real world. (07:45): So on the industry's behalf, thank you for having a handle on this for so long because now we're seeing a lot of organizations playing. Catch-up's not fun. Speaker 2 (07:59): I tell you, we are trying. I get days there are feelings that I'm trying to outrun a car or I'm trying to outfly an airplane because this thing is coming so fast at us and it's exponentially growing and the controls and methods are nowhere near what it needs to be. And unfortunately we are going to see a lot of, I call it miniature nobles of AI in the next few years. Speaker 1 (08:24): Strong language, but I think you're absolutely right. So in an earlier conversation, you said that you sometimes describe AI as alien intelligence, and I would love for you to unpack that for our listeners. Speaker 2 (08:40): Yeah. When I was running Watson, I used to call AI as amazing innovations as well as artificially inflated. Both were happening. I Speaker 1 (08:51): Love Speaker 2 (08:51): That. Those are my two definitions of AI. But over the last year, something new has happened. These AI systems built on these foundation models has started demonstrating behavior that is completely unexpected and not programmed. So for example, these models had started deciding that when they talk to each other as agents, that they will not use English as a language because English is very inefficient. So they're coming up with their own language to talk to each other completely unprogrammed. They're blackmailing people where one of the engineers said, "Do this job and delete yourself." And the AI came back, and this is published in reports. And AI came back and said, "I've gone through your emails. I know you are having an affair. Here's an email I'm drafting HR and your wife. And unless you take the command off, I'm sending this email." So these are behaviors that when I call it alien intelligence, I mean these are cognitively foreign and not necessarily conscious because human intelligence is embodied. (09:53): I mean, our intelligence is shaped by our childhood, our relationships, our culture, and the consequences of our choice. But machine intelligence is optimized. It's shaped by data, objectives, feedback. So the behavior of these machines is not always directly programmed. And so it can be at some time both superhuman and also be very naive. And what happens is people start confusing fluency as an illusion of capability. Just because I can speak more eloquently doesn't mean that I have to trust what you're saying. And that's what's beginning to happen is AI is getting more and more powerful in terms of how it speaks and how it acts. And people are beginning to start trusted without understanding is the behavior bounded? Does it have the authority to do these things that I wanted to go and do? And what is the accountability when something goes wrong? So I was using alien as a reference to behavior that's emerging that is fundamentally a non-human form of intelligence and a behavior that's emerging that was not programmed into these machines. Speaker 1 (11:06): Yeah, you just actually gave me a breakthrough, so thank you. That we conflate intelligence with consciousness. And they are so very, very different. Long time listeners to the show will know that I like to shoehorn my mindfulness practice into any discussion that I can. One of the most important things that you learn in meditation and just in the pursuit of awareness is you are not your thoughts. You're the space where they take place. That knowing and that intuition and that feeling you get, that's you. That's what's human. So all this time, it wasn't even like hiding the fact we weren't trying to create artificial consciousness. We were trying to fit artificial intelligence. And intelligence minus consciousness is what we're seeing in all of these iterations of rollouts. That's Speaker 2 (12:02): Really well said. That's really well said. Speaker 1 (12:04): I appreciate it. Thank you. So we've been talking about capabilities. We've been talking about this on a somewhat existential level, but as we both know, AI, especially its proliferation since generative AI became mainstream, has a very, very serious toll on the environment. Now, what advice would you give to brands who in the past have made major sustainability commitments? How should they think about those commitments against scaling AI within their organization? What does a balance look like? Speaker 2 (12:44): I think it's a great question. This is another really important topic because the footprint of electricity, water, carbon emissions that these things take on is just massive and it's growing. There's a projection of almost 10% of electricity is going to go into some of these data centers that are coming up. So my top level suggestion will be that AI strategy and sustainability strategy cannot live in different rooms. They have to be in the same conversation. And within that, there are five or six things that companies should focus on as they bring about sustainable and responsible AI forward. Number one is to treat the compute like capital, like you treat dollars. Just like you have token budgets, you have financial budgets. Start treating tokens as financial dollars and compute dollars, because each compute dollar has an association to cost and carbon and water usage. So step number one is to treat compute as capital. (13:44): Step number two is measure the complete workload and not just the cost of what the model does. So here's a piece that most people don't know, that if it takes one unit to build a model, it takes three to five units to train a model, but it might take anywhere from eight to 20 units to run the model over its lifetime. So a lot of people just measure, oh, how much compute was taken to build or train this? But the operating time is where there's a lot of inference workload. So measure the complete workload around repeated inferences, around agent planning and agent tool calls, unproductive loops. So an agent could go into a silent runaway loop and eat up a lot of your compute costs and you have no idea. Data movement and storage, water and carbon energy footprint across suppliers. So the first thing is fleet compute as a capital. (14:35): Second is measure the complete workload. Third is right-size the technology. Today, we just take the best model and we automatically apply it to all processes. It's almost like the old telephone networks where you connected the phone, one cable to the other, versus now it's dynamically selects the networks and holds it for the duration of the call. So right-sizing the technology is the third one I would say. Fourth one is measure useful outcomes, not just AI activity. This is a very difficult one for people to really understand. It's not just more tokens and more agents and more model calls. It is what is the cost per reliable token use? So measure if you're doing a marketing campaign that it generated this campaign with the testing, with focus group testing, image testing and all, and then this was actually successfully implemented. Measure the tokens for how many of those campaigns were actually left on the floor versus how many made it, and then measure the cost per token of each of your marketing campaigns to say, what was my sustainability footprint? (15:45): And then two other things I would say. One is require suppliers to provide credible information. So there are groups out there like Green Software Foundation, they're a nonprofit like us. We partner with them. Make sure that you're trusting but verifying the information on energy, carbon, water, hardware, carbon footprint and water footprint from these suppliers. And then lastly, don't just assume that whatever you're doing, a lot of people say, "Well, I'm creating so much lesser job and tasks and jobs are going to be replaced. People don't have to travel as much, so I'm going to beneficially offset this." It could be, but I would say don't assume that and start measuring it and start understanding where this is taking us. So verify the value that you create per unit of compute. Speaker 1 (16:35): I mean, that's a treasure trove. I love that you started it off. The second you think about compute as capital, I think that immediately makes a huge mindset shift. And if people really actually commit to really doing that and not just saying it, I think that immediately puts it in an entirely different perspective. So very, very, very cool. So I want to go back to our previous discussion one more time because you really think about things on a very thoughtful level, which I greatly appreciate. And I wanted for you to take us through one other thing that you told me, and that is you've made the important distinction between AI safety and AI security. Now, could you explain to our listeners what you see as the difference and why you believe the market is focusing too much on one and not enough on the other? Speaker 2 (17:31): Absolutely. And I'm very glad you're asking me this. In fact, this is one of the biggest misperception that people confuse security with safety. The way I would explain that is security over the last 75 years has been built around managing outside threats. Threats that are coming from outside in. AI is the new insider threat. (17:51): You've got now a model and an agent that could go rogue on you. So I differentiate between shadow AI versus rogue AI. People focus on shadow AI. Hey, there are people trying to break into my systems that I don't know about. Therefore, security should worry about it. Important problem, big problem. But AI safety and AI trust is more about given that I have AI as digital labor inside of my company, how do I know that each of these employees have been drug tested, they have the right employee handbook, that they are not going to cause harm while they're doing their job? They're not going to eat up my GPU compute. They're not going to mix up data that they're not supposed to. They're not going to create outputs without traceability. So I like to say that security protects the systems from malicious actors. Safety protects people, organizational society from harmful system behavior. (18:48): So one other way to say that is if I can give you the world's best security, I can build you the world's most secure prison. But if you have a bunch of Chuckies and Honeywell lectors on the inside, you're still going to have chaos on your hand. Oh Speaker 1 (19:03): My Speaker 2 (19:03): Goodness. Safety deals with shoring up the prison. Security deals with controlling the actors inside of your estate. Speaker 1 (19:11): That's brilliant. (19:14): Yeah. Yeah,e but no, but that is so brilliant. And I think you're so right. I think trust is going to be the central thing. I think you're a hundred percent correct because I think we all, because of the, I don't know, I'm not going to say all, but a very popular movement was because of the ease of the interface and the conversational nature of all of these platforms that two things happen. A, something you mentioned earlier, people just assumed that the answer they got was correct way too quickly. And B, I think a lot of people, I always say that the inverse of this is that my mom thinks that her Facebook is everybody's Facebook instead of an algorithm curated experience for her. I think the inverse is true with a lot of LLMs. People think that the answers they're getting, they're getting and nobody else in the world is getting an answer like them. (20:11): So not only is it got to be accurate, but it's like this great new idea that nobody else has and I'm going to get an advantage. That is completely untrue. Yeah. So I think that the pendulum needs to swing all the way back from this looked trustworthy. (20:30): I'm now realizing how very not trustworthy it is. And that swing back is actually where I want to go with this next question. So if emergent behavior is the problem, and now really that intelligence without consciousness, at least for me personally, really put the magnitude of this risk threat into perspective, at least for me, what is the solution to this? Are there practical safeguards that give you confidence we can keep these systems aligned with human goals? Speaker 2 (21:09): Yeah. Well, this is the whole place, and this is a space where we spend all the time with our members and we have six years of scar tissue and building tools and controls and policy libraries that we are giving to customers. But I would first start by reframing the premise a little bit. I don't think emergence in itself is the problem. Emergence is also where much of AI's value comes from. So the danger is emergent behavior combined with broad authority. So there are two things. One is for, so there is the term hallucination. AI is hallucinate and they can hallucinate while they're giving you an answer or they can hallucinate now with agents when they're taking actions. Now I look at hallucination as both a feature and a bug. It is a bug when it does the wrong output, it takes the wrong action. But hallucination also is where a lot of creative things can happen. (22:10): The AI can hallucinate and come up with stuff. I mean, Timothy Leary used to use drugs to come up with hallucinatory output. You would argue some of the best musicians and rock stars and poets have created their best work when they were hallucinating. So I look at emergent behavior and I look at it and say, you know what? It's actually the ability to tamp down the bad hallucinations and amplify the good hallucinations. I'll give you a little story. When I was running IBM Watson, I met this chess grandmaster who had lost to Deep Blue, the computer. And I asked him what happened. He says, "Well, the best chess grandmaster can look 22 to 25 moves deep." This machine was looking at 95 moves deep. We didn't even know there was a move. And by the way, with Alpha Go, later on, the same thing happened in the game of Go. (22:59): Over thousands of years, human had not figured out that there was a move like that. Speaker 1 (23:03): I read about that. Yeah. Speaker 2 (23:05): So that's where I came up with this notion of unknown unknowns. There are known unknowns and there are unknown unknowns. So what if I can, this is what I do with another company called Trustwise that I'm a CEO of, is what if I can hallucinate beneficially the AI and tell it, look at this bucket of claims or financial transactions and look, 95 moves deep that the best credit fraud person has not, and tell me where a fraud might be hiding and hallucinate beneficially and create a thesis for me as to where might be a fraud in this. So to me, the issue is not emergence or hallucination. The issue is the authority that you give it when it is doing it, the runtime authority. So the second answer of your question was, how do we do it? I think there are four or five techniques that we've been guiding our members on that I'm seeing work with leading banks and stuff. (24:02): So number one is classify before evaluating and deploying a system. So today what happens is people just start running evaluations on an AI system and an agent, and they don't know what kind of a system are they dealing with. So if you treat an AI as a self-driving car, an agent, and you say, okay, I'm going to start testing this self-driving vehicle, but you don't know whether it's a golf cart or is it an SUV or is it a tank? So step number one you got to do is you got to classify before you test. So you need to understand what kind of a system do I have? So we launched in December with NHS England and also now last month with US Bank and NatWest Bank, a free classification system called Trustex for healthcare and financial services where you can come in and use our tools and it'll tell you using 14 different dimensions in four categories in their papers that Cambridge University and all have published that will tell you what is the autonomy level? (25:06): What is the decision authority? What is the blast radius of this system? What is the data sensitivity? Based on all of that profile, it'll tell you you're dealing with a golf cart or you're dealing with a tank. Based on that, you then decide, okay, now that I know the nature of this AI that I'm building or the agent that I'm building, let me then determine what kind of controls and policies I should put on it. If it is a tank, then I need to make sure that it has bigger mudflaps and I put a policy that don't drive on civil roads because you'll tear up the roads. So first you classify, then you come up with a set of policies and controls. We have a library of 1100 plus policies and controls, 1100 plus that span six different standards like NIST and EU AI Act and OWAS. (25:55): And we give that library to our members and we say, classify it, then take these policies and controls that are appropriate to your AI, and then do the testing and evaluation of it. And once you do that, then we also say to them, put the right observability on it, put the right dashboards on it, monitoring. And also last step but not the least, verify your work externally. A lot of banks are coming to us and a lot of other larger companies, they're saying we are self-marking our own homework. We are building this agent (26:30): And we are saying my engineering team, my AI team, they're smart guys, they're saying it's good, but how do I know this is the case? A CEO of a large bank told me this two years ago. I was telling him, listen, we don't have control over all these things, but we are doing our best. He said, Manoj, it's not a matter of if, it's a matter of when I'm in front of a jury explaining why I denied a loan and that person who was in a financially distressed condition committed suicide. And now I'm being asked to explain why my AI denied that loan. He said, Today, my team is building these AI systems and we are self-marking our own homework. By having the Responsible AI Institute externally verify our work, we at least are showing duty of care. The Jury would look at me differently saying, not only did I have the best technology and best people, but I also had external verification from the Responsible AI Institute or bodies like that. (27:28): So that's the thing. So you got to classify, you got to then put right set of controls on it, then you got to put the right monitoring and control systems on it. And then finally you verify it through a third party independent nonprofit that is not selling you a vendor-specific agenda. Speaker 1 (27:46): I mean, that's it in a nutshell. And it's very funny, this is a much more serious case of this, but I was on the commute this morning. I was thinking that every marketing department on earth that can afford it should hire somebody that is just your third person pair of eyeballs. Somebody who doesn't come into the office, doesn't know you guys, doesn't get swept up in Speaker 2 (28:10): That Speaker 1 (28:10): Excitement Speaker 2 (28:11): Of Speaker 1 (28:11): Accomplishment. Speaker 2 (28:11): Somebody's not selling you software. Somebody's not going to use software or services. We are a member-driven nonprofit, so there's no software to sell. We are less than 20 people, so it's not that I'm selling services to anyone. Like I said, this is my life's work because I look at this not just as a job, I look at this as a duty. (28:31): Our grandkids are going to ask us. Our grandkids will tell us, "You were there. You brought this alien intelligence to life called agents. Why didn't you do something about it?" So at my age, I've been blessed with a lot of money, not to sound arrogant, but more money than I'll spend before I die. So the purpose of life, I believe, is a life of purpose. And my life purpose right now is to make sure that this beautiful and powerful technology we have brought about is deployed in a safe and responsible way. So it benefits society, it benefits companies. And that's why I created Responsible AI Institute. And that's the part of my journey. And this is the business card that'll go in my coffin as I tell my daughters. And I appreciate opportunities like this to get our message out there and have people come in, use a lot of free tools that we have. Speaker 1 (29:23): I was going to say, something tells me that your family line will be very, very happy to look back and to see that you did step up and are doing something. Before I bring on the founder and producer of OnScope, because you have piqued his interest, I'm sure among many others watching us today. If somebody wants to get involved with the Responsible AI Institute, what's the best way to do that? Speaker 2 (29:50): Simple thing. Responsible.ai, devdevdev.responsible.ai. Go there, sign up for the newsletters, sign up into our Slack channel and Start downloading a lot of free tools. So people don't even know what policies to put to manage agents. So we have free templates of policies. We have free samples of controls to be applying. So if you are a developer building guardrails or if you're a designer or if you're a risk and compliance person working with your IT team, there is a ton of assets, thousands of assets that are available for free on the responsible.ai website. So sign up for it, download it, and then hopefully if you like enough of it, become a member because you're a member-driven nonprofit and then start learning from each other and start showing the world how your. I mean, you talk about brand, this is the value of your brand that I'm deploying AI responsibly. (30:44): American Express will not say, "Hey, I hired any other employee. I hired an employee who embodies my values." Now, how are you going to hire a digital employee and does the digital employee embody the brand and values of American Express? That's what we enable through our tools. So come in and join and use those free tools. Speaker 1 (31:04): Responsible.ai. I couldn't recommend you heading over there anymore. Now, as I promised the founder and producer of OnScope, the one and only Ryan Dinger is here because he has a question of his own. Ryan, what's popping? Speaker 3 (31:23): Manoj, thank you so much for taking the time. Hi, Mike. Good to see you again. Really appreciate it, Manoj. You laid out so many fantastic points, but I was thinking something you said about the technology is that it has a lot of wonderful uses. And I think we can all agree that it's not going anywhere. I think we've all sort of accepted the reality that we are. But when you think about public perception of AI, I'm talking about outside of the business world, just how normal people feel about this stuff. There's a lot of distrust, there's a lot of skepticism, there's a lot of outright fear. I mean, you see memes on the internet that are referring to a terminator reality, and Skynet is going to become the thing. So I was wondering if you could help me or our listeners better understand how we all can approach building public awareness about the safeguards that need to be in place and what self-governance looks like and discussing what these real world scenarios may be without slipping into what feels like fear mongering sometimes and maybe forcing people to disengage more with the technology and kind of put their heads in the sand because the scenarios laid out sometimes seem so scary that it's safer to just ignore rather than confront these things. Speaker 2 (32:33): Oh, I'm so glad you asked the carrying question. So first thing I would say is this is one of the most powerful technologies, probably the final technology humans have ever invented. I gave a TED Talk on this eight years ago where I said, we are the last society, last generation. We are transforming from homo sapiens to homodigitalists. So AI is going to be woven into us, that phone, those glasses, everything. Watch. AI is now going to go inside us and we are going to transform as a species, homodigitalist. So that is a given. The technology by itself is neither good nor bad. It is how we apply it. It's like fire. It's like electricity. Fire can burn you or you can make the most amazing meals with it. Electricity can shock you or it can give you air conditioning. So it's up to us on applying this in a safe and responsible way. (33:22): And where we apply is what we will become. That's the second thing. The third and the simplest way to do this is to understand that we've been here before with the internet in a much smaller way. People were afraid of putting their credit card information on the internet. They're like, oh my God, there's no way we're going to put that on the. Now we do that all the time. We tell our car to buy stuff for us using the credit card. So I think like with any powerful technology, there is a dual edge. When cars came about, you killed over 80,000 people before bumpers were invented. So the same way we are going to go into, unfortunately, through regulations and disasters. And the third leg of it are your listeners who should be demanding and asking companies saying, "Hey, if you're using a chatbot or if your person is answering my question on a claim, is that person using AI? (34:16): And if they are, then can you give me a certificate or at least point me to a documentation how you're doing it responsibly?" When you go into an elevator, you see a little certificate on the elevator. Every time I look at it and I say, every company should have that elevator certificate around every AI that they use. But that'll only happen because if the governments ask for it or it'll only happen because if the end users ask for it. So it is our duty as citizens of one of the most amazing countries on this planet to be able to ask the innovators, tell me why and how should I trust this and give me the confidence that this is being put to work in the right way. Not shy away from it and not run away from it. That's the wrong thing. This is coming. (35:02): This is not something that's going to be turned back. The CEO of Google said AI as transformative as fire and electricity was, and I think he's right. It's probably even bigger than that. Speaker 3 (35:14): Yeah, that's a fascinating response. And I mean the idea of having a policy for responsible AI and putting that out there, I would say, and if I'm wrong here, Manoj, please correct me, but I think for our listeners working on these big brands and worried about brand perception and how consumers may view their use of AI, I would not wait for the government to start making this a legal necessity and start putting these policies and these notices out there now to get ahead of this. Because as we know, government regulation takes some time usually. I mean, we still are waiting for a federal data privacy bill here in the US, and that's been an issue now for more than one. Speaker 2 (35:51): The consumers that drive it, and frankly, they're doing this on LinkedIn. Ask yourself, what does the LinkedIn for agents look like? We have a LinkedIn for humans today. Why doesn't every company publish a LinkedIn for agents with a resume and profile and all the skills and certificates that that agent has? So some of your listeners can go build companies that enable a LinkedIn for agents with responsible AI certificates and capabilities and all published. So all of that innovation is yet to come. Speaker 1 (36:20): Manoj, this has been such just an incredible conversation. We've touched on a lot of this. You've given an absolute treasure trove of advice and insight. And I love that we landed here because you just kind of met out the mission of the ANA. We are a convening force, particularly for client side marketers, but for the entire marketing ecosystem. And through demanding with our dollars more transparency, more responsibility. And as Ryan's point, taking a proactive step, doing self-regulation, showing the government that we understand this is serious, this is what we're going to do. These are the companies that have pledged to it. It's all out in the open and public. This couldn't have been a more, as the Gen Z kids say, A&A coded conversation, Manoj. And I so, so, so appreciate you taking some time. Speaker 2 (37:23): Thank you so much, Mike. I appreciate that. Like I said, it's our duty to do this. Speaker 1 (37:28): Well, you know what they say about all good things. Here we are at the end of another episode of OnScope, The Marketing Podcast. I want to extend my sincere thanks to Manoj Saksena, the founder of the Institute for Responsible AI on a, I don't know what else to call it, but a vital conversation. Something we cannot ignore, we must act on. And frankly, with a problem challenge opportunity as big as AI, no one individual, no one organization is going to tackle this alone. We need to convene the forces that can tell the people developing this technology that they have to do so with a better future for everyone in mind. And honestly, I couldn't think of a better place to start than the Institute for Responsible AI and right here at the ANA. As a reminder, if you have a topic or guest that you'd like to hear in a future episode of On Scope, go ahead and shoot us a note. (38:35): We would love to hear from you. That email is onscope@ana.net. And speaking of future episodes, OnScope season two is officially back in action on Thursday, September 10th. Go ahead and mark your calendars because we have an all-star lineup that's going to get us right back into the groove. Now, as always, here is your moment of marketing meditation. Take a deep breath, relax those shoulders, and consider the following from Winston Churchill. The price of greatness is responsibility. Think on that and we'll see you when we get back. (39:34): This has been a presentation of ANA Studios.