Building a responsible AI to partner with humankind
Episode Details
Transcript
Since the beginning of Natasha Crampton’s career in law, she’s had an interest in the areas where law intersects with technology and its effects on society at large. And as technology has evolved, so has her focus.
Today Natasha is Microsoft’s Chief Responsible AI Officer, and no two days are the same. Internally she works shoulder-to-shoulder with engineering teams to build new AI technologies in a way that grounds AI’s awesome capabilities within Microsoft’s Responsible AI Standard. Outside of the company, Natasha contributes to a broad industry, community and government discussion about the laws and standards that will be needed to bring AI to life in a responsible way.
It's a fascinating line of thought, and as she says, there is much more ahead of us than behind. Tune in to this episode of Pivotal to get a glimpse into the ethics of AI and how a diverse set of perspectives and skills are coming together to make this exciting new tech frontier a reality.
Link to full episode transcript.
Produced by Larj Media.
Hayete
Welcome to Pivotal. Hayete Gallot, corporate vice president for commercial solution area at Microsoft. I work with customers around the globe to transform their business through technology. At the center of every transfer information are people who give technology its purpose, and that doesn't change with the advent of AI. It's actually being accelerated. People sparked visionary ideas for leveraging technology. The release of AI technology like Chat GPT, this year is exciting, but it has led to big question as we grapple with the best way to harness those tools to enhance and support the people behind the work. We like to talk about technology. I love to talk about it. But we often forget that technology is most when it supports people with purpose. This season will demystify AI by talking to the innovators using new AI technology to uplift their industries and augment their people from education, to journalism, to surfing. And it just illustrates what AI is about. Everybody thinks it's about tech. No. Everybody's using AI. And that's what we're gonna show you on this season. Today, we're doing things a little differently. We're publishing my conversation with Microsoft chief AI responsibility officer, Natasha Crumpton. We discussed the current state of AI and legislation. What it means to be a chief responsible AI officer, and how to help your company navigate the change successfully. Natasha brings fantastic insights on the importance of balancing innovation, and responsibility and transparency. I'm excited to introduce Natasha Crumpton. She's our chief AI responsibility officer at Microsoft. And I have to say I've kept you very, very busy with all my customer requests. So thank you for the partnership, and thank you for joining us today. Natasha
Of course, Hayete. It's been a pleasure of partnering with you, and I've been so energized by not only how excited our customers are for this new wave of AI, but how committed they are to doing this next wave responsibly as well. Hayete
I have to say it's impressive. Right? Like, the the level of questions we're getting is not just going about the tech right now. They're they're asking you all the right Natasha
Absolutely. And that, of course, is the first step to getting it right. Hayete
You and I work together a lot. Our customers, they want to to go with AI, but they do have responsibility in mind in making sure they do right with this powerful technology. Natasha
They're absolutely motivated to do the right thing here and very eager to learn. And what I've found is that many customers come into the meetings that we jointly have together, and they've already thought about the basics. They've thought about how they can perhaps leverage their privacy program to, apply it in the responsible AI space. They've thought about what their company values when it comes to their own sense of responsibility. And so we can have these very rich discussions about the how? How do we move to the next stage of the program? And it's always really intriguing for me too because it's informative to hear the types of issues and questions that are on our customer's mind, that actually really helps us in our work within the company as well, because fundamentally our engineering teams, are building the safeguards into the systems that preparing their transparency documentation because they want it to be of service to our customers. So it's really helpful for me to hear firsthand from our customers what they're seeing from us that's working really well and what more they would like as well. So you have an unusual job, and we're gonna get to it But before we get Hayete
to it, I'd love to see how you got here. Natasha
So since the very beginning of my career, I've had an interest in this intersection between technology and law and society. So in the earliest days, it was doing things like steady and internet safety and security laws. But, of course, as the technology evolved, so did my focus. So about five years ago now, I I moved from Microsoft and Australia to Microsoft's headquarters here in Redmond, to do something that was very new at the time. There was just this brand new engineering team that had been established called AI and research and my job was to come along and be the counsel, the lawyer to help this group work through some pretty tricky case studies. About a year into that, we realized if we wanted to meet our ambition in this space, we would need to make sure that we had a dedicated policy and governance efforts. So when we stood that up in the middle of twenty nineteen, I stepped in to be Microsoft first chief responsible AI officer and also to lead our office of responsible AI. Hayete
So why ai? Why did you pick the job? You're in Australia, the job came up, where did you pick it? Natasha
It seemed pretty clear to me even back then that AI was just gonna have an outsized influence on all of our futures. And I am a big believer in the positive potential of AI, but I also recognize that the best futures don't just come to us. We have to shape them. And so, I was committed to trying to do my part to apply my background to this, work. And, you know, I had an early experience, when I arrived, that I think really helped me see how we could influence AI. So There was a bit of a, controversy in the middle of twenty nineteen, here in America because there was a suggestion that some facial technician technology was being used at the southern border at the height of the crisis there. Now that didn't involve Microsoft facial recognition technology, but it did make it clear to us that we would need to do two things at the same time. First as a company, we would need to step up to our own responsibilities. So we developed a set principles, and then they worked very closely with the engineering teams to implement them. But we also needed to call for new laws and so we also engaged in an effort to land those new laws. And ultimately, we were successful in that And that for me was just a playbook that I thought, yes. This is something that's both important to do for facial recognition, but also important much more broadly for other AI technology as well. Hayete
And we're thankful for that. Natasha
Thank you. I mean, look, we are We are in a good place now because we have been working on these issues for now almost seven years. So although chat GPT was a big surprise for lots of people. We we knew what was coming. And so some of that foundational work has put us in a really good way to make this new generative AIA error. Hayete
So I'm sure there are a lot of people like me originally who wondered, what does the job of achieve AI responsibility officer look like every day. What do you do? Natasha
Well, no two days are the same. That is for sure. I really have two big parts to my job. So the first part is an internal facing role So I work shoulder to shoulder with the engineering teams as they are building our new AI systems. And that could be right at the earliest days of conceiving of the system and thinking about what are those benefits we want to secure? What are those harms we want to avoid? Or it could be working with the leadership of the company. So we're fortunate that our responsible AI if it is free closely supported by Brad Smith, our vice chair and president, and also Kevin Scott, our CTO, and EVP of AI, So I could be working with both of them to report in on our progress, make sure that we're aligned about our priorities. So a lot of internal work making sure that we're living up to our commitments because fundamentally, you know, that's core to who we are and what we're about. The second external part of of my job is all about contributing to the discussion about the new laws and norms and standards that we need in this space. So what we do there is we try and contribute what we've learned internally about building these systems in ways that uphold our principles. And help policy makers with ideas about how can they secure AI systems to make sure they are fair and private and secure reliable and safe. So these are two both very fascinating parts of of my job, and one's I'm very grateful to have an excellent team supporting me with. Hayete
So when you think about number two, our customers are quite anxious because there is no official regulation, whether it's in the US or rest of world, It's still out there. Right? And so you're doing work. Natasha
What does it look like when you're working with government and policy makers? So often a starting point is thinking about how existing laws might apply to AI systems. And there's quite a lot to draw from there. So if we think about, for example, the GDPR, Europe's well known, privacy law which has become the gold standard everywhere. That type of law is neutral in the way in which it applies, which means it applies to AI systems as well as much more traditional forms of of software. So policy makers often think about look how can I use existing law to begin with to try and regulate AI systems? Often they can lose existing law alone is not going to be enough. And so then the conversation becomes what should those new laws look like? And oftentimes, the policy objectives are quite clear. Policymakers want to make sure that AI systems are safe, they're secure, they're trustworthy. But they haven't yet figured out a clear path to making sure that that is gonna happen via law. Now In Europe, in fact, they've been working for about two and a half years now on a comprehensive piece of, legislation, and that might land towards the end of this year. But it's not just in Europe that this action is taking place. In the US, we've developed a blueprint for, an AI bill of rights that's been developed by the White House there's been NIST has published an AI risk management framework. The White House earlier this year, worked with leading AI labs like Microsoft to come up with some voluntary commitments. So overall, I'd say that the pace of policy activity has really picked up. And what I think is really promising is that most policy makers are now realizing but, you know, they can't do this work well alone. Policymakers need to work with industry counterparts who have deep knowledge about these systems have been built, but also members of civil society in order to get this regulation right. It's definitely going to take a multi holder approach. This should Hayete
be an interesting point because I think you always worry with tech. Right? You want policies to make sure we're safe. At the same time, you don't wanna prevent the technology from happening because you're so stringent. That's right. Then you can't use it anywhere So how do you work that lane? Natasha
Yeah. That's right. Look, my strong view is that innovation and regulation, compatible with one another, there's not an inherent trade off between the two. But it is important to make sure of two things. One is that the new laws are clear and certain. So, you know, organizations, particularly small and medium enterprises are not left wondering whether they've met the law on or not. And the second is that they need to be proportionate. Right? They need to be well structured so that they achieve the policy objectives but they do that in the most efficient way. Hayete
So you had to establish some rules for Microsoft years ago for us to developed safe AI systems. How did you go about it? Natasha
We did. And we developed something that we call our responsible AI standard. So what we did at the company was we started from our six principles that we adopted back in twenty eighteen. These servers are Northstar. So we believe that AI systems need to be fair, private, and secure, reliable, and safe, inclusive, transparent, and accountable. Now those six principles have been durable over time, but in reality, they're not sufficient on their own to guide engineering teams because engineering teams ask me questions like, what does it mean for the data that I need to collect? What does this mean for the testing I need to do before I make this available to our customers? So what we did is we took those six principles and we double clicked down on them and we thought about what are the steps that our engineering teams need to take in order to uphold these principles. We're actually onto the second version of our responsible AI standards now, in part because we learned that the first version wasn't concrete enough. The engineering teams were thirsty for very actionable guidance. So we've provided that in the second version, and we keep evolving it as well. So just recently with this wave of generative AI systems as we've been building all of Microsoft's cool pilots, for example, we've taken our general framework, the standard, and we've developed some more specific guidance for our engineering teams about generative AI application specifically. And that really combines the policy requirements with the tooling that we've been building internally within the company to carry out those policy requirements. So the engineering teams essentially have a one stop shop where we say, this is the policy requirement, and here's the tooling to help you to do that. So it's been a journey, a journey only made possible by the fact that throughout the whole process of developing the standard, we've had three groups of people working very closely with one another. The first is our researchers who help us look around the corners, understand what the mitigations are. The second has been my policy and governance team who have been pulling in the signal about where the law is going, and had lent on the expertise of how to govern and enable this type of effort, and then, of course, the engineering teams. Fundamentally, we're a product building company, and so we've gotta make sure that we've got engineering teams in us working shoulder to shoulder on how to make all of this real. Hayete
So when you look back, where did you get it wrong at first? Cause I think it's always good to learn from our sense. Right? Natasha
Yes. Absolutely. Look, I think the place where we got it wrong was that, we provided a lot of consideration Right? So for each of the principles, like fairness, for example, we would say, consider how the system could behave unfairly. Consider these past examples of how things went wrong, but considering a loan is not enough right, the engineering teams were really thirsty for specific steps that they could take. So what we did is we tried to, look cases where we'd had, fairness challenges, for example, with our facial recognition technology, and we tried to pull out the passions as to how we went about identifying what the problem was measuring what the problem was and then ultimately mitigating the problem as well. By going back and learning from those past experiences, pulling out those patterns, we were able to come up with a standard that was just much more concrete and specific, which is what really helps our engineering team skip to the next level. That's very interesting. Hayete
So now how have you operationalized that? Cause I know there is going from standards Yes. To being very precise on what you're recommending. And then people, actually, on their day to day lives, actually using it absorbing it and just adopting it. What is different? Like, how did it change the way we develop software? Natasha
Yeah. That's a great question. You're absolutely correct that it is a lot more than what just writing the policy. Right? I'm not interested in having an excellent policy that stays on a piece of paper. Right? And we want to make this real and just an everyday part of the way in which we build products here at Microsoft. So we've worked on really, I would say, quite a comprehensive approach. The first is we've put people in designated roles across the company to make sure that there's accountability across the company. So for example, we have a network of responsible AI champions and all of the engineering teams and also our sales organization around the world. Those champions report to division leads who help to make sure that there's a consistent approach to implementing response bull AI across their division. And then ultimately, we have accountable, corporate vice presidents who are part of the leadership group that works with me and with Brad and Kevin to make all of this real. So getting those governance structures right is one important step. The second is to engage in training. You know, we're very conscious that few of our employees across the company have had the benefit of specific, you know, ethics training, although that number is now increasing. So we really actively train our community on our approach And that's an ongoing process where, you know, we start with general training, and then we get into more specific role based training. And then I say a third part of it is that we've learned over time we've needed to develop templates and tools to help our teams carry out this work. So for example, we require the teams to carry out an impact assessment, just one of the first steps of, following the responsible AI standards. And we have a template for, teams to do that and also very detailed guidance to help teams work through it, both of which we've actually made available publicly to help our customers in business with this work too. But in addition to that type of template, we also need actual tools. So when we're measuring, for example, the fair performance of our AI systems. We need tooling to help our engineers and our product building teams do that. And our general approach has been to both build those tools for Microsoft. Often we've made them available on an open source basis, And then ultimately, we have incorporated them into Azure, ML, our AI development platform here at Microsoft. And the reason for that is that we know that if we need to build and use these tools, then our customers and partners are going to need to use them as well. And so we really want to build an AI development platform that has all of these toolings baked in from the very beginning so that others can get the benefit of Hayete
it as well. I'm so glad you talked about this because I remember years ago when we launched this concept, remember people saying, oh, no, the tax. This is this random thing. We don't know how to act on it. And you've made it so real. It's impressive now when I see engineering teams from the get go saying, what, where is my researcher? Where is my policy maker? How do we work together? And they have tools. I mean, it's so much more concrete from where we started years ago. It makes it more real. Right? Natasha
I I think that's right. And, you know, for sure, we still have much more ahead of us than what we have behind us, but I think we're on the right path here because you know, to make, responsible AI an everyday part of the way in which we build products at Microsoft, then we do need systems that really docking very closely with the engineering systems that are in use already, because that's the way to really bake it and I do feel like, yes, we've made some some great progress, and that's put us in a good place to be able to meet this next Genote of AI wave and and build the same set of tooling that we will need for this next generation of AI systems. Hayete
So when you start something new, you always have the naysayers, you always have the difficult moment where you're like, I'm not gonna make it. How am I gonna get there? Yes. When was your moment or proudest moment where you're like? We got there. Like, it's working. It's starting to work. It's not perfect, but it started to work. Any moments you can recall, you're like, I feel so good about it. Natasha
Yes. Look, as we were, landing the second version of the responsible air standard, we had, as an actually a preview period with the engineering teams first where we have had a very advanced draft. But we thought, look, even though we have engineers and researchers and policy people in the development of this standard, we'd probably better put it to work in a couple of actual scenarios first. And that the feedback came back and for sure there were aspects that, you know, we got feedback on that people said, you need to change this or that or whatever. But what was encouraging to me was that the fundamentals were right. And I felt like that was a big part of progress since the first version of the responsible AI standard, which we knew was, you know, a guide of sorts, but it wasn't sort of driving the sort of change that we wanted to see. So I did feel pretty encouraged by the time we finalized the response layer standard. And then right away, we were demonstrating to the world that we were prepared to actually apply the responsible layer standard and also make decisions that would mean that we wouldn't necessarily continue in the same way that we had in the past. Just one example of that was that as we were working on our facial recognition technology and applying the standard to it, we we realized that we had some capabilities in our facial recognition technology that didn't line up with our commitment for the technology to be fit for purpose. And so we decided to retire some of our capabilities, including in particular, our emotion recognition capability. We concluded that essentially, you know, that technology was not underpinned by a robust scientific basis. People expressed their emotions differently around the world. And that was a really important moment for me because I think it really demonstrated that the company was prepared to reconsider decisions that would be made in the in in the past that we were prepared to take a principled stand and retire capabilities that we had made available previously. And it's been encouraging to see that The decision that we made several years ago now is now a decision that I start to see reflected in actual policies being developed by external parties. Hayete
It is very cool to see how because of our AI standards, we were able to recognize that this technology is not fit for purpose, and we should acknowledge that. Right. That is very exciting. Mhmm. Now if you're a customer, You're going down this journey and you're trying to also be responsible. How do you guide them? What would you say to them? How do should they get started? Natasha
Well, I think the first and most important thing for our customers to do is to ask this initial question, which is essentially What are you trying to solve for and how can AI help with that? And with the benefit of those initial opening questions, I usually suggest a couple of things. The first is to find a low risk use case where you can start experimenting right away. You might want to use, for example, the Azure open AI service to start analyzing some of the customer feedback that you've received at a scale that you haven't been able to in in the past and to glean some new insights from it, or you could start using the m three sixty five co pilots just to make your workforce more productive, for example, doing an everyday task like preparing meetings summaries and and action item list. These low risk use cases really help you to get a feel for the technology, what it's good at, what it's not so good at, And it just allows you to build momentum quite quickly and make sure that your responsible AI approach that you've started in the company is actually taking hold in the way that you intend it to. I think it's really important at the beginning to assemble and multi disciplinary team. Make sure that you've got not just your technical people, but also try and find some people who have some experience in the in the humanities or the social scientists. You might have policy people. You might have designers or user researchers. These colleagues really bring essential skills to the mix as well. And then I would say just call on the resources that are already available. For Microsoft part, we have published our response blast standard externally. So you could pick that up and try that out and adapt it as you need. As your own playbook for how you build your, systems responsibly. And and by all means, take advantage of the tooling and the template that are out there as well. I think that helps you get a head start. It makes you realize that, you know, in addition to learning from sister domains, like privacy or accessibility or security. There's lots of materials already out there that you can take advantage of, and that will, get you up and running a bit faster. Hayete
Exciting. Yeah. Now our customers, they wanna know how do they get organized, but you think you gave some context how to get started, which you're provided. But then they always ask me, how do I measure success? How do I know that I'm progressing? What metrics? So how do you know? How do you think about Microsoft in our progress? Natasha
Yeah. That that's a great question. So Look, I think we we think about metrics in essentially two dimensions. The first is to measure essentially the rollout and the strength of our program from a governance point of view. So we look at things like, do we have all of the right people in the right role and are those communities sort of growing in line with our expectations? We look at things like what is the, coverage of the training that we've, put out there, have we got good completion rates there? So those types of metrics, Gosh, that sort of overall health of the governance approach across the company. And then we look at specific steps that might be taken in relation to the actual products. So for example, we do, in the case of some of our new generative AI systems, we have developed metrics to look at the content that is produced by these systems. Is there any illegal or harmful content coming out of those systems? And we measure that before we release the products, but importantly also on an ongoing basis as well because, you know, the world doesn't stand still. It changes the nature of the of the challenges that we're trying to address changes, and we want to make sure that our systems are working well in the real world and not just in the labs when we're measuring these things ahead of time. So I think trying to think about both of those approaches, how do you measure your progress in implementing a responsible AI governance program within your organization and and second for specific products or AI systems that you're deploying How do you measure the health of those particular systems? That will vary depending on the context because most things about our systems are very context dependent. But doing that work is is important as well. Hayete
When we think about AI, our customers have anxiety, they have excitement, the industry is the same. The consumers are the same. There is anxiety while there is excitement. As a chief AI responsibility officer, what is top of mind for you when you think about AI? Was keeping you up at night? Natasha
So part of my job is to have a healthy dose of that anxiety so that we can make sure that we are staying ahead of the risk of of these AI systems. You know, for my part, I I find that the most empowering thing I can do is to channel that anxiety into really concrete actions that we can take, to secure our system benefits. So what I'm really thinking about here is how can I take a challenge, for example, the you know, the production of harmful content out of a new generative AI system or a challenge, like, you know, an AI system not being properly grounded in the data that we want it to be grounded in? So I take those sorts of challenges, and then I think about Right. What can we concretely do that reflects the best practices here and put that to work across the the company? So just to give you one example, as we were thinking about, you know, our new generation of AI systems, like bing chat enterprise or the m three sixty five co pilots. It was really important that we figured out how to measure and mitigate different types of harmful content, you know, hate speech, violence, and the like. What we did is we took our examples of what we thought might happen in this space. We fan them out. We expanded the the number of examples we are working with. And ultimately, we built our own systems to try and get ahead of those types of content. Because we're committed to doing this not just for ourselves, but for our customers and partners, we then moved from taking that content safety system that we built for ourselves and making it available to our customers for their own use as well. So just last week, we announced the general availability of our Azure AI content safety system. And that's really the sort of path that I want to take across the board to move from being rightly concerned about certain risk surfaces, building those tools and then making them available. I mean, I I think The energy that I get from working on these challenges is that fundamentally, I believe that this technology is gonna be transformative for all of us. And that we're only really going to advance it responsibly if we do the hard work internally within our organizations to create these new practices and make sure that we're having a broad and diverse conversation about the best way to do that. Hayete
And when you think about your work on policy, are there specific areas where you're more concerned about where you're leaning in specifically as you think about the potential challenges or biases that we could amplify or create. Natasha
Yes. One of the most important things I think that we can do in the policy discussion right now is to do two things at once. The first thing that we need to do is we need to recognize that there are very much here and now challenges associated with AI systems that we need to be squarely focused on. These are things like making sure that we are not repeating, you know, the discrimination of the past and amplifying it via new AI systems, also making sure that people's you know, personal data is is respected and we're putting the right sort of privacy guardrails in place. So those are very real challenges that we need to confront today, and I believe we've got good solutions to address them. But at the same time as doing that, we also need to look ahead and make sure that for the very most cutting edge systems that are right on the frontiers of AI technology. These are increasingly going to present new and different challenges. Some of which will eventually go to national security issues and international competitiveness issues. Now I think one of the most important lessons from past technology errors is that we need to make sure we are today building the policy frameworks that are needed to deal with those challenging issues of the future. So we've gotta keep our eye on the here and the now and also build for the future where we've got even more capable, even more powerful systems, and we need to think about a different range of risks. Hayete
So I know you think of AI as the positive and ESA transformative force for the world. Mhmm. So what do you look forward to when you think about AI? Natasha
Oh, look. I think I am most excited about the ability of AI to help unlock some of the biggest societal challenges that we face, whether that's, for example, making government services more accessible, in India to stakeholders who previously just didn't have access to government grants and the like or helping Canadians fight fire and get ahead of some of the risks there, or even on a basic productivity level, helping us have the productivity that we need as the world faces the shrinking working population in many countries around the world. So all of that you know, is so exciting to me. And I'm just determined to roll up my sleeves every day and make sure that we can put these guardrails in place so that we can secure those beneficial uses of AI. I think we've got an incredibly bright future ahead of us if we do this type of work now. Though as a mom, Hayete
when you talk to your kids about AI, what do you tell them? Natasha
Well, in my first conversations, with my kids about AI were a lot about facial recognition where they found it quite amusing to see what was possible and not and whether they really looked like me enough to be able to unlock my devices and they learned pretty quickly that that that was not the case. So what we talk about now is the fact that they are going to live in an AI mediated future, and that's going to take different skills on their part to really make the most of that technology. So, you know, we talk about the fact that there are some really important parts of being human that no AI system is ever going to replace. Whether it's, you know, creative endeavors, it's their empathy, it's their ability to join dots and think of novel ideas haven't been conceived of of the past. So we talk about what it means to live and thrive in an AI world. And also how it's gonna help us with some of our greatest challenges. Right? I'm optimistic about the progress that we can unlock on things like climate change or some of our greatest health challenges that we have in front of us. So I think they're excited. I think it's become a little bit less mysterious their mom does after the advent of a chat, GPT, and the like. And I think they're excited for it. Hayete
Isn't that cool when they know what you're doing? Sure is. My kid used to tell me you just talk on calls during COVID is that you talk about customers. That's what he thought I was doing, which was very interesting. Natasha
Yeah. Absolutely. So what about you, Hyatt? I'm sure you also have conversations with your children about all things AI and I'm intrigued to hear what what you discuss. Hayete
I think it's kind of similar to what you said, but, like, I have a fifteen year old, and when Chad GPT came out, and I had to sit down with him. And I was like, I I had to really work the fine line with him saying it can be helpful. It have it's here to inform. It can help you get faster to the data. But it does not replace your work. You gotta do the work. You gotta design the assignments. You this is just inform. You're the person who has to be behind it, connecting those dots, as you said. And then for the seven year old, it's been interesting. It's just getting to learn how to use the web and type for search, and he always has ideas about the things he's searching. So for him, it's like, he's learning, and I'm helping him right now. How do you ask the right questions? And it's actually been in a very interesting conversation. How do you ask the right questions? And for a seven year old, it's a very interesting look in watching him learning and improving. It's been exciting. And, you know, I'm also saying be safe, and we all this conversation come with AI. But, yeah, no, I mean, I've been quite positive about it, but at the same time, I've been very clear it does not replace you. And I want you to make sure you use it, and you have to be smart about how you use it, but don't forget you do the work. And it's about you and your creativity behind it. So that's how I I have the dialogue. Natasha
Yeah. Makes complete sense. I mean, I think the more that we instill all of us, the idea that AI systems are a tool for human beings. Humans get to put these tools to work to help us to augment us. And that I think is just so core to Microsoft's way of thinking about these systems. Right? They're our co pilots. They're not auto pilots. We remain in charge. Exactly. Hayete
So you said your job is a lot about anxiety. Did you turn into action? Yes. And it's all about rolling up your sleeves. Mhmm. To make sure that the the world is really Natasha
getting the value of AI and not the potential dangers of it. Thank you. Thank you for what you do. Thank you so much, Hey, Ed. It's a real privilege to have this role, and I'm surrounded by a lot of people who, you know, are just there in the in the trenches every day doing the same work with me. So I'm incredibly grateful to be able to sort of apply Microsoft's commitment in this space and excited for what we can achieve together. Hayete
Thank you. And thank you for the partnership, and I look forward to bringing you to even more customers. Natasha
Thanks, Hayid. It's always my pleasure. Hayete
Thank you for listening to pivotal. I'd love to hear your story in your pivotal moments. Don't hesitate to follow me and share on LinkedIn. Audence information is also available in the show notes. Our show is produced by large media. That's L ARJ Media. Special thanks to Lynn Yang and our partners at We Communications. 
