Beena Ammanath - Executive Director at Deloitte AI Institute

(Joel Beasley at 00:00:00) Hello, my friends. Today we are talking to Beena, the Executive Director at Deloitte AI Institute, and we discuss the ethical questions that surround AI and other new technologies, why it will take an all-hands approach from humanity for AI to reach its true potential, and how cultural diversity of thought will contribute to the success of all technologies. All of this right here, right now on the Modern CTO podcast. Here we go.

(Joel Beasley at 00:00:31) This is the Modern CTO podcast. I sent my team, I said, "Hey, go find me some of the brightest people in the world on AI," and they dug up you. And so now I'm interested. I want to know, how did you get started with artificial intelligence?

(Beena Ammanath at 00:00:54) Oh my god, okay. So I studied back in the early nineties where I started computer science, and part of it was AI. And this was back in India. And at that time, it was more theory than actually doing anything, right? There was no compute power, no easy access to the massive amount of data that we have available now. So I'd studied in theory. I love history, so I did read up about artificial intelligence, and I said, "Wow, this is amazing." Right?

(Beena Ammanath at 00:01:28) And I'll tell you, my first job out of college was as a database SQL developer, and I was working on SQL Server and a software called JAM, which doesn't even exist anymore. But it was back in those days. Right? And when we used to even talk about personalized ads, it seemed very futuristic. You know, is it even going to happen? And now, fast forward to today, we are right in the midst of it. All the things we could have imagined, we were imagining thirty, forty years ago, is now becoming real, or at least it's heading towards reality. Right?

(Beena Ammanath at 00:02:12) I say that there's a lot of ideas people have probably had for decades but didn't have an opportunity to actually try it out. And now it's providing a forum for us to try this out in the real world. So I think we live in very, very interesting times. We are so fortunate to have seen this whole explosion of technology as a whole. Right? I remember growing up, you know, listening to radio stations. We didn't even have a TV at that time.

(Joel Beasley at 00:02:40) Mm-hmm.

(Beena Ammanath at 00:02:40) And the first year was black and white TVs, is what I was looking at. I'm totally dating myself here, Joel. But yeah, that's my early memory of listening to radios and having black and white TVs, and then came color TV. And now look where we are. Right? So I think my experience with AI, my curiosity with AI has not been so much about AI, but about data.

(Joel Beasley at 00:03:08) That's the food for AI, though, right?

(Beena Ammanath at 00:03:11) Exactly. It just happened to be because I've always been a data geek. I've been interested in things that you can do with data. Started out as a database developer, SQL developer, went on to be a DBA, and then managing a data team, data architecture, doing all these different roles within the data domain. Then came the era of BI and data warehousing. Remember Snowflake schemas and the way we used to normalize the data to enable better reporting? Then that was the whole wave of BI, right? Business intelligence.

(Beena Ammanath at 00:03:47) Then came big data, which enabled machine learning and AI and the things we're seeing today. So I say that I still am very passionate about data, and AI is just another manifestation of how data can drive outcomes and change the world to be a better place.

(Joel Beasley at 00:04:08) So what are you doing on a day-to-day basis?

(Beena Ammanath at 00:04:12) Right now? So I have been at Deloitte for about eighteen months, and I lead Deloitte AI Institute. And Deloitte AI Institute is focused on connecting all the different dimensions of the AI ecosystem. Prior to this, I was the CTO for AI at Hewlett Packard Enterprise. Before that, I led data sciences and innovation at GE. So I've been in different industries and different verticals, and one of the challenges that I saw was the AI ecosystem is exploding. Right? It's just growing really rapidly and fast. So it is impossible, even if as a business, if I know of the business challenge, I don't know the fastest way to solve for it. I'll take an example.

(Beena Ammanath at 00:04:59) For example, I know I want to predict when a jet engine might fail so that I can prevent unplanned downtime. Right? As a business, that's a problem that I want to solve for. What I don't know is how do I get to the solution in the fastest way, assuming that I know what data I have, how much historical data, where is it stored? I understand the data part. But how do I get to this predictive analytics, predictive insight? Is there a product out there that I could buy off the shelf, train it on my data, and then deploy it? If there is no such product, is there a startup working on solving for this that I could potentially invest in, do an acquihire, or partner with them to get to that solution faster? The goal is how do I get to the solution the fastest way? So if there is no startup, is there a PhD student, or is there somebody in academia trying to solve this problem for predicting a jet engine failure?

(Beena Ammanath at 00:05:59) Right? So is there somebody in academia working to solve this problem that I could look at? If none of this exists, then, okay, that means I need to have a team that's going to solve for it. And that's where the question comes. Do I use my internal data science team to solve for it, or do I partner with a vendor externally to solve for it?

(Beena Ammanath at 00:06:22) So once you figure this out and say, okay, I'm going to maybe leverage my internal teams to solve for it, then comes all the nuanced topics around it as to what are some of the regulations around using this data to predict this kind of behavior? Is there any regulation that I should be aware of in different parts of the world? Is there any policies? Are there any best practices? Should there be ethical concerns around this? Right? So understanding all the nuances of solving. So we started with the core business problem, but to get to a solution that can be scaled, it's a very complex ecosystem that you need to navigate.

(Beena Ammanath at 00:07:03) So the Deloitte AI Institute is really focused on the applied AI. So starting with the business problem, connecting all these different dimensions of the AI ecosystem, bringing together solution in the fastest possible manner. I also lead our Trustworthy AI practice. And what Trustworthy AI is really focusing on is the AI ethics aspect. Right?

(Beena Ammanath at 00:07:29) Because I feel, let me take a step back, Joel, and you'll appreciate it being a geek in this space. Really, if you look at the times that we are living in today with AI, right, think of it: there are three parallel streams. The first stream is really the core technology itself. Right? Whether it's deep learning or quantum, the core technology for AI, that is still evolving in academia and research groups. And there is a second stream, which is the applications of the core technology. Right? It is being applied today in the real world across industries, across functions to drive real business outcomes.

(Beena Ammanath at 00:08:09) Right? So it is like you are developing the car engine, and in parallel, you're using it to drive in a massive way. And then there is a third stream, which is really on the consequences of the applications of the technology. And the consequences include things like ethics, regulations and policies. What are the risks associated with using a developing technology in the real world? What are the consequences beyond that value creation, which is what everybody focuses on? But by now, we all know there can be other consequences which are beyond that value creation. So that's the third stream. So how do you solve for this third stream where there is no such playbook today? Right?

(Beena Ammanath at 00:09:00) We don't even track for it. We don't measure for it. We don't know how to solve for it. That's what we call Trustworthy AI, is to really think about ethics, regulations, policies. The challenge with ethics is there's a lot of hype around it. You have a lot of headlines, and that's also a big concern because ethics is not just about bias and fairness. I think we have to get to that more nuanced discussion on what does ethics mean for a manufacturing plant versus what does ethics mean for a jet engine manufacturer versus what does ethics mean for healthcare. You know, bias is honestly not relevant if you are looking at predicting when a factory floor machine might fail. Right? Bias and fairness don't matter.

(Beena Ammanath at 00:09:51) What matters there is the robustness, reliability, and security of AI solutions that you put on. So that's where Trustworthy AI is really addressing that third stream of all the consequences of the technology beyond value creation.

(Joel Beasley at 00:10:07) At what point does, or what point in the future do you think that government will get involved with AI policy or ethics?

(Beena Ammanath at 00:10:17) I think they are involved with it already. You know, there's a lot of catch-up that is being, is happening right now. And I think they're already involved. Unfortunately, it is not something that can be figured out in isolation. Right? There needs to be more robust discussions between the technologists and the policymakers and the researchers to really come up with these well-thought-out regulations. I think we cannot ban any technology as is. It has to be applied more from a use case lens or from an industry lens. There's a lot of discussion around facial recognition, for example. And I think part of the challenge of just negating a whole technology is that there might be actual relevant use cases where the technology can operate in an ethical manner.

(Beena Ammanath at 00:11:14) So having those thoughtful discussions and building our policy, which is very nuanced, as opposed to having a broad, overarching regulation. I think we have to get to that next level of detail, and I'm certainly seeing a lot of movement in that space, which makes me very hopeful.

(Joel Beasley at 00:11:35) So is Humans for AI the same thing as Trustworthy AI, or are they separate?

(Beena Ammanath at 00:11:40) They're separate. It's, Humans for AI is a nonprofit that I set up back in 2016. And one of the basic ways we can solve for ethics or bias in AI is by increasing the diversity of the teams. Right? We know there is lack of diversity in tech in general. We know there's lack of diversity in AI. But I also think AI provides an opportunity to solve for including more diverse candidates in AI. Here's why I say it. Basically, when I say diversity, I am thinking of diversity of thoughts. So that includes—

(Joel Beasley at 00:12:26) Yes, yes.

(Beena Ammanath at 00:12:27) Yes, yes, yes. From a different gender, which gets a lot of attention. But people from different genders, races, ethnicities, geographic backgrounds, educational background, cultural backgrounds. You know, the more diversity of thought we can get to AI solutions, the better, the more robust that solution is going to be. Right? So that's why it's called Humans for AI, because we want all humans to be part of the design, development, and deployment of AI.

(Beena Ammanath at 00:12:58) And what our goal is really—and I've built a few data science teams—and what I've noticed is, yes, you absolutely need data scientists. You absolutely need people, experts who have that PhD in machine learning and AI. But you also need UX. You need designers. You need software engineers. You need product managers. You need project managers. You need QA. You need people who are deploying. So you need all these ancillary skill sets to really productize or, you know, put it into production. So why don't we surround this homogeneous group of data scientists with diversity? So that diversity of thought will naturally filter in throughout the process of the AI development. So that's the mission behind Humans for AI. We focus on women and underrepresented minorities and help them prep to fill in these surrounding roles around the data scientists, but being part of the AI team.

(Beena Ammanath at 00:14:05) Does that make sense?

(Joel Beasley at 00:14:06) Yeah. What type of education, what do the education programs look like in practice?

(Beena Ammanath at 00:14:12) It's really basic AI literacy. What's lacking today is there's a big gap between people who understand the core concepts of AI and people who don't. Right? Like, when you read a newspaper article about AI, you know, to fundamentally understand what does machine learning mean, translate it into real language that clicks in your head. So basic AI literacy on what are all these different terminologies that's associated with AI.

(Beena Ammanath at 00:14:40) And the goal is, the first goal is when you read a newspaper article about AI, you understand what they're talking about. And then the next step is really then connecting to this diverse ecosystem. Once you have the basic AI literacy, if you're interested in product management, here is the list of courses to take. So it's really about, the fundamental goal is to drive more AI literacy so that everybody can understand what AI is. Everybody can be part of the discussion.

(Beena Ammanath at 00:15:10) Because if you don't understand, if you think AI is this big concept, this complex concept that only people trained or educated in this understand, then you're left out of the conversation. And I think we need to level that.

(Joel Beasley at 00:15:29) And people fear what they don't understand.

(Beena Ammanath at 00:15:32) Exactly. Exactly. I am a technology optimist. I'm definitely an AI optimist. I think there are amazing things that AI can do for us. But for that, we need all humans to be part of that discussion, to be channeling that discussion. The other thing, Joel, you know, and you've been in this space a while, is I also believe that for AI to reach its true potential, we are going to need these diverse voices. Here's why. You know, a lot of AI product ideas today come from technologists.

(Beena Ammanath at 00:16:06) Right? But for AI to go deep into a domain and really reach its true potential within a domain, you need the people who are working in the domain. You need the domain experts, because they are the ones, once they understand the basic concepts behind AI, they are the ones who will be able to look at their job and say, "Oh, yeah, I think this can be solved by using NLP, because this part of my job is really boring and can be automated."

(Beena Ammanath at 00:16:37) Whereas instead of kind of turning that conversation to be, it's not the data geeks or data scientists who are coming up with the product ideas. I truly think the next wave of AI product ideas will come from the domain experts who understand their domain. So it will be the teachers, lawyers, doctors, the nurses who will be able to help take AI to the next level within their field. So if that's a hypothesis, then let's make sure that these domain experts are as diverse as possible. Here's our opportunity to actually fix for it. Right?

(Joel Beasley at 00:17:18) Yeah. And I'm curious, you're getting me thinking. Is there any education happening in the public school system on AI? Because it has been a long time since I've been in public school, so I don't—and my kids are very young. They're under the age of five. So they're not in, you know, elementary or middle school. But do you know of any education programs in school?

(Beena Ammanath at 00:17:42) Yes. So I have two teenagers, so yeah, and they're very much there, and I keep an eye on it. And I think it is very focused on, you know, building towards becoming a data scientist, or it's very focused on coding. And I personally have, what I've experienced is that you don't necessarily need to be a hands-on coder. You don't need to be a data scientist. But what you fundamentally need is understanding of the concepts of AI. So it is very focused on going towards a data scientist career path and not so much on what if you want to be an AI product manager or an AI designer. You still need to understand the core concepts of AI.

(Beena Ammanath at 00:18:29) So.

(Joel Beasley at 00:18:30) Do you think that would fall into, like, social studies, histories, mathematics? Which teacher would be best suited to describe these basic concepts?

(Beena Ammanath at 00:18:42) You know, that's a great question. I think it could be really any of those. If I could wave a magic wand, I would say that AI literacy should be taught to every student, like universities and then schools. It doesn't necessarily, when we club it under mathematics or science, we're putting a lens on it. Right? Even if we put it under sociology or philosophy, we're putting a lens to it. And that's why I say the educational background has to be diverse as well. Right? So it's basic AI literacy because we use AI in our lives every day. So you need to know what that thing in your pocket is capable of and why it works the way it is working.

(Beena Ammanath at 00:19:29) Yeah.

(Joel Beasley at 00:19:30) I was thinking what would be the best path to educate the general public that aren't necessarily looking to specialize in data science or AI as they go through the schooling system. So that just that they have a general awareness of, hey, this area of study, it exists.

(Beena Ammanath at 00:19:49) Yeah. It exists. And here are all the possible ways you can get engaged.

(Joel Beasley at 00:19:56) That's important.

(Beena Ammanath at 00:19:57) If you don't, that's fine. You know, here's a way to go. If you are more interested on the philosophical aspects, absolutely. We need that input as well, right, from an ethics lens. If you're interested more on the legal side, great. You know, we need to figure out regulations and policies for it. There are so many possible ways you can go down the AI career path because we have all these potential roles, which is not just about coding.

(Joel Beasley at 00:20:24) Yeah. You mentioned legal. I was talking a while back with another Joel. It's, like, the only other Joel that I know really from the podcast. And he's the CTO of a company called ThoughtTrace and they do AI for document sorting and searching. It seems really useful if you're in the legal industry. That's where they had their niche. Right? And I was curious, like, thinking about the legal industry, also thinking about the MRI type industries where you've got these white collar professions. Right? These specialized top level job professions that AI is becoming extremely useful. And at the same time, about a week ago, as I was preparing for this interview, I saw an article, I think, in the New York Times about AI is coming for billing and accounting or something like that. And they were discussing AI replacing white collar jobs. What are your thoughts on that?

(Beena Ammanath at 00:21:21) There will be some role changes. Right? The job description might change for an accountant or even a CFO. Right? Because we are truly in the age of humans with machines. Right? We have to know how to use that machine, the AI, to do our jobs better. Think of AI as a tool in your arsenal of productivity kit that you can tap into, but you need to understand what that tool is capable of. And you need to be able to provide feedback, inputs to make that tool better. And what do you need for that? The AI literacy. I think, you know, the clickbait headlines kind of drive that fear, but it is never as simple as a headline. Right? It is much more nuanced as to what are all the additional doors that will open once you take away some of these monotonous parts of your job. Plus, then you should be training on how to work best, right, with the machines that are there to help you.

(Joel Beasley at 00:22:30) I agree. And I am an optimist as well. I understand that fear spreads really fast and people use it a lot for headlines. But I also have a deep understanding of AI and the people that are making it. And, overall, I'm optimistic. I think that it's largely misunderstood by the general public. And if they could get simplified explanations of it and become comfortable with the understanding of what it's doing today and how it works and how it could progress, I think that would be very useful.

(Beena Ammanath at 00:23:02) Yes. And it'll also help AI grow as well. Right? Because as long as there is fear, there will be some limitations put in place. If there are no regulations or policies or best practices, that will hold us back as well. So how do you get more thoughtful? How do you have these discussions to move the conversation forward and actually put these guardrails in place so that you can innovate faster with AI?

(Joel Beasley at 00:23:33) The other day, I saw a graphic showing going from single-celled organism life all the way to humans and all the different speciations.

(Beena Ammanath at 00:23:43) Yeah.

(Joel Beasley at 00:23:43) And I was like, I want somebody to take that and do AI on that and make it go forward because it ends at humans. Like, let's make, we have really smart AI people. We have this data, all of this DNA data. Who's running the DNA simulation of what we will become?

(Beena Ammanath at 00:23:59) Yeah. Yeah. That's fascinating.

(Joel Beasley at 00:24:02) Yeah. Do you know anybody that's doing that?

(Beena Ammanath at 00:24:05) I haven't seen that. I know, I'm thinking more from a skills perspective. I don't think it will be a data scientist by ourselves. It will have to be, you need to get in historians, you need to get biologists all to be part of this team that thinks through that.

(Joel Beasley at 00:24:24) Right? Because, I mean, it's DNA. And we have DNA samples all the way back, and we can see how the DNA changes. It'd be interesting. It's one of the things I love about technology is it shows us how much we don't know. If you do research on how much we know about the brain and consciousness, it's way less than what we want. If you do research for how much of the DNA is decoded and understood, it's way less than what we want. So I'm actually really excited about these tools coming in and getting to work with them. And also, one of the reasons why I'm not scared is back about ten years ago, I did a software and the result of this software was it reduced an accounting team, the size of an accounting team. Let's say, for conversational purposes, like, from 30 people to two people. Okay? And it was my job as the software builder to go around and do some implementations within organizations. And inevitably, every time we would go do this, we would go into the accounting team and say, hey, this is new software. It's coming. There would be two groups of people, two types of people. The people that hated the change, which was over 80% of the people and they didn't want to participate and they were, oh, my job's going to change, I'm going to learn something. And then there was the two or three people that were like, oh, this is so cool. This is going to solve so many problems. This is really interesting. Show me how it works. Can it solve this problem this way or that? And I'd be like, oh, it can. Yeah. And it will make this part easier. And they would get all excited. And subsequently, they would learn how to use it through that excitement and wanting to understand it. And they would be the ones that I left running the system, and they would be the ones that would have jobs running that system. And other people would either be reallocated within the organization or let go, whatever it may be. But having, the reason why I share this story is because having that mindset of the curiosity and the continuous learning, that's what keeps you relevant. So we fast forward twenty years in the future, and Elon Musk has Neuralinks in all of us. And we're working with the AI. Like, it'll be the people who are curious, who are interested in moving things forward that remain relevant.

(Beena Ammanath at 00:26:32) Yes. I completely agree. I think it comes back to the point. Right? The technology is evolving so fast. Right? Like, we've talked about lifelong learning in the abstract, but now it has to be, I mean, for you to no matter what role you're in, even for technologists like you and me, if you don't adopt to becoming lifelong learners, we'll get outdated. Right? It's not just the accountants. It's also the technologists. I mean, you have to have that mindset of lifelong learning. And I've looked at that every time I hire a team member is to look at not just what your current technical chops are, but also how likely are you to learn and stay on top of the technology advances that are happening in this space? So curiosity, I agree, is one of those big traits that's going to keep us all going and thriving in this era that we've entered now.

(Joel Beasley at 00:27:33) In your role, are you writing? Can I follow your content? I mean, people ask me all the time, CTOs, CIOs, VPs of engineering, we're all in this competitive business. Right? And we're all always watching the landscape. How do we stay up to date on advancements in AI?

(Beena Ammanath at 00:27:52) Absolutely. So I think I'm the only Beena Ammanath in the world. Google me, you'll find all my social. I'm pretty active on LinkedIn and Twitter. Deloitte AI Institute is really about putting out cutting-edge POVs from a very applied AI lens on how AI is used in businesses. So we have a Deloitte AI Institute website. That's a great way to stay plugged in and on Twitter as well, Beena Ammanath.

(Joel Beasley at 00:28:23) Do you go around and give talks to kids, high schools, college students?

(Beena Ammanath at 00:28:29) I do a lot of keynotes and panel presentations. Now it's all, you know, standing right here at my desk and going around virtually. But, yes, I do, both for Deloitte and also for Humans for AI. We actually had a session with Girl Scouts where we were talking about having AI-related badges to get them introduced to the concepts behind AI very early on. So, yes, I do do that quite often.

(Joel Beasley at 00:29:01) There you go. I love that. That's way better than social studies. Right? You could go to a Boys and Girls Club, like after-school programs, places that have national presence, the Girl Scouts, Boy Scouts. That is so smart. I love that. Who came, did you come up with that, the badges?

(Beena Ammanath at 00:29:17) The idea is there. It is, I don't think it's finalized yet. We also do a lot with universities. So there is an Alliance for Inclusive AI, which is currently with UC Berkeley, but we're planning to expand it to universities. It's basically a foundation that gives scholarships to women and underrepresented minorities to study AI in that university. But here is one which gives me goosebumps to this day. So we did our first cohort before COVID hit. So it was an in-person one at the campus. And we were really targeting people who would otherwise not ever learn about AI. And we worked with this group and part of that 40-member team, there were seven of them who were from a human trafficking rescue center because that nonprofit actually focuses on getting them to high school education. And what we did was to get them literate on AI as they start their new life. Right? Which for me was very humbling and really made me happy to see us taking AI to a demographic which otherwise would never even learn about it.

(Joel Beasley at 00:30:34) Yes. And if everybody just did a little part like that and found it in their own way, then this world would be a lot better place. So you're setting a fantastic example.

(Beena Ammanath at 00:30:45) Thank you, Joel. Absolutely.

(Joel Beasley at 00:30:47) When you're speaking, you get to do a lot of talks. Right? And you get to do a lot of Q&A. But you will also pick up as, let's pretend you're an AI. You'll pick up on trends in these question sets that people ask you. What are some common questions people are asking you?

(Beena Ammanath at 00:31:03) Yeah. The most common question that I get, of course, it depends on the demographic, but most common question I get is, how do I become a data scientist? The other one is, which programming language should I be learning so that I can succeed in AI? And I always give the example that, look, when I was studying, I studied COBOL, assembly language. Pascal was my favorite language. None of them exist today. So, you know, it's about understanding the concepts, right, on how a computer works as opposed to going deep into one language, which you should if you're looking at becoming a hands-on developer data scientist. But, you know, you also need to understand the core concepts. But those are two top questions that I get.

(Joel Beasley at 00:31:56) I love that because we have an Intercom chat on our website, on our Modern CTO website. And I would say one of the top three questions we get is how do I become a data scientist? People just come to the website, they're listening to somebody or they found a data scientist that I interviewed and then they ask us that question and it's, I have, like, this templated response, but, yeah, it's super common.

(Beena Ammanath at 00:32:22) Yeah. And I think part of it is, you know, we've created this whole narrative around data scientist being the sexiest job of the century. But really for data science scaling, there's a whole array of roles around it. Right? Data scientist is crucial, but there are data engineers, data designers, and roles that don't even get get mentioned. The ones who do the data labeling, data janitorial work on building those pipelines and making sure clean data comes in, governance, architecture. There are so many roles in a data science team, and it's not just the data scientist. So I think there needs to be an effort to drive more awareness of all the roles that exist in an AI team and the pathways to get there.

(Joel Beasley at 00:33:10) That's true. When I first found out how large the data labeling industry was, I was blown away. They had just massive buildings full of people doing labeling on, well, there's, I think the one I saw specifically was image labeling. They would tag objects in images. And then I got to talk to a couple different people. There's companies that'll need these algorithms trained and the way that they're training them are tons of humans training the algorithms. It's fascinating.

(Beena Ammanath at 00:33:42) Yeah. Yeah. I know. And they don't get as much attention as a data scientist does. So I think it's up to us, people like you and me, to say, hey. Look. It's, you know, beyond data scientists, there are so many other roles in AI. Find a way to get involved.

(Joel Beasley at 00:33:59) Alright. So I'm going to ask you some good questions here. You are, like, you're at the top. Right? You've eaten, slept, breathed AI. You're at Deloitte at the AI Institute. There's definitely some young people listening to this podcast. They want to be like you. Right? They want to know how you got there. They want the tips. They want the tricks. They want the one thing that they have to do. What sort of advice do you have to the next generation?

(Beena Ammanath at 00:34:29) And I get asked this question as well. So, you know, and I truly believe in this. I did not plan my career path. I did not say that this is where I'm going to start as a SQL developer and move to this and this. It was more of what drove my career was always curiosity. If I got too comfortable in a role, if I felt that, oh, you know, I am in a banking environment doing data warehousing. You know, I'm really curious of how they do it in the retail world, or I'm really curious how they do it in health care. How do they use data? Right? So that curiosity has kind of shaped my career and helped me move forward.

(Beena Ammanath at 00:35:11) I refuse to get too comfortable in a role or in a way where, you know, I think comfort actually slows down progress. So it's always about challenging myself, whether it is learning new things or curiosity to explore newer industries. So that has kind of shaped my career until Deloitte. What has happened at Deloitte is, you know, Deloitte works across all the industries. So now I'm like that kid in the candy place where I get to work across all the industries and look at these large, icky problems like ethics and think about how do you actually solve for it.

(Beena Ammanath at 00:35:49) Right? So my advice would be, you know, find something that you're passionate about. Don't go and study computer science because that seems to be what everybody wants you to do. Identify where your passion lies, and it is completely fine if your passion changes over time. Let your curiosity drive you to the next challenge.

(Beena Ammanath at 00:36:12) Keep challenging yourself. Don't get stagnant.

(Joel Beasley at 00:36:15) I love it. I don't think you have a choice either. I think it's just kind of how you're wired because you sound a lot like me.

(Beena Ammanath at 00:36:25) Yes. That's true.

(Joel Beasley at 00:36:26) This is great. This is great. What are you really excited about? What's the thing happening that you're allowed to talk about? Like, the project or the technology, what's something that's getting you up out of bed in the morning?

(Beena Ammanath at 00:36:37) Oh, there is this cool new thing that I'm working on. It's really looking at technology ethics broadly. So today, I think with AI, we are playing catch up with ethics. We are trying to figure out what is that third stream of ethics and how do we solve for it. But part of my charter at Deloitte is to look at the ethical implications of broader technology.

(Beena Ammanath at 00:37:00) So what does ethics mean for, say, virtual reality and AR? What does ethics mean for quantum? Right? How do we get ahead of the curve and plan for it as the technology itself is also developing? Right?

(Beena Ammanath at 00:37:16) You know, because everybody stays focused on technology development and the value creation from the technology. The part that's keeping me excited is the third part of the ethical implications of that technology.

(Joel Beasley at 00:37:29) And so do you know, by the way, do you know Bill Briggs?

(Beena Ammanath at 00:37:36) Yes. He's awesome. I love him. He is really cool. He and I work closely together. So for example, Bill is focused on quantum compute and the different ways we can use it.

(Beena Ammanath at 00:37:46) And I partner with him to say, "Hey, Bill. Okay. These are the ways. What are the ways this could go wrong? What are the negative impacts?" Forcing him to think through what are the ethical implications.

(Beena Ammanath at 00:37:58) Right? So think of a world, right, we might not be having this conversation in 2D in a few years. Right? You might be, you know, here in a 3D form and, you know, we might be having a lemonade together in 3D form. Right? But, you know, I think fairness would still be an issue in that case. Are we able to project people of color in the same schematic as the others? Right?

(Beena Ammanath at 00:38:27) So, you know, thinking through the ethical implications of technology. And, Joel, the coolest part of it is, you know, there is no search that can give me those results. I really have to go back to my computer science roots and think through the technology itself and the negative implications of that so that we can prevent it and we don't have to play catch up once this technology becomes scaled out broadly.

(Joel Beasley at 00:38:53) Yeah. Because people got pretty upset. There were, you know, issues early on with the facial recognition and the training data. And I'm not extremely well read on it just to be transparent. I looked into it here and there.

(Joel Beasley at 00:39:06) There seems to be a lot. The thoughts seem to be advancing farther than I'm keeping up with them. Right? But at first, it seemed like, okay, we're just building something and then, uh-oh, this happened. Now let's go fix it.

(Joel Beasley at 00:39:18) But then, you know, because we like to build cool stuff as technologists and then we realize, oh, that shouldn't be happening. Let's go change this. And then there's always people that'll say, oh, it was intentional or whatever it will be. So being able to predict what might go wrong from our past experiences and learn, is that something you guys are doing?

(Beena Ammanath at 00:39:35) Yes. That's the entire focus of my time is to really think through, how do we prevent it from going wrong? What, you know, then that means thinking ahead of the ways it could go wrong. So it's really something that's very exciting, and it's cool because it's beyond, you know, the obvious good things technology can do. But thinking about what are the bad things this could do and how do we prevent it as we scale this out.

(Joel Beasley at 00:40:09) Who are some of the other colleagues that you get to collaborate with? You get to collaborate with Bill and who else?

(Beena Ammanath at 00:40:15) So in my role, I work across all our businesses of audit, tax, advisory, and our consulting arm. So I get to work across all of our businesses and all of our industries too because what I'm looking at is ethics as applicable to all of them and then also being able to dive into each one of those. And even on the Deloitte AI Institute, it is applied AI for finance, applied AI for health care. So being able to look at the AI applications and use cases by each industry. And then also looking at how does AI impact audit, how does AI impact tax.

(Beena Ammanath at 00:40:56) So being able to dive deeper into each one of those. So I get to work with pretty much all the leaders within Deloitte.

(Joel Beasley at 00:41:05) That's excellent. So you get to work with a bunch of great people.

(Beena Ammanath at 00:41:08) And I get to learn about all their businesses, which is—

(Joel Beasley at 00:41:11) Come on.

(Beena Ammanath at 00:41:12) Very exciting for me.

(Joel Beasley at 00:41:14) See, that's why I love this job. Before, I built software, and I had different projects all the time, and so I got to learn about finance and fitness and real estate. And then this provided that, but on steroids. Now it's, like, multiple times a week I get new industries, new problems, people storing data inside of DNA and planting things in people's brains.

(Joel Beasley at 00:41:35) It's crazy what's going on out there.

(Beena Ammanath at 00:41:37) Yes. But that's why it's the need for an ecosystem, Joel. I truly believe, you know, it is impossible to have, like, one company leading AI in every aspect. Right?

(Joel Beasley at 00:41:52) I agree.

(Beena Ammanath at 00:41:53) They'll have, you know, several leaders emerging, and they will be in specific domains because AI itself is so domain specific. So, you know, that need for an ecosystem is more relevant than any of the past technologies we've seen.

(Joel Beasley at 00:42:09) One thing that I did see in my show prep is that you guys recently conducted a survey, the State of AI in the Enterprise Report. What were some of the key takeaways?

(Beena Ammanath at 00:42:19) Oh, here was one that, you know, was actually surprising, but at the same time, I was very hopeful after seeing that number. So almost all, 95% of the companies that are ahead in their journey with AI, 95% of them expressed concerns around ethical risks for their AI initiatives. So there's been a lot of talk about AI ethics. Right? But it's been more in the philosophical debate phase.

(Beena Ammanath at 00:42:51) It was very hopeful for me, an opportunity that drove optimism that most companies are thinking about ethics and it's top of mind because that means now we can actually solve for it. We can operationalize it. So that was a great statistic that just stood out for me. But, you know, in general, you know, we are seeing companies investing more into AI compared to prior years. I would say 71% of them were planning to increase their investment over the next fiscal year, and 74% of them were also looking at integrating AI adoption. AI adoption is growing.

(Beena Ammanath at 00:43:43) Ethics conversations are coming upfront so we can get to the solutions. So I feel very optimistic about where we are heading with AI.

(Joel Beasley at 00:43:53) Do you follow any of the projects like OpenAI or, you know, what Elon Musk was helping fund?

(Beena Ammanath at 00:44:00) We do. We are very plugged into the AI ecosystem because we have to be aware of what different organizations are doing, where. Where are opportunities to collaborate? How does it fit in with specific industries? Where is it relevant?

(Beena Ammanath at 00:44:16) And having a point of view on the advances happening in there, again, from an applied AI lens.

(Joel Beasley at 00:44:24) And you mentioned, like, jet engines as an example. Have you seen any other, like, really cool examples?

(Beena Ammanath at 00:44:32) Oh, yeah. There are so many. Maybe I will stick to the two things that actually I wanted to say earlier. Let me ask you a question, Joel.

(Joel Beasley at 00:44:43) Okay.

(Beena Ammanath at 00:44:43) What are the ethics implications for a jet engine? What are the, what are some of the negative consequences? What are some of the ethical implications one should consider when you are building an AI solution to predict jet engine failure?

(Joel Beasley at 00:45:02) I don't know. I mean, are we talking about, like, socially ethical? Like—

(Beena Ammanath at 00:45:06) I'll give you an example because this is something we ran into. When you are looking at the data, the data discovery phase to find the correlations, right, on what might be causing an engine failure so that then you can predict for it. When you look at that historical data, we could actually find out how the pilot was flying the plane. Was he or she flying the plane with, you know, hitting the thrust with the right amount of force? Or was he just hitting it hard, which caused more deterioration? So you could actually see the pilot's behavior in flying the plane. And the conversation came up on, you know, should this be part of his or her performance review?

(Beena Ammanath at 00:45:53) If you are not following guidelines, does that impact your performance review? Right? And there is no regulation for it. We went to FAA, and there's no guidelines around it. But that's the nuances that exist in ethics that you need to think through.

(Beena Ammanath at 00:46:09) It is so much more than bias. Right?

(Joel Beasley at 00:46:13) Yeah. And I think that's one of the misconceptions. We connect ethics and bias, like, immediately.

(Beena Ammanath at 00:46:21) Yes. Yeah. And so help me define ethics.

(Joel Beasley at 00:46:25) Ethics is that third stream of applying and evolving technology in the real world without thinking about all the negative consequences that could still impact you.

(Joel Beasley at 00:46:40) I get it because that pilot having a bad review is a negative consequence.

(Beena Ammanath at 00:46:46) Yes. It could be a health, you know, if you're looking at, you know, it could be a health consequence. Right? Using a screen, using your app too long, there's a health consequence. So thinking about all the negative consequences of using that technology that is being built.

(Joel Beasley at 00:47:03) It's almost like to be really geeky, it's like, where's the deploy hook for thinking about this? Right? Because, like, at what point are we trained? Because I think we're not. Because I'll tell you this much.

(Joel Beasley at 00:47:13) If I were on a project, I would say, you know, I was on the jet engine project. I would try to figure it out and then I'd be like, oh, the behavior of the throttle has to do with this. Great. We'll mark that as an indicator. I wouldn't have thought deeper, like, oh, yeah.

(Joel Beasley at 00:47:27) It's then the behavior of the person that's actually making the throttle change. You're a smart person.

(Beena Ammanath at 00:47:33) And you think about, okay, how do you solve for it. Right? Because, and there when there's no guidelines, obviously, we didn't share that with, you know, so that it would impact. We stayed focused on solving for predicting jet engine failure even though the data told us all the other things that we could do. Right?

(Beena Ammanath at 00:47:51) So making a conscious decision, yes, that, you know, is out of scope. So, you know, we're going to leave it till there are some guidelines around it. But then, you know, using that data to actually provide better pilot training so that you could change that behavior and use gamification to motivate the pilot to fly in a certain way. That's a better solution than going through, you know, the performance review. Thinking through how do you change behavior, how do you provide more impactful training, and use AI to actually drive that training and measure it is a much better way to solve for it.

(Joel Beasley at 00:48:34) This is interesting because we're talking about things that are fairly intangible. They're kind of hard. There's not a framework necessarily where it can give me an exact answer on what I should do.

(Beena Ammanath at 00:48:48) Yes. Absolutely. And that's why you need thoughtful, a lot of thought put into it. And that's why diversity of thought becomes super important because it's impossible for any one of us, no matter how smart we think we might be. You need that diversity of thought to really think through all the negative consequences. I'm sure you heard of the classic case of the robotic vacuum cleaner.

(Beena Ammanath at 00:49:12) Right? That goes on the floor in an automated manner and sucks up dust and hair and, you know? But in South Asia, a lot of people sleep on the floor. So, you know, there have been incidents when somebody's sleeping on the floor, their hair is getting sucked because there were no guardrails.

(Beena Ammanath at 00:49:36) Right? If you had that robotic vacuum cleaner, it was doing that because, you know, you had not thought about that case. So, you know, that is an easy one to relate to, but, you know, that's why you need diversity of thought.

(Joel Beasley at 00:49:51) I wouldn't want to be on the other end of that customer service call.

(Beena Ammanath at 00:49:56) Yeah.

(Joel Beasley at 00:49:57) Oh, this is great. Beena, we did it. We made a podcast. How do you feel?

(Beena Ammanath at 00:50:02) This is great. Thank you, Joel. I really enjoyed our discussion. I mean, you are truly a curious person and, you know, this was great. Loved it.

(Joel Beasley at 00:50:13) Thank you so much for listening. And if you found this episode useful, please share it with a friend or colleague who you think would get value from it. And if you have topics that you'd like to hear discussed on the podcast, either add me on LinkedIn or send me an email, [email protected]. Every time I get an email or LinkedIn message, it absolutely makes my day and inspires me to keep going.