E061 - AI or Not - Zoher Karu and Pamela Isom

Welcome to "AI or Not," the podcast where we explore the intersection of digital transformation and real-world wisdom, hosted by the accomplished Pamela Isom. With over 25 years of experience guiding leaders in corporate, public, and private sectors, Pamela, the CEO and Founder of IsAdvice & Consulting LLC, is a veteran in successfully navigating the complex realms of artificial intelligence, innovation, cyber issues, governance, data management, and ethical decision-making.

AI is moving fast, but most organizations are still stuck at the “more dashboards” stage. We sit down with data analytics and AI executive Zoher Karu to talk about what actually moves the needle: decision intelligence that helps people make the call, not just justify it later. Along the way, Zoher shares a career journey across consulting, retail, finance, and healthcare, and why a business-first mindset beats a technology-first rollout every time.

We dig into a real product example with Taylor.ai, a men’s clothing rental subscription that blends human stylists with algorithms. It’s a surprisingly hard matching problem: understanding a person’s fit, taste, and goals while also understanding the clothing itself and the context around it, from weather to travel. That leads to a bigger takeaway for anyone building AI products: clean feedback loops, clear definitions of “better,” and data that is organized enough to support real decisions.

From there, we challenge a common pattern in enterprise AI strategy: using generative AI to shave minutes off manual work while missing the larger transformation. Zoher argues the bigger prize comes from redesigning the whole process, even asking why the work exists at all. We also get candid about trustworthy AI, security, and why models will always produce an answer even when they are guessing, which raises the stakes for data quality, access controls, and critical thinking.

If you’re leading AI, analytics, or digital transformation, this one is built for you. Subscribe, share it with a teammate, and leave a review with the biggest decision you want data and AI to improve. 

Transcript

[00:00] Pamela Isom: This podcast is for informational purposes only.

[00:26] Personal views and opinions expressed by our podcast guests are their own and not legal advice,

[00:34] neither health, tax, nor professional nor official statements by their organizations.

[00:42] Guest views may not be those of the host.

[00:50] Hello and welcome to AI or not, the podcast where business leaders from around the globe share wisdom and and insights that are needed right now to address issues and guide success in your artificial intelligence,

[01:03] your data and that digital transformation journey.

[01:07] Now today I have a special guest with me and that is Zoher Karu and Zoher is a data analytics and AI executive and I am Pamela Isom, your podcast host.

[01:24] Zoher welcome to AI Or Not.

[01:27] Zoher Karu: Thank you for having me. Pamela. Happy to be here.

[01:30] Pamela Isom: Thank you and thank you for joining us. As I said, you are joining us as we kick off Season three and so we are excited to have you and appreciate you taking the time to share your thoughts and your perspectives with our listeners.

[01:45] To start out with, let's have you tell me more about yourself,

[01:49] your career journey and how that has shaped where you are today.

[01:56] Zoher Karu: Great. Thanks for that.

[01:58] You're right, my career journey has been a little bit of a winding road. My educational background is I did a Bachelor's, Master's and a PhD in electrical engineering and Computer science, but then didn't end up using it directly.

[02:11] Went straight to McKinsey. Did management consulting for several years in a variety of industries and a variety of problems.

[02:17] Got to work as a product manager and lead product management for a company that tracked customers inside of retail stores using cameras or where you walk, where you shop, how long you spend in certain parts of the store.

[02:27] So that was quite fascinating.

[02:29] Got to work for Sergio Zeman briefly. He was the chief Marketing officer at Coca Cola who had left Coke to form his own company and got to work directly for him helping companies grow their brands and marketing strategies.

[02:40] Worked for another small startup analyzing phone conversations and this is large for large language models appeared on the scene. But like what is the subject of the conversation? Are customers upset?

[02:51] What are they upset about? What agents are better than other agents? So having worked in consulting in a bunch of smaller companies,

[02:56] I then moved to bigger enterprises where I was basically leading large data and analytics teams and most recently some of the AI work.

[03:04] So I worked in Sears holdings, the retailer, then ebay, then Citibank, which was actually based in Singapore. So I worked overseas and then Blue Shield of California Health Insurance.

[03:14] So worked in a variety of places all around data and analytics and trying to help businesses maximize the value that they have from their data and make better decisions and drive their business forward.

[03:27] Recently and recently, just after Blue Shield, I've been doing some independent work helping a variety of companies out again along the same themes of trying to maximize the value from their data.

[03:38] Pamela Isom: That's exciting. Wow. Well, congratulations on a successful journey. How has what you've done in the past really helped shape where you are today and where you're headed?

[03:51] Zoher Karu: So look, I think people are of course, recognizing that there is more and more value to the data that they have. They're doing a better job at collecting it. There's more and more sensors that are available everywhere.

[04:03] Like your watch is collecting data now and your cars are collecting data and there's a lot of data in the world and that just continues to explode. But I think the key remains around,

[04:13] are you organizing it well, do you know where it is? How are you going to use this information?

[04:18] And most people, not most, but many people are stuck in just producing reports.

[04:24] Okay, fine, that is one way to look at data, but that is not actually using your data to propel your business forward.

[04:30] And so what I'm trying to do is really take a business focused lens, not a technology focused lens, but a business focused lens. To what problem are you really trying to solve and how can we of course apply technologies to help you solve them?

[04:42] AI and generative AI of course, are brand new technologies that have appeared in the last couple of years to help you get there faster.

[04:49] But at the end of the day, it's just a tool like other tools. You still need to know what problem you're solving and not get caught up in just shiny object syndrome, but really focus on what you're trying to achieve as a business.

[05:01] Pamela Isom: Do you think we're doing a good job of using data to propel our business in general?

[05:07] Zoher Karu: I mean, I think it happens in pockets. Okay, but there are still many, I'll call them legacy companies that are still, I'll call it using data to maybe check their decision, but they're not using data to make their decision.

[05:19] Okay. And the reason that they're not necessarily using it to make their decision is it's hard to access or hard to analyze. Like, okay, should I stop this marketing campaign?

[05:29] Well, I'm not sure.

[05:31] Okay, maybe I gotta ask this other team and wait a few days for the answer, but forget it, I'll just decide on my own. Right? And so it's, you gotta put decision making capability,

[05:41] decision making intelligence in the hands of like everybody in the enterprise. Right? So you need cleaned up data that of course,

[05:48] that everybody operates on. You need to connect those dots across the company. Like, did you connect your website data with your call center data?

[05:55] Connect everything together and give people a way to ask and answer sort of intelligent questions. And some of them are predictive in nature. Like, is my call volume expected to go up next week?

[06:06] Okay, that's a legitimate question. Well, how are you going to get the answer to that question? Right. And looking at a report, it's probably not going to get you that Right.

[06:13] And so how everybody's making decisions on a regular basis in any company.

[06:18] I think the question to ask yourself is, are you making those decisions based off of data and intelligence, or are you making them and then, you know, based off of just your own experience?

[06:28] Pamela Isom: Okay, that's good. So tell me more about this rental subscription algorithmic activity that you have going on. Tell me more.

[06:39] Zoher Karu: Yeah, so one of the companies I'm helping out is called Taylor T A E L O R Taylor AI and it is a men's clothing rental subscription company designed to basically deliver,

[06:55] you know, human stylus slash, algorithmically chosen clothes to you to help you kind of succeed in your life. Right? Like, maybe you're trying to do better in your career, maybe you're out on a date night, maybe you're whatever.

[07:08] But like, clothes are a means to an end.

[07:11] And like, you want to present yourself as more confident or more successful or whatever you. Your reasons are. But like I said, clothes is one way to get there. And we help you get there through clothing.

[07:23] And so we try to understand who you are as a person, of course, your measurements and all of that, but actually recommend clothing to you. And they ship it to your house,

[07:31] you wear it, you ship it back. No laundry, no anything, just ship it back, you get new clothes. Right? And so the hard part about this problem and algorithmically is you're trying to both understand a person,

[07:43] which is its own complexity, right? Like, what's your style? What do you like? Of course, what is your size and things like that? What, what brands do you like? What do you.

[07:51] Why do you need these clothes with the clothes themselves? Like, is this edgy shirt, a conservative shirt? Is it long? Is it short, is it stretchy? Like, everybody who's ever shop for clothes knows that a small is not a small is not a small, right?

[08:03] Or a medium is not a medium. Right? And so how do you both understand the clothing and understand the person and make this match right together? So it's a hard problem.

[08:13] And then you layer in other things like contextual information, like the weather Outside matters. And if you're traveling on vacation that month, then that matters. And there's a whole bunch of things that matter.

[08:22] Pamela Isom: So.

[08:23] Zoher Karu: But it's an interesting problem with, I think, an interesting problem space. It's targeted at men who don't have the time or the interest or the skill, frankly, to really dress themselves, if you will.

[08:34] Right. And it is about not left to your own devices, you might keep wearing the same stuff you wore in college, but like, how do you really push your. Push the boundaries of what, what could look good on you?

[08:45] And like I said, we're using a combination of human stylists and algorithmic approaches to try to try to solve that problem.

[08:51] Pamela Isom: This includes AI and data and any, any unique algorithms or anything you can share there.

[08:58] Zoher Karu: Well, of course, it's using AI and some of the large language models to try to determine, you know, what typically goes with what. Right. Like you wouldn't wear the Hawaiian shirt with the dress pants.

[09:08] Okay, got it. Right. Like bar what. What things typically go together.

[09:12] So it's using AI and of course, a whole bunch of data about you as a person.

[09:17] Also,

[09:17] you can review items and like, I like this shirt. I didn't like this shirt if this was a long, short, tight, whatever. We use that as feedback as well to improve the algorithm and make better recommendations in the, in the future to you.

[09:32] Pamela Isom: What are some lessons learned that you can share with us?

[09:35] Zoher Karu: Look, broadly speaking, not just about this rental problem, right. But lessons learned.

[09:40] Some of the big lessons are it is one is easy to get caught up in shiny object syndrome, if you will, and say, look, we need to do some fancy AI stuff like some vendor presented to me.

[09:52] We got to try that. And you're just doing things for the sake of doing them. And you've lost sight of what problem you're actually trying to solve. And it's always important to keep that front and center.

[10:01] What problem are you actually trying to solve? Because like I said, this is just a tool to help you solve it.

[10:06] There are many tools that you could use, so never lose sight of that.

[10:10] Always. And also another lesson is how will you know it worked? Like, people implement things without any sort of measurement. If you will plan behind, like, okay, we get it, like, you're great.

[10:21] Are things better?

[10:23] I don't know. Are they better? How would you know? Right? And just because you've turn something on or piloted something may or may not mean you made an impact.

[10:31] And so thinking about that, another lesson learned is the data is critical. People often ignore what I'll call The janitorial work of cleaning up their data.

[10:41] And they're like trying to point these powerful tools that data that sometimes there's multiple sources of truth in an organization and then even the AI doesn't know whether is this column called Sales the right one?

[10:53] Or is this column called Sales the right one? Like I don't know which one's the right one. Right. And so if you haven't really figured it out internally,

[11:03] you're just going to point machines at this problem and create confusion. Right. And so that is another, certainly a lesson learned.

[11:11] Pamela Isom: I like it. I appreciate the comments here.

[11:14] So now you and I, we had some interesting dialogue and one of the things that we talked about is missed opportunities with respect to AI and business.

[11:26] I would like you to describe and share with the listeners some of some real and missed opportunities that some of those that we talked about, but.

[11:36] Zoher Karu: And then some the missed opportunities haven't happened yet. So. But I think you. Look, the broad statement I would make about AI is that a lot of people are using it to improve existing business processes.

[11:53] I currently have this business process.

[11:56] It takes me a lot of manual effort to do this process. I wish I had a tool to help me do it, like summarize this legal document or help the agent write notes at the end of a phone call or whatever.

[12:10] I mean, there's a series of things every company does.

[12:13] And you are absolutely right. If there's something that's very manual, there is likely some sort of,

[12:19] I'll call it AI approach that could help you make it faster, better, cheaper. Right.

[12:25] And a lot of places start there. A lot of AI projects are around cost efficiency and trying to remove costs from the system. It's a perfectly reasonable place to start.

[12:35] But that will generate some value. Hopefully if you do it right and scale,

[12:40] actually scale it and actually try to take cost out of the system.

[12:45] If you're not careful, you just add cost to the system and you didn't change anything. Like, okay, change management is a big part of implementing AI.

[12:53] But the real,

[12:54] the bigger value in my opinion is not using AI just to improve existing processes,

[13:00] but it's using AI to rethink your entire process. Right? Like, I mean,

[13:05] just to take an extreme example,

[13:07] yes, I could use AI to help call center agent do their job more effectively by helping them write their notes after the call, for example.

[13:17] Maybe the real question you should be asking yourself is like, why am I even taking a call in the first place?

[13:22] Right? Why? How do I really think about rethinking my company to get rid of my call center or automate it in some way completely. Right. And so that kind of thinking really requires a lot of business transformation and is definitely much harder to execute.

[13:41] The reason it's harder to execute is it typically crosses organizational boundaries. Right.

[13:47] Like,

[13:48] eliminating the call might mean making our website more functional so I can get rid of my call center, but that means the digital person and the web call center person have to coordinate with each other.

[13:58] And so crossing these organizational boundaries in any organization is typically harder than just trying to improve a process within your own organization. Right. But if you can crack the code, I think there is more value to be had in those complete overhauls or redesigns or transformations or whatever word you want to use,

[14:19] rather than incremental efficiency improvements, which will add value, no doubt. Okay. But the bigger prize is rethinking the way you do your business. Right. Like an insurance, car insurance, for example.

[14:31] They used to have to send an agent out, look at your car, you have to take it to a shop and blah, blah, blah.

[14:37] Now you just take a few pictures using your phone, upload it, Boom, got your estimate right there. Right. That's totally different than the way it used to happen. Right. And so really thinking about how do I change the way I operate, I think will yield the most value.

[14:51] Pamela Isom: So that's good. I like it.

[14:54] I appreciate that.

[14:55] I don't think about.

[14:57] Well,

[14:57] maybe I do. I mean, I'm so tired of hearing about how I can use AI to take notes. Oh, my gosh. Like, that was the best example. Because there's so much more to it.

[15:07] There's so much more value there.

[15:09] That.

[15:10] That was a good point. I appreciate that.

[15:12] Zoher Karu: Look, AI is a really, at the end of the day,

[15:15] ultimately is about making decisions. Right. Even this agentic AI kind of thing on the scene now is agents will carry out actions for you.

[15:24] They will make decisions for you and act on it, not just give you information. Right?

[15:30] Yes. Giving you information is still one step of the process, but a human still has to do something with that information. Right.

[15:35] So as we get further and further down this maturity curve, actions will start happening automatically. Right. And so it'll automatically reorder the part. It will automatically change the image on the website.

[15:47] It will automatically do things automatically. Right.

[15:50] So.

[15:50] Pamela Isom: And you're comfortable with that?

[15:52] Zoher Karu: I think I. Yes.

[15:54] You know, we don't want machines running wild and free necessarily. There's, like,

[15:58] human in the loop. There's human watching the loop. There's human out of the loop. Right. And there's a probably A different mix of these things that are going on, right.

[16:05] And certain industries, you have to be more careful in others like healthcare and so forth. Right. But I think, think the technology is improving every day, as we well know, right?

[16:16] And it's getting better and better and better.

[16:19] And does it make mistakes? Sure. Can make mistakes. Do humans make mistakes? Yes. Humans also make mistakes, right?

[16:26] You'll see it in the headlines. Tesla car, just to pick on them for a second. Tesla car on autopilot runs into whatever, XYZ truck, tree, whatever, right? And it makes headlines everywhere.

[16:41] Humans have car accidents every day, like at a factor of like 100,000 more than a Tesla.

[16:46] Nobody does not make a single headline, right?

[16:49] And so there's this expectation that machines are perfect.

[16:53] And I think that is a false or a dangerous expectation. It's never going to be perfect. Perfect.

[17:00] But there are many cases where it's already better than humans, right? Like reading X rays or looking for this or like there's already places where it surpassed humans. I mean look at all these games like chess and go and whatever else.

[17:14] And like, yeah, it's better than a human, right? So I think people are right to be critical,

[17:20] but I don't think that means being so critical that you never trust. Right. I think that's going too far.

[17:28] What I will also say is we as a society are in danger of losing our critical thinking, right? If we just turn everything over to whatever, ChatGPT or your favorite model and just believe what it says all the time without thinking about it.

[17:43] That's not good for society either. Like what? Humans still have to think critically. They still have to imagine new things and question things and explore and all of those things matter.

[17:54] Not just blindly letting machines just do everything for you. And then what are we going to do? Right? And so that is an important skill and I worry a little bit about that, that we are, we don't want to become so dependent that we lose critical thinking skills.

[18:09] The advantage of AI is it can automate a lot of the manual work to free up your time to be able to do that critical thinking as long as you keep doing it right.

[18:18] As opposed to,

[18:19] well, I have nothing to do now. So.

[18:22] Pamela Isom: Yeah, I agree.

[18:24] Yeah. And it's true. And it's so easy to fall into that trap. And it's a trap and it's easy to fall into that space and you're making big time mistakes.

[18:36] Let me just say this. I know somebody.

[18:39] This is your time to talk. But let me just tell you this.

[18:41] I know somebody and we all do. And they.

[18:46] I call it AI etiquette,

[18:48] but. Or digital etiquette.

[18:50] So two things have happened and I'm just like, I'm asking to teach a class on it. I'm about to go teach a class on this.

[18:57] So one is they get AI to write the emails and then they turn around and send that email to you. And you can tell that they haven't read your email, they haven't read their email, they haven't read their responses,

[19:11] they haven't read anything and they turn around and send it back to you. So then when you ask them about it, they can't answer you because they don't realize what they've written.

[19:20] It's just nuts.

[19:22] You don't do that. That just seems like it's not good etiquette.

[19:27] Zoher Karu: Yeah, yeah, yeah. No, you're right. I think it's.

[19:30] In many cases it's assist and augment, but it's not necessarily replace. Right. And so can it summarize the main points from this 100 page legal document? Sure.

[19:41] Does it mean it should go to court for you without you showing up? Okay, maybe not. All right. And so.

[19:47] And we're. We're all collectively still trying to figure it out. Right. What is the right role and what's the right mix of how much automation and how much manual and how much this and how much that.

[19:57] And so we're all learning as we go and it changes daily as we talked about.

[20:03] So it's an interesting time we're living in. No doubt. Like things are changing fast. No doubt. Okay. But the truth is the pace at which they're changing today is actually the slowest it's ever going to change.

[20:14] It's only accelerating. It's not slowing down. Right.

[20:17] Pamela Isom: So true. And you don't want to be left behind,

[20:20] even though we don't know what that is, because you can be caught up today and just totally out of the loop on tomorrow.

[20:26] But the second example that I had was very similar. Like the person gave me some materials to review.

[20:35] But the thing is that as I was reviewing the materials, I could tell that it was not what she intended to give me.

[20:43] And so we have to be really careful. So critical thinking is so important. This is not just etiquette. This is like, okay, so this context is not fitting.

[20:54] Sounds really good.

[20:55] It's not fitting.

[20:57] And then I had to make a decision on what to do with this information.

[21:02] I'm not opposed to universities allowing the use of AI to help fix grammatical errors,

[21:10] to help students come up with a Thesis. I'm not opposed to that because I feel like that's like telling them they can't use a calculator, they're going to do it,

[21:19] so. Or just going to drive shadow AI, Right? So we. I'm not opposed to any of those.

[21:25] I am wanting us to be strategic about how we use the tools and the capabilities.

[21:33] And I want us to be deliberate and intentional about advancing our critical thinking skills. You mentioned earlier.

[21:43] And so I think in school,

[21:46] in universities, in school, on the job,

[21:49] there should be exercises for our development so that we keep our. That requires us to look at how we cultivate our critical thinking skills in the midst of automation. Like, there should just be some intentional activity.

[22:05] So that's my perspective.

[22:06] Zoher Karu: But no, it's like the calculator is a good analogy. Like the fact that you have a calculator saves you from. Multiply this, carry the two, add this. Like, okay, fine, it saved you from all of that.

[22:15] But that just means you can now do more things faster, right? Like it. You would have spent an hour doing this calculation.

[22:23] Now you only spent like 30 seconds on it. So now what are you going to do with the other 59 minutes? Right? Like, and so it's important, right? And in society right now,

[22:32] like let's say the Industrial Revolution years ago, it affected a section of the economy, right?

[22:38] And things went to factories and cars came on the scene or whatever, and it affected a section of the economy, and it took several years to make those changes.

[22:47] What is happening today is that we are in the midst of another revolution, but it's impacting all parts of the economy,

[22:54] like hr, legal, finance, engineering,

[22:58] medicine, whatever, like everything.

[23:01] And it's happening at a speed that we've never seen before. Okay, so humans have never, have not been able to keep up with this change, frankly. Right. And so will jobs get eliminated?

[23:13] Yes. Will new jobs be created? Yes. Okay. But we're in the middle of it right now. Okay? And so it's unlike something that took five, ten years. This is happening like in 18 months.

[23:23] Like what? What's happening? Right.

[23:25] So that's what's going on.

[23:28] Pamela Isom: So let me ask you this. Think about this for a minute. What is the most unique application of AI that you're aware of and why?

[23:41] Zoher Karu: You know, I attended a talk recently about people using AI to help them maximize the yield of a winery.

[23:51] All right? And the way they did that was they had little robot go up and down the vineyards taking pictures of every single vine and the grapes associated with them, the leaves and whatever,

[24:04] and they combined with their location of where they were, they knew exactly how much water or how much fertilizer or how much whatever each individual plant needed to maximize the yield from that plant.

[24:19] And they, for the same acreage, they could boost production significantly just because they understood it and before that wouldn't be possible. Okay. Or having a human look at every vine and study it and write, take some notes and what like let alone applying the different treatments.

[24:38] So I thought it was quite innovative using AI and agriculture, frankly. Right. Like we're not talking about using AI and like so called white collar jobs only. Right. Like AI can be used everywhere.

[24:51] And I thought that was pretty interesting and innovative and demonstrated results. And,

[24:58] and I had some of the wine tasted great.

[25:01] Pamela Isom: That's awesome. When you go back to your example of real opportunities and AI and business, that's a really good example there.

[25:11] Also your rental subscription too, it's a good example.

[25:15] And so because it gets us thinking outside of the normal and thinking about what some of these possibilities are and more. So the reason why we're having this discussion is, is because it's real.

[25:27] In fact, it's really happening. So you're sharing experiences that are real and are happening today. And that's good for us as we are looking at sharpening our business models and strengthening our capabilities.

[25:39] Zoher Karu: Yeah, look, I mean, look, the easiest way to get your head around it, in my opinion, is AI and its various forms are a way to help you make better decisions.

[25:50] And you're making decisions all the time in your professional life, in your personal life.

[25:56] What am I going to cook for dinner tonight? Like, can AI help you? Yes, of course it could help you. Right. It knows what's in your fridge. It knows what?

[26:03] That it could figure it out for you. Right. I could help you make that decision.

[26:08] And so that's the easiest way to think about it is what decision am I making?

[26:14] Could something help me make that decision? Right. And so that, and yes,

[26:19] it can. And I think we're just scratching the surface frankly of everything we could do.

[26:23] Pamela Isom: So. So is there a myth that you want to clarify today?

[26:32] Zoher Karu: Okay, I'm not 100% sure this is a myth, but this I believe. Okay, this is a myth that applying AI is always cheaper than having people do it. I don't think that's quite true.

[26:43] Right.

[26:44] And especially if you think about the worldwide population and their skills and of course their labor rates and things like that,

[26:52] right now people are paying a lot of money for cloud infrastructure and data centers and paying Snowflake and databricks and Azure and everybody and Nvidia and whatnot. Right. And so they're spending lots and lots of money and it's in the headlines all the time.

[27:07] And some of it, of course, rightly so.

[27:09] But they're also having a hard time justifying the ROI of that spend. Like, hey, you just spent $100 million on ABC.

[27:19] Did you generate more than a hundred million dollars of value?

[27:22] Okay. Some of these things are hard to measure. Right. And I think a force fitting AI on top of every single thing may not necessarily be the most cost effective way to do it.

[27:35] Right. And there's more things to consider than just cost. I recognize that. Right. Accuracy and speed and everything else,

[27:42] but we don't need as many people because we're going to use AI.

[27:46] Okay.

[27:47] I think,

[27:48] I think some of it has been carried to an extreme. Right.

[27:51] Pamela Isom: So I like that. Clearly, I like that. Okay, so what are your thoughts on making AI trustworthy? Is it possible?

[28:03] Zoher Karu: Yes, I think it's possible. It also define. Depends on how you define trustworthy. If you mean perfect. Okay, that's a tall order. We talked about that. I think.

[28:11] Look, all of these AI algorithms and generative AI algorithms, they're all at the end of the day, statistical in nature.

[28:18] You ask chatgpt or like I said, any other model Claude, or whatever,

[28:23] a question,

[28:24] it will give you an answer. It's never going to say, I have no idea, I give up. Right. It will definitely give you an answer.

[28:30] The question is whether it's, quote, the right answer. Right. And so, but you have to keep in mind is it's always making its best guess at what the answer is.

[28:40] Always. And it's only. And it's making that guess based on the data that you provided it.

[28:44] And what is actually almost completely magical to me is if you watch sort of anything do text responses,

[28:53] all it's doing behind the scenes, and I programmed neural networks long ago in my undergrad days,

[29:00] all it's doing is predicting the very next word. Like if you type happy,

[29:04] okay. The next word is likely to be birthday. Right. And it's guessing and it's saying birthday. If you ever watch ChatGPT, respond, it looks like a typewriter, like typing across the screen.

[29:12] Right. And it's. All it's doing is predicting the very next word. Of course, it's using a huge amount of information to try to predict that next word. Like your last several conversations, whatever it knows about you, the world going on today, whatever it uses, lots of input.

[29:27] But at the end of the day, that's the only thing it's doing is predicting the very next word.

[29:31] And the fact that it's able to spit out coherent paragraphs and everything else is honestly magical to me. Right.

[29:36] But is it trustworthy?

[29:38] Look, it's as trustworthy as the data it was trained on. You give it a bunch of biased data, it's going to give you biased answers. All right? And so.

[29:46] And you give it incomplete data, it's going to take its best guess. Is it going to be wrong? Okay, it might be wrong. Right.

[29:52] And if you don't have the critical thinking skills,

[29:54] you won't notice it's wrong. Right. Like, there are many examples on The Internet of ChatGPT Getting basic math things wrong, like, add these numbers together. I mean, maybe not that simple, but it gets it wrong.

[30:04] And if you don't know any better, you're like, okay, looks good to me. Right. And so I think,

[30:10] is it trustworthy?

[30:12] Yes.

[30:13] Can it be hacked?

[30:14] Yes. Okay. There's a whole area of, like, prompt engineering and whatever else going on, and people injecting bad things to data lead you down the wrong path.

[30:25] So that's evolving. Security side of the world is also evolving.

[30:28] Data access is also an issue. Like, should everybody be allowed to see all data?

[30:34] Okay, probably not. Like, in the health care world, you know, certain people, like, let's say the clinicians can see your full medical record.

[30:42] Is anybody in the company allowed to look up Pamela's blood pressure? Like, I don't know, should they be able to? Maybe not. Right. And so how do you control who sees what?

[30:53] When you had sort of straight databases and say, okay, you're allowed to see these tables? Not these days. Great. Now you've got this little prompt says ask any question you want.

[31:02] What is Pamela's blood pressure? How does it know to. Okay, don't look over here, just look over here. Right. It's a little harder of a problem.

[31:10] So the problems continue to evolve, honestly,

[31:14] as the technology evolves.

[31:15] Regulation is of course trying to step in, but regulation is always behind technology.

[31:20] And so interesting ties. Right. That we're all living in.

[31:23] Pamela Isom: It's definitely complicated.

[31:25] Trustworthy is overloaded anyway. But then to associate it with that is I think it's good intentions. But I believe that we have to get there as a people at large on what do we mean,

[31:39] and then how do we continuously evolve it?

[31:43] I like to integrate AI security. So I like the example that you just used because we have inferences that are made by these models and we don't always understand what is behind the inferences that it's making,

[31:57] but pulling pieces of information from here and from there.

[32:01] And imagine you forget that the agent has access to B,

[32:06] but it actually is communicating data based on C that it doesn't have access to. So, so how'd that happen? That's what's happening. Right. And so because it's piecing and pattern matching and doing more than that now.

[32:19] And so it's aggregating information from many, many different sources.

[32:24] Zoher Karu: That's where it conscious, it forms its opinions based on more data than you could think about consciously.

[32:30] Pamela Isom: Right, exactly.

[32:31] Zoher Karu: Yep, exactly.

[32:32] Pamela Isom: And that's where I think that's where AI security in my perspective and trustworthiness, if we're going to get there, we're going to drill into that. Right.

[32:42] That information, that data, which is your area, your lane. So that's why you're here, because you know about data.

[32:48] So my question to you is, is there a call to action?

[32:52] And usually I ask, and I'm asking you now, or words of wisdom or both that you would like to leave with the listeners. But before that,

[33:01] is there anything that you wanted to talk about? Anything else that you want to hit me with?

[33:08] Zoher Karu: No, no, I think we talked about a lot of things, from philosophical societal implications to practical data problems and so forth. So I think we, we had a good conversation. I enjoyed it quite a bit.

[33:19] In terms of call to action,

[33:21] I guess from a broad perspective, I would restate some of the things I talked about earlier in terms of things people miss, like don't just chase a shiny object.

[33:32] Think about the problem you're trying to solve and then not just the problem, but how do you know you will have solved it? People skip that step a lot. Right.

[33:41] And so,

[33:43] well, this pilot went live. Great. Is it better?

[33:46] I don't know.

[33:48] Okay, great. So you've got to really think it through and also think through the change management.

[33:54] It's easy to say at the top level, sure, let's do this.

[33:57] But the people who have to change their behavior are somewhere in the middle.

[34:02] And people don't like change. I mean, it's just human nature. They don't like change.

[34:06] So if you don't give them an incentive to change,

[34:09] they're going to resist it. Right. Like why should I change?

[34:12] Right. I don't. It's been working fine for me. I don't need to do anything different. Right. And so because you'll run into brick wall,

[34:20] like head first if you're not careful,

[34:22] if you don't think through the change needed and just because it seems better or faster or cheaper. Like, of course this algorithm is better. Why? I don't understand why nobody's adopting it.

[34:32] Well, that's because you didn't think through the change. Right. And so that's a big piece of it. And then just from a super tactical standpoint, a call to action, I would say, is for all the men or all the women who have men in their lives, Go visit Tailor AI T A E L O R AI and check out the men's clothing rental business and see if it could be a fit for you or at least give

[34:54] it a try.

[34:56] Pamela Isom: Okay.

[34:57] All right. Well, I certainly appreciate you joining me. This is kind of fun. So I appreciate you joining me. And I like all the ideas that you have.

[35:06] So thank you so much for being here.