Responsible & Inclusive AI Innovation in Analytics with Mary Kern and Rumman Chowdhury – Tech Talks

Qlik is an unusual company in that they bring in a lot of outside expertise to help them understand the real world perspective on the, the areas in which they do business. One of the best things about Qlik is that they have this incredible range of outside experts who bring, uh, knowledge and perspective on AI and data into the company. And that's why I'm super excited to, once again, welcome Vermont Choudry from Humane ai and Mary Kern from Qlik to talk a little bit about the international implications of AI and how AI is affecting us all on a daily basis.

So first off, um, thank you for spending a little bit of time with me. Yeah. Thank you so much for having me.

I, I'm really excited to be here with you And thank you for, uh, for, for bringing her along. You're Welcome. I appreciate, I know.

So, um, as I said, I mean, you, you've been focused now for a long time on how, uh, data and AI can serve basically everyone throughout the world, um, underprivileged, the privileged, you know, the first world, the other, the rest of the world. Um, what do you, what is it like in 2025 when you look at the world of AI and the world of, I mean, how data is affecting us all? Well, it's interesting.

2025 is a year of almost contradictions, right? So we do have these hyper powerful centralizing powers such as these big, uh, frontier AI companies that have built these, uh, incredibly expensive models. But at the same time, we're seeing the rise of things like small language models, uh, increased use of open source, more countries and more individuals even being engaged in, in, in the AI narrative.

And, and at Humane Intelligence, we like to work with a wide range of people to really bring them into the AI conversation. So it's almost paradoxical that yes, we have a few companies driving so much innovation, but that there is a such a degree of enthusiasm all over the world. Yeah.

And, and I'm sure that, that you see that as well with, with Click. 'cause I mean, click is a very global company, more so than many other companies, like sort of Silicon Valley centric companies. Yeah.

I think for a lot of our customers, there's a ton of excitement. There's always a ton of excitement. We've got a lot of ai, um, built into our platform, and there's always a lot of excitement about how you can do, like, transform data into decisions.

How can you can do that faster, transform into outcomes. So there's always a sense of how can we do this more effectively? How can we do this more efficiently?

And AI holds a lot of promise. Um, and so there's a lot of excitement, there's a lot of experimentation, but, you know, companies still really struggle with, and how do we do this in a responsible way? How can we like implement it with our systems, with all of our data?

So there's a lot of excitement. There's still some challenges, but, um, yeah, we're boldly moving forward, Aren't we? Well, it does seem like there's a lot of, yeah.

Like you say, optimism. Yeah. But also, frankly, a lot of pessimism and a lot of people that are very concerned about how our data is being used and how it will be used in the world when all these models that you speak of these foundational models that are available in many cases, uh, widely available as open source that could be applied in all sorts of different ways.

Um, you know, what are you hearing from around the world about, uh, ai? Well, you know, interestingly, there was some talk of open source and concern about open source, but the reality is the open source community is the lifeblood of how technology is built. You know, we started a little over 35 years ago with some of the, you know, first connected computers in the earlier days of the internet.

And now we live in a world in which we don't know how to operate without these phones in our hands or without our computers, right? Nearby. And that didn't just happen because big companies made big products.

It actually happened because of the open source community. And I think we're seeing the same thing as it comes to ai. There's a lot of experimentation.

Mm-hmm. Uh, you know, as Mary's mentioning, um, there's a lot of, you know, curiosity, enthusiasm, and really open source is this great democratizing tool by which a lot of people can play with AI models, learn more about them, build their own, et cetera. So I'm, I'm very bullish on the value of open access, uh, better tools and just more democratized learning and understanding of how AI systems work.

Well, speaking of democratized learning, I mean, one of the things I think that, that you've been involved in for a long time is essentially figuring out how data can improve people's lives. And, um, AI is sort of the next step in that process, at least in my perspective. Essentially, you went from, you know, collecting data to figuring out how to extract wisdom from the data set.

And now we have this, uh, new technology that can maybe help us. Are you optimistic of the ways that AI can leverage data to improve people's lives? A absolutely.

And at the end of the day, it's human beings that have to make the decisions on how we're using the technology. AI is a tool and a tool that enables really great things. Um, I think the, one of the most fascinating things, and maybe sometimes again, this is 2025, you have contradictions.

The most fascinating, maybe the most frightening thing is AI is not automation work, it's knowledge work. I think that kind of scares people, but it also opens up an immense amount of possibility. 'cause now we talk quite seriously about the potential for AI tutors or AI based research in science.

Yeah. Yeah. Well, I actually had a conversation with another one of the Click AI council on another episode of this same, uh, series, uh, where we talked exactly about that, how AI can be used to improve science.

Um, do you have the same perspective on this? You know, absolutely. I mean, technology has, oh my goodness, without technology, we would all be farmers.

It has really transformed our lives, our jobs. And I think for ai, what the, you know, the promise is that it really allows us to just be our better selves and to, to do it in a more natural way in terms of interacting with tech. I mean, I've been in analytics for a long time and I love analytics, but the truth is not for everyone, and it's not easy for everyone.

And so being able to lower those barriers of entries make it more accessible for more people to take advantage so they can be, uh, more performant, they can be more efficient, they can do their jobs better. Mm-hmm. Um, I think that's great for all of us.

It's really about being able to harness AI in a way that kind of lifts you up and improves you. Instead of saying like, oh, I'm gonna let AI do my job. It's really about how can it enable you to do it even better in a bigger, better way than what you could do before.

Yeah. That's a really important perspective. And I think that what I'm hearing you say, and I love this idea, is that essentially, uh, especially language models can act as basically a user interface to these tools that we have and these tools that we have had.

That's a much more effective and accessible user interface, not just for sort of people here, but for people everywhere. Because it can meet people in their own language, it can meet them in a way that they're used to interacting, whether if they, if they don't have a lot of experience with the physical devices that we're using, and it can translate that into something that is, that gives them power, gives them power to use these tools. It's, it's kind of like the, the, the evolution from like the command line to the GUI, having a computer that you can talk to in your own language is, it's really powerful.

And I could think that that would bring a real, you know, more accessibility to technology internationally, right? Mm-hmm. Absolutely.

I, I, and you're spot on. The ability for these lang these models to, uh, interact in multiple languages is very, very powerful. Um, I think one of the other places we're going to see a lot of improvement is, uh, accessibility for individuals with disabilities.

Um, you know, we already have like different technology that enable people, uh, who cannot see to be able to navigate spaces. And AI is just going to make a lot of this better now. But again, this all has to be intentionally done.

It has to be inclusively built. These communities need to be part of the conversation. And as Mary's pointing out, analytics is not necessarily for everybody, right?

Um, the ability to interact with and give feedback to how AI models are built needs to meet all of these people where they are not necessarily require everyone to be a programmer. So natural language tools, you know, all of these kinds of interfaces really enable anybody to be part of the conversation. Yeah, that's, that's a really great perspective.

I have, I have a, personally, a lot of, uh, I do a lot of work with this disability community here in the us and indeed it is really amazing how some of these tools have, I don't wanna say unintended because that sounds negative mm-hmm. But, you know, surprising, uh, ability to help people to navigate the world in a way that they could not have before. And I could see that as well happening, you know, opening doors all around the world.

So one of the things as well, that I worry about though is that by expanding access to some of these technologies, it may, um, I don't know, bring a certain, uh, uh, you know, perspectives of one country globally. You know, if, if an AI model is built in, you know, country A and it's used in country B, is it really gonna be appropriate? Are you worried about things like, you know, how AI and data can maybe harm people?

Or, or, or just homogenize the international perspective? That's actually one of the things that we work on at my nonprofit. Um, so Humane Intelligence partnered with the Singapore Infocom Media Development Authority, IMDA.

And we did, uh, across all of aan, we tested multiple different large language models for multilingual and multicultural bias in nine different countries and nine different languages. So we brought in cultural land linguistics and sociologists experts into doing this event that we, we, we do a practice called Red Teaming, where you basically, uh, interact with a piece of software until you break it. Like the goal is to go and break it.

Uh, it's a common practice in cybersecurity. It, it comes from the military, and now we bring it into AI testing, but not just to test for your firewall or whatever, but also to test for things like cultural awareness or cultural biases. And you're right, like, it's very interesting, a lot of these models don't know cultural awareness, but what we're also seeing is a rise of hyper fine tuned models in different countries, in different languages.

Um, so the example I always give is that, uh, in America, the term cricket would probably bring up Jiminy Cricket or Cricket the Insects. However, in any, um, British colony, it would bring up cricket the game. And I think that's quite important in a place like India, the world's largest democracy.

Yes. Right. Where you have hundreds of millions of people who are very technologically capable, right.

And very interested in cricket And, and very interested in cricket. And you wanna make sure they're getting the right cultural reference. And even if you think of things like, if you have it, write a story for your child, like a bedtime story, what sort of story is it writing?

What cultural references is it drawing from? And those are maybe kind of mundane examples, but this matters when we talk about things like jobs hiring. Um, when we test for biases and models, we test for the biases that we understand in an American English context.

We're not necessarily testing for, uh, biases on different past systems, which may happen, for example, in a language journal that's used in an Indian context. Um, so bringing that kind of nuance to testing, uh, is hyper-specific, but it also requires the kind of people who have that cultural knowledge and background. And again, as Mary said, breaking down the barriers to access technology is one of the only ways in which we can make these better models.

Yeah. Well, Mary, I I, I think that it's actually speaks well of Qlik that at least you're bringing, you know, this kind of conversation to the table when you're thinking about this because the tools that that Qlik is building are going to be used globally. Yeah.

And they need to have this kind of perspective For sure. I mean, one thing that's really nice about like is we are a global company. We've got, our product teams are largely based in Europe.

We've got some in the United States, we have some in India. So we do have some diversity that we bring in terms of the overall development, who are we designing for? And I think it's also something that we, when we're designing our products, we try to keep in mind because we do have things like predictive ai, and I, I honestly believe you need to have trust in your data.

You need to have trust in your models as well. And so being able to look at your models operationally to be able to detect if there's bias, you know, um, to be able to make sure they're trained on your data, they're accurate, they're not drifting over time for the large language models. Mm-hmm.

Putting the guardrails in place, right, to prevent maybe answers you don't want to have at an enterprise perspective is really something important, um, to us. And it's critical in terms of how we build our software. This is absolutely incredible.

Um, Mary, thank you to click for inviting us to come along with you and, and making these introductions. It's so valuable, uh, for us to, to get the perspective of people on the AI council and, and, and to learn more about how all of this technology affects the world. Mm-hmm.

And thank you so much as well for bringing this. Where can people follow you, connect with you, and learn more from you? Uh, sure.

I'm mostly on LinkedIn. Um, you can just look me up by name Ramon Chowdry, and my organization is Humane Intelligence. org.

Okay. And, um, where can we follow your work? Hi, well, obviously also on LinkedIn, uh, Mary Kern at Click.

Excellent. com. com.

Thanks for listening in.