MOHIT MOKAL:
Hi Smita. Thank you for coming here.
SMITA GUPTA:
Hi, Mohit. Thanks for having me. I'm very excited about this.
MOHIT MOKAL:
Same here. Likewise. So, to give a general introduction about Smita, she is a lawyer and technologist who is working as part of OpenNyAI from the Agami team. Well, welcome to this edition of our interview on OpenNyAI. And I just want to get started with a very basic introductory question. Can you tell us what OpenNyAI is and what exactly it does?
SMITA GUPTA:
Sure, thanks, Mohit.
So, OpenNyAI is actually a combination of "open" and "Nyai," and we pronounce it as "n-y-a-i" because we are focusing on open-source AI models and solutions that can empower and enable access to justice for many more citizens than currently have access to it.
The intention of this initiative, which started two years ago, was to bring together a community of lawyers and technological experts so that they can communicate and understand each other's language. We wanted to break the silos between lawyers and technologists, making their processes more efficient. Technologists were building solutions for lawyers without knowing what problems to solve, and lawyers were unaware of how technology could aid them. We noticed this gap between the two communities and wanted to create an ecosystem where they could collaborate, work together, and generate innovative solutions.
Along with community building, we also focus on providing foundational building blocks of technology that are available to everyone. These building blocks lower the barrier for innovation and allow anyone, whether it's a startup, a research group, a law student, or a tech student, to understand the legal space and build upon it. These building blocks are entirely open source, like the entity recognition for Indian judgments. It helps researchers who want to analyze specific aspects of the legal system, such as which sections of the income tax act are most litigated. By distilling information from massive amounts of judgments and cases, researchers can gain powerful insights into the justice system.
OpenNyAI is a collaborative mission that started with several members, including Agami, ThoughtWorks, Ek Step, National Law School Bangalore, and the Rohini Nilankani Philanthropies. Now, we are a rich community of over a hundred organizations.
MOHIT MOKAL:
Wonderful. Thank you. I remember reading about this, and I will mention it in the new blog article on OpenNyAI. Now, coming to the second question, can you give us one or two examples of the work you are doing at OpenNyAI?
SMITA GUPTA:
Certainly. One area we have been focusing on since November last year, after the launch of Chat GPT, is how to leverage generative AI for the legal space. It requires trust and faith from the legal community to accept information generated by AI. Thankfully, the legal industry has a lot of structured data that we can use.
We have been conducting use case analyses to understand the needs and pain points of various stakeholders in the legal industry, such as NGOs, courts, government departments, law firms, and practicing lawyers. We identified the need for enabling conversations in different languages and providing information that can help people access justice and obtain relevant information about government schemes, Aadhar updates, or applying for programs like Aayushman Bharat. We categorized this problem as an information retrieval and extraction challenge and explored the creation of chatbots that can converse in languages like Hindi, Tamil, Telugu, Odia, and other constitutionally recognized languages. For example, a farmer could ask, "Mujhe tractor kharidna hai, mere liye koi scheme hai jo meri madad kar sakti hai?” (I want to buy a tractor, is there a scheme that can help me?)
Right? So once you do that, you receive an answer in voice as well. This is one example of government schemes, which has sort of, you know, taken, caught a lot of people's attention. But what it also allows is, you know, specific wearing on the Domestic Violence Act on Motor Vehicles Act.
So we built some of these demos and reference solutions for different departments for the Delhi High Court, for Karnataka High Court. We are really working on Jiva. So Jiva is the judge's intelligent virtual assistant, which again uses large language models, generative AI capabilities, along with identifying one of the biggest pain points of lawyers, which is a comprehensive place where they can find out the entire law.
If, for instance, I want to understand the Indian penal code, seeing all the variations, various amendments that have come into IPC, seeing that over the years in one place and then being able to instantly get Section 3 read with Section 5, read with Section 10. Right. So just that ability to read laws together is such a crucial one, which takes up a lot of time for judges.
And what we've realized, and we've been speaking with judges from the Karnataka High Court especially, and they articulated that just to find out the law applicable for a given case and to find it and read through it in a courtroom, it takes about five minutes per case.
And if for, for approximately, if they're doing 20 to 30 cases, that's more than an hour per day that they're spending on just information retrieval, right? If you want to put it technically. So these are some of the things that we've been playing around with. How can we aid government departments, government schemes, judges, get quick access to documents, really synthesize legal processes essentially?
MOHIT MOKAL:
Makes sense. I recently, it's just because you mentioned Jugalbandi and how learning like using AI to give access with learning language models or large language models algorithms. I recently read an interview. I saw an interview with him, and he mentioned that you know how AI always has this sort of black box that they're not exactly sure how it does it, but, you know, it does it.
And so they realized that even Google Bard, which was recently launched, when they gave a few prompts in Bengali, even though the Bard was not trained in Bengali, after a few prompts, it completely understood and started translating into and giving outputs in Bengali. And now they're actually, they're not exactly sure how it happened, like how the AI adapted, but now they're sort of working to make Bard available in over a thousand languages.
SMITA GUPTA:
Just something interesting, I read this, I just wanted to let you know. Yeah, no, that's so cool. And you know how that happens. Nobody really knows how generative AI is generating stuff. Right. And I was watching this interview with Sam Altman, the CEO of OpenAI, and one very interesting thing he said was that when you go to school as a kid, right, you're just learning maps and maybe spelling and grammar, and science.
And then the next year you start maybe studying civics and history, and the next year you start studying geography. What happens is that your brain just starts expanding more and more, right? When you start studying math more in detail, you suddenly understand history better, right? Because your cognition increases.
So that's exactly what's really happening with the machine, but we don't really have a way of articulating it as yet. Yeah. And I think a lot of people are afraid of this because historically with technology, it's like you develop something and you know how it's working. Now with this new AI environment, it's like we have a good idea about how it's working, but we can't exactly pinpoint how it works.
And so that is what I think scares a lot of people about, oh, what's the future of AI going to be? You know, I think a good safeguard there is just employing some responsible AI practices and what we are learning. And something I read recently, it was very interesting, is there are more than 20 plus frameworks for responsible AI practices in the world, and none of them talk to each other.
None of them are congruent. One is saying X, Y, Z, the other is saying A, B, C. And as a manufacturer of an AI service or an AI product, I can just rely on either to sort of justify my stance, right? And what they do is they don't really even give you any practical insights into, okay, you failed this framework, you failed this metric.
Now how do I apply some ethical practices into my building, into my making practices? There's nothing available. So I mean, of course, there's a lot of stuff on the internet, but really practical advice on how to apply it, that's really missing. And we are definitely in OpenAI also focusing on why we are building out.
So one of our biggest ambitious projects right now is building out a large language model for Indian legal text in India. And so you have ChatGPT, what if you have a legal GPT? Right? And while we build it out, we are crowdsourcing some data. We are definitely trying to employ all those responsible AI practices and sort of learn from our own mistakes and then put it down to paper and be like, okay, this is how we did it.
Can somebody help us make it better? This is one way of doing, going about it. So I feel like the real value in social entrepreneurs utilizing this technology comes in because they're the ones who will actually tell you where it's failing, how it's failing, and how to make it better.
MOHIT MOKAL:
Hmm. Yeah, definitely. Very interesting. You know, but coming back to the questions we have today, otherwise we might get slightly sidetracked. So the third question that I have is how can people use or benefit from OpenAI?
SMITA GUPTA:
Right. So when it comes to benefiting from it, I think definitely join the community. We are a very open group of people, always willing to get more people in, and we always do some gatherings. We have OpenAI learning circles, which we do very regularly, almost on a monthly, quarterly basis where we try to bring the entire community together to discuss what's new, what are the new developments in the area. So that's a good space to just come in, absorb, and see what's happening with the people involved.
Second is we have something called OpenAI Labs, where, you know, just imagine like a scientific lab in school where you're working alongside your peers, building something up, learning from each other. So just imagine that for AI, for law, and we have a lot of interesting folks: startups, entrepreneurs, researchers, journalists who have really come together and are trying to build stuff out in that way.
We also have an OpenAI Residency coming up in the next couple of months, which will be an in-person gathering of entrepreneurs and innovators working in this area, just really hacking it out. It's like a hackathon, a residency of five days to come and build together. You can find details of that on our LinkedIn, on our website, and if you have any questions, just shoot us an email. We are very happy to figure out how you can contribute to the community and the space.
And when it comes to using OpenAI, all the models that we've built in the last two years are all up. They're all open-sourced and published under the CC BY 4.0 and MIT license. They're available on our GitHub page, which is again, totally open and available on our website. So if you have any questions on how to use them, we are very happy to talk about that as well.
MOHIT MOKAL:
Wonderful. Thank you. I think that's definitely very helpful. Now, in the last question I have, how do you see the future of AI being impacted by OpenAI? Or let's keep it in India.
SMITA GUPTA:
Yeah, that's a big one because what we are doing right now, OpenAI, it's a snowball effect, right? Every single day we talk to somebody new, we meet somebody new, and we try to espouse those principles of openness, collaboration, transparency, and inclusivity. These are the principles we build with, interact with, and engage folks with. So what we are really hoping is that more and newer ideas come out, more innovation happens in the legal and justice space in the country. We have the unlikeliest of stakeholders, people really on the grassroots. We have social entrepreneurs coming in and adopting AI practices to make their workflows more efficient, make their lives smoother. We have solutions coming up in sign language, in tribal languages.
We want solutions coming up in online dispute resolution, in digitization of codes. How can we make those processes easier? So, definitely, I think the future, the next five to ten years, is very bright in that it will really unlock human bandwidth and productivity. And I'm not talking just about OpenAI itself, but the entire ecosystem. The idea of AI really coming in and making processes smoother for lawyers, technologists, and more. So I feel like the future is bright, and it really hinges on a lot of collaboration and community work that we do today.
MOHIT MOKAL:
Wonderful. I definitely see the future of AI growing with the kind of support OpenAI is offering to entrepreneurs. Even the AWS challenge is something very exciting. And I remember at the ODR forum, now I'm also working on something, but I would rather not discuss it in this interview. But yeah, is there something you would like to tell our viewers as closing words before we end this interview?
SMITA GUPTA:
I think that's a great note to end on actually. We are always at OpenAI, always doing some challenge, some hackathon, some residency, some lab, or the other. So we'd love for everybody to always reach out to us. We also have the Jugalbandi API, which we have released on our GitHub as well. And yeah, we are always looking to build more solutions, work with partners, and spread the good word. Hope to meet new people too. Thank you for doing this and for spreading the good word.
MOHIT MOKAL:
Yeah, same here. Thank you, Smita, for taking the time for this interview. Thank you for being here.
SMITA GUPTA:
Thanks for having me. Bye, Mohit.
Comments
Post a Comment