Inclusive AI: Spreading the Algorithmic Wealth to Underserved Communities
In my day-to-day affairs, I navigate between two vastly different worlds. Thus, I’m constantly juggling many roles and responsibilities. On one hand, I work with a successful farmer and entrepreneur from a rural village in India who grows an exceptional crop of turmeric. Despite her expertise, she struggles to sell her product against more prominent brands.
On the other hand, I am immersed in conversations about the potential of ChatGPT and Artificial General Intelligence (AGI) and the role each may have in the current cognitive revolution.
I observe the struggles of a restaurant owner who cannot afford professional photographers to capture the beauty of his dishes for his menu. At the same time , I read about artists who fear that AI will create soulless copies of their work. I also hear from an online clothing retailer who is unsure about how best to promote his business. All the while, GPT-4, OpenAI’s new language model, is passing every professional exam with ease.
As I reflect on the current state of affairs, I am reminded that every technological revolution in history has primarily benefited the elite in the developed nations. From the advent of paper to the printing press and ultimately to the internet, the most privileged individuals have reaped the rewards. However, something has changed in recent times. A vast country like India skipped the entire computer generation and has embraced mobile phones as its primary mode of technology. Indians have demonstrated an exceptional ability to adopt new technologies. We are now at a point in history where the next productivity revolution can be executed in parallel among the developed and underdeveloped populations of the world.
In the bubble of San Francisco, where I live, discussions revolve around whether AI will take over the world or raise us to level 1 on the Kardashev scale of civilization. But people in other parts of the world simply want to get things done affordably. They don’t care about LLMs, Diffusion, Prompt Engineering, or training parameters. They are more interested in applications that can help them in their daily lives.
AI-based tools for the underserved
Through my work with underserved communities, I have discovered that small incremental AI-based applications can make a significant impact on their lives. In 2019, when we began, we discovered that India’s agricultural land plots were not digitized. To address this, we developed OpenCV and machine learning-based solutions that utilized images of village boundaries with plots, subsequently geolocating them with plot boundaries on OpenStreetMap. During the pandemic, we focused on creating a platform that enabled farmers to establish their own independent stores, helping them reach customers while leveraging the latest market prices for guidance. We launched this platform in May 2022 and have consistently grown it by adding more farmer stores.
When the generative AI text-to-image model, Stable Diffusion, was released in August 2022, we recognized the potential for creating automated AI images for labeling and packaging. We promptly developed and launched a solution within the first month following the model’s release.By generating more than 15,000 unique stock images of around 1,000+ commodities, we were able to help farmers create custom-made labels for their products. We also discovered that the product images used for social media and marketplace websites were not professional enough to compete against bigger brands with large marketing budgets. To address this issue, we created a product showcase tool that allows farmers to take a picture with any background. The application creates professional-looking images that it can distribute to gain better consumer attention.
While this tool was initially designed for agriculture customers, it garnered interest from other domains, leading us to develop the Cognify Studio app, which is now used by smaller sellers in retail, e-commerce, and food industries. Our goal during the development process was to keep the design simple, frictionless, and as inexpensive as possible, with the needs of this demographic at the center.
While ChatGPT and its GPT APIs are revolutionary developments in AI, they are mostly used by tech-savvy individuals who understand prompt engineering. To address this, we recently launched Kissan GPT, a voice-based assistant to ask and answer questions in native Indian languages and English. Kissan GPT, a ChatGPT-based AI application, specifically designed to empower Indian farmers, supports 9 Indian languages. Through this innovative approach, the underserved farming community can now access accurate and valuable agricultural information using voice commands, bridging the technological gap and aiding them in making informed decisions.
Kissan GPT has already made a significant impact in its first week since its launch, receiving widespread acclaim from farmers, agricultural professionals, and university professors alike. The application’s remarkable accuracy, language translation capabilities, and prompt responses have garnered admiration across the agricultural domain. We gratefully received some constructive feedback that has served to improve and refine the system further.
This process taught us to have the willingness to listen and learn from those who are not technologically inclined but may greatly benefit from AI in their daily lives. This is not a recipe for instant unicorn startups, but it promises to uplift a vastly underserved community. Our journey of building AI applications for underserved communities has been a humbling and eye-opening experience. We would encourage others to start similar initiatives too.
The views expressed in this article are the author’s own and do not necessarily reflect Fair Observer’s editorial policy.