In a development, that is concurrence with the artificial intelligence boom in the world, Indian entities have jumped into the AI bandwagon. This includes OLA founder's Krutrim and a Reliance-backed chatbot Hanooman.
These developments come in the light of the frequent backlash that mainstream, and fundamentally American chatbots have been facing. The recent of all, includes significant bias discovered in Google's Gemini. OpenAI's good old ChatGPT is also perennially mired in controversies, with the recent episode of ChatGPT generating balderdash as a result of a crash due to a DDoS attack garnering a lot of traction.
Although this could be just be discounted as niggle at the beginning, the fundamental reality of the mad rush behind the development of the technology, along with the immense investment cannot be ignored. In such a situation, the preliminary foibles of the enterprise need to be addressed. After all, these investment should add to the common cogent of evolution, along with pecuniary potential of it. One of the prime aspects of the blemish is the ostensible bias, that has plagued the AI horizon. The data base used for feeding and training these bots and the process used is often under the radar.
This is where homegrown, Indian equivalents come in. Starting with Krutrim AI, Krutrim, an AI startup founded by Ola chairman Bhavish Aggarwal, unveiled an AI chatbot in public beta on February 26. It is supposed to take on the likes of ChatGPT and Gemini.
This launch follows Krutrim's recent disclosure of a USD 50 million financing at a USD 1 billion valuation, making it India's first unicorn of 2024. It boasts the capability to comprehend 22 Indian languages and generate text in 10 Indian languages.
Then we have Hanooman, put together by BharatGPT led by IIT Bombay, Seetha Mahal, with backing from Reliance. This chatbot also claims to be trained in 22 Indian languages.
Language in this case, although an advantage, is not the important aspect of it, in a country with a population of over 1.4 billion, with dozens of language, having a multi-lingual system is fundamental and not exactly exceptional. Beyond these lofty numbers, the bigger question is, as to what these bot can differently from the existing system. And, perhaps, more importantly can these chatbots not commit the same errors, that ChatGPTs and the Geminis of the world have done until now.
The database used to feed and train these bots, again is under scrutiny, as what do they mean and how they can build a system to mitigate bias should be examined. In the rush to build the next big thing for the future, many forget the fuel, that runs them, it is in yesterday, and what is today.
At a time, when the AI era stands to expand and perhaps holds the potential to subsume many aspects of our, it is crucial to build a system, that is capable and extensive, 'in what it knows'. This is also the time, when the risks of the system have been enumerated by many, leading to rise in voices urging for regulation. Under such circumstances, it is crucial to understand the system that is being created, while it is being done, rather than leaving the avenue of deciphering the system for posterity.
