New Delhi : While tech giants are pushing platforms like ChatGPT, Bing, and Bard, India’s top engineering colleges are launching a growing number of generative artificial intelligence (AI) research projects, many of which are technology I’m trying to understand how can help create tools like OpenAI. ChatGPT but in Indian language.
New Delhi : While tech giants are pushing platforms like ChatGPT, Bing, and Bard, India’s top engineering colleges are launching a growing number of generative artificial intelligence (AI) research projects, many of which are technology I’m trying to understand how can help create tools like OpenAI. ChatGPT but in Indian language.
Generative AI platforms have been all the rage since late last year, with Microsoft and Google pushing these programs into existing services. Even the Ministry of Electronics and Information Technology (MeitY) said on February 3 that it was “aware” of the emergence and spread of generative AI, saying AI could “dynamically enable” India’s growth. I was.
Generative AI platforms have been all the rage since late last year, with Microsoft and Google pushing these programs into existing services. Even the Ministry of Electronics and Information Technology (MeitY) said on February 3 that it was “aware” of the emergence and spread of generative AI, saying AI could “dynamically enable” India’s growth. I was.
subscribe and read more
However, researchers at research institutions highlight many challenges for generative AI projects in academia. Its greatest challenges lie in procuring sufficient data in Indic, the cost of such projects, and the scale of computational power required. The Indian researcher has been working on projects like this for more than three years.
“In academia, we use language models, or Transformer Architecture techniques, for various tasks such as classifying data, answering questions, machine translation, and building chatbots.” Rourkela, University of Technology (NIT).
Transformer AI models are the underlying algorithms for generative AI tools. After understanding the context, it can process human conversational input and generated output. The global platform works primarily in English, but researchers under Mishra are working on languages such as Hindi, Bangla and Kannada, and have developed a model that can take questions and produce output in these languages. He is creating, says Mishra. In English. They haven’t used OpenAI’s tools for this, but they have achieved a “very good” score according to his industry-standard BiLingual Evaluation Understudy (BLEU) test.
NIT Rourkela scored 25 to 30 points in Hindi to English and 19 points in Bangla to English, he said. For reference, his GPT-4 model in OpenAI with English to French output he has a score of 22.9. Last month, the institute, along with the Association for Computing Machinery, the US science education community that publishes research on natural language processing (NLP), published a research paper on Hindi to English translation.
NIT Rourkela isn’t the only one doing this. Students from the Indian Institute of Technology (IIT) Madras are also working on such projects. Harish Guru Prasad, Assistant Professor of Computer Science Engineering at IIT Madras, says his one such project includes “a better translated YouTube video in Tamil.” .
“Students mainly took up this, comparing their own research language model with GPT-4, and finally published a paper on a new approach to translating videos into Indian languages,” he said.
