Ola Unveils ChatGPT Clone, Calls it Krutrim
Indian ride-hailing giant Ola has introduced Krutrim, a new artificial intelligence (AI) solution that bears a striking resemblance to OpenAI’s ChatGPT. Despite the similarities, Ola’s chief Bhavish Aggarwal claims that Krutrim surpasses GPT-4 in various Indic languages. According to Aggarwal, Krutrim AI is trained on a massive dataset of 2 trillion tokens and can comprehend over 20 Indian languages while generating content in 10 different languages.
Aggarwal stressed the importance of developing an AI model that captures India’s rich cultural diversity, stating, Current AI models cannot capture India’s culture, knowledge, and aspirations due to our multicultural, multilingual heritage. He emphasized that Krutrim is not just a superficial addition to an existing API but is built from scratch, with a focus on changing the math and algorithms to make it more relevant for Indian languages.
During the launch event, Aggarwal took a subtle swipe at an undisclosed American AI company, questioning their claims of building AI for humanity and doing so democratically. Ola’s achievement in creating India’s first generative AI model entirely in-house is noteworthy. However, the speed at which Krutrim was developed, especially when compared to its Indian counterparts, raises eyebrows.
While Ola plans to release Krutrim Pro, a multimodal version of the AI solution, in the next quarter, questions remain about the dataset on which Krutrim AI is trained. Aggarwal did not disclose this information during the event. Training a model on multiple languages poses unique challenges, particularly for languages with intricate structures and scripts such as Hindi, Kannada, or Telugu. The financial implications of tokenization disparities further complicate the training and usage costs of AI models.
Meanwhile, Ola’s decision to refrain from disclosing the number of GPUs acquired for training their model and the absence of a published research paper or dataset details add to the skepticism surrounding Krutrim’s development process. Additionally, the team leading this initiative seems undecided about whether to identify themselves as Ola or Krutrim.
Critics argue that Ola’s unveiling of Krutrim appears to be more of a marketing ploy aimed at securing additional funds, as no solid product has been released yet. Furthermore, the demo presented during the launch event left many unconvinced.
Despite these doubts, Ola’s entry into the AI space and their efforts to cater to the Indian market’s linguistic diversity should not be disregarded. As Ola continues to navigate the development and implementation of Krutrim, its impact and effectiveness in revolutionizing the Indian AI landscape remain to be seen.
In conclusion, Ola’s introduction of Krutrim, a ChatGPT clone with a particular focus on Indian languages, has grabbed attention within the industry. While skepticism surrounds the claims made during the launch event, Ola’s endeavors to create an AI model that captures India’s cultural nuances are commendable. As the development progresses and more information becomes available, the true potential of Krutrim will be revealed.
[Word Count: 514]
Disclaimer: The above article is based on the details provided and does not reflect the views and opinions of any specific entity or news agency.
