Krutrim Si Designs, popularly known as Krutrim, is gearing up for the debut of its maiden product—a cutting-edge generative artificial intelligence (AI) chatbot app. Bhavish Aggarwal, the founder of the startup, took to X (formerly Twitter) to unveil the launch timeline for this innovative AI solution, offering a glimpse into its functionalities through shared screenshots. The startup asserts that all its AI products will be proficient in comprehending India’s 22 scheduled languages while also generating content in eight languages.
Aggarwal’s social media announcement on February 4th teased the imminent release of the @Krutrim app, accompanied by an assurance of ongoing enhancements to the AI models post-launch. Subsequently, a series of posts showcased the chatbot’s generative prowess, demonstrating its ability to respond to inquiries and provide suggestions.
However, one specific post drew criticism from netizens when the Krutrim AI chatbot’s response to a historical query raised eyebrows. In response to whether India existed as a country before independence, the app asserted that India indeed possessed its own political system, culture, language, religion, and tradition before British rule. Aggarwal’s accompanying caption advocating the need for India’s indigenous AI further fueled the debate, contrasting with the response from OpenAI’s ChatGPT, which provided a nuanced historical context.
The ensuing discussion highlighted concerns over potential instances of AI hallucination and raised observations about UI inconsistencies within the app interface.
Krutrim aims to revolutionize the AI landscape with its family of Language Models (LLMs), comprising Krutrim base and Krutrim Pro, touted as a multimodal foundational model. Leveraging training on over two trillion indic language tokens, the startup promises voice-enabled functionality for its generative AI apps, although specifics regarding this feature’s availability remain undisclosed.
The company’s website showcases benchmark scores of its AI models, positioning them against industry benchmarks such as Meta Llama 2 7B. While Krutrim outperforms Llama 2 7B on several metrics like Arc, Hellaswag, Copa, and Jeopardy, it exhibits comparatively lower performance on benchmarks such as MMLU, Bigbench, and Lambada_open.
As anticipation mounts for Krutrim’s AI chatbot app release, industry watchers await its impact on the burgeoning AI landscape, poised to redefine user interactions and pave the way for enhanced linguistic understanding in digital environments.
Leave a Reply