Krutrim AI to launch chatbot app soon, says founder Bhavesh Agarwal; Preview capabilities

Krutrim Si Designs, otherwise known as Krutrim, will soon launch its first product – a creative Artificial intelligence (AI) chatbot app. The startup’s founder, Bhavesh Agarwal, took to X (formerly known as Twitter) to reveal the launch timeline of its first AI product, and shared screenshots to preview its capabilities. According to the startup, all the AI ​​products launched will be able to understand the 22 scheduled languages ​​of India. The product will also be able to generate content in eight languages.

Agarwal posted on his social media handle on February 4. Posted, “Testing the @Krutrim app before release! Releasing next week. Looks good though we will continue to improve the models even after the launch. Later, in a series of letters, he also revealed his creativity. Chatbot.

While a couple Posts Highlighting that the chatbot is capable of answering basic questions and even offering suggestions, a special Post It received a lot of criticism from netizens.

When asked by the Krutrim AI chatbot if India was a country before independence, the app replied “Yes, India was a country before British rule. It had its own political system, culture, language, religion and tradition.” That’s it. Captioning the reason why we need to build India’s own AI,” Agarwal also posted a reply to OpenAI’s ChatGPT that said, “Before the establishment of the British Raj, the region now known as India There was not a single country known from In the modern sense but a collection of different empires, kingdoms and kingdoms.

Many users responded to the post calling it a case of AI hallucination (a phenomenon where large language models (LLMs) create incorrect or non-existent information) and that ChatGPT’s response was more accurate. Some users also pointed out a UI bug where the ‘send’ icon was not aligned with spaces in the text box.

According to the company, the AI ​​firm has created a family of LLMs that includes Krutrim base and Krutrim Pro, the latter of which is claimed to be a multimodal foundation model. The models are trained on more than two trillion Indic language tokens. The startup has also said that its generative AI apps will be voice-activated, though it’s not known if the feature will be included at launch. On its website, the startup also posted benchmark scores of its AI model, comparing them to the Meta Llama 2 7B model trained on more than 500 billion tokens and a coding assistant. The chart posted shows that Crotram beats the Lama 2 7B on Arch, Hellswag, Koopa, Jeopardy, Pica and other benchmarks. However, it performed worse on the MMLU, Bigbench, and Lambada_open benchmarks.

---