Zoho’s Sridhar Vembu Reveals Plans for Smaller AI Models with 7-20 Billion Parameters
Zoho’s founder, Sridhar Vembu, has announced that the company is currently working on developing smaller artificial intelligence (AI) models with 7-20 billion parameters. The goal is to create models that are specifically designed to solve domain-specific problems for Zoho’s customers. Vembu made this announcement during his speech at the CNBC-TV18 and Moneycontrol Global AI Conclave held on December 16.
According to Vembu, the company believes that smaller models are more effective when it comes to addressing specific challenges within various domains. As a result, Zoho has decided to focus on these smaller models instead of pursuing the more massive 500 billion parameter models. In addition to this, Vembu also shared that the company aims to own its own graphics processing units (GPUs) infrastructure, as it is a cost-effective option in the long run.
Vembu expressed the importance of brain capital and its role in obtaining financial capital. He believes that by solving brain capital problems, financial capital will naturally follow. He also mentioned that in Bengaluru, funding of $50 million to $100 million is not impossible to find. Vembu further added that Zoho is working with AMD, a well-known chip manufacturer, to address the supply chain issues related to chips.
During the conference, Vembu emphasized that a significant portion of software development work is centered around creative tasks, which accounts for approximately 20% of the work. However, a staggering 80% of software development work consists of data bookkeeping, which Vembu believes can be automated and deemed as low-hanging fruit.
Zoho, known as a software-as-a-service (SaaS) giant, recently introduced 13 generative AI extensions and integrations for its applications. These additions are powered by ChatGPT, an AI model that enhances Zoho’s AI capabilities. The company aims to further enhance its AI offerings by integrating third-party intelligence into Zia, their secure cloud-based AI engine.
Notably, Zoho has been working on building its own large language model (LLM) since June 2023. The LLM is similar to OpenAI’s GPT and Google’s PaLM 2 models and focuses on understanding natural language. This project, overseen by Vembu and carried out by Zoho’s R&D team in India, is expected to enable tasks such as language translation, summarization, and question-answering.
Vembu has also been an advocate for the implementation of a comprehensive AI policy in India. In collaboration with Rajiv Kumar, former vice chairman of Niti Aayog, and Sharad Sharma, co-founder of iSPIRT Foundation, Vembu submitted a proposal to the central government in May 2023. The objective of this proposal is to establish an AI policy that prioritizes openness and transparency in the industry.
Zoho’s efforts in developing smaller AI models and owning GPU infrastructure demonstrate their commitment to providing targeted solutions for specific domain problems. With a focus on automation and integrating AI capabilities into their software suite, Zoho aims to stay at the forefront of AI advancements and deliver innovative solutions to their customers worldwide.
