Bengaluru-based startup Kogo AI has announced the imminent launch of its platform, Kogo OS, based on its large action model (LAM).
Raj K. Gopalakrishnan, scheduled for release this week, revealed that Kogo OS will be offered as a platform-as-a-service (PaaS), catering to enterprises and small businesses seeking to develop artificial intelligence (AI) agents.
Large Action Models (LAM) are an extension of the capabilities of large language models (LLM) like GPT (Generative Pre-trained Transformer). While LLMs excel at generating human-like text based on input prompts, LAMs are designed to not only understand language but also perform actions based on that understanding.
This encompasses a wide range of tasks, from answering questions and providing recommendations to more complex actions like composing music, generating code, or controlling virtual characters in games.
LAMs are built upon the same transformer architecture as LLMs, utilizing self-attention mechanisms to understand and process input data. However, they are often trained on a diverse set of tasks and datasets to develop a broader understanding of context and improve their ability to perform various actions.
One key challenge in developing LAMs is ensuring that they can generalize their learned capabilities across different domains and tasks. This requires extensive training on diverse datasets and fine-tuning techniques to adapt the model to specific tasks or domains as needed.
LAMs have significant implications across various fields, including natural language understanding, human-computer interaction, and artificial intelligence. They have the potential to automate many tasks currently performed by humans, ranging from customer service and data entry to creative endeavors like content generation and design.
Overall, LAMs represent a promising direction in AI research, aiming to bridge the gap between natural language understanding and action execution, enabling more versatile and intelligent systems.
Kogo OS has undergone testing in 14 proofs of concept across diverse industries, including travel, mobility, legal, manufacturing, and e-commerce. Gopalakrishnan revealed that eight of these proofs of concept are slated for deployment in production by the end of this quarter, with the remainder to follow by the next quarter.
Founded in 2020, Kogo AI has secured approximately $2.5 million (about Rs 19 crore) in funding from MapMyIndia. Gopalakrishnan anticipates a significant expansion of their business over the next few years, driven by the introduction of the new platform.
Large action models, which extend beyond the content generation capabilities typical of large language models, have been gaining traction globally. This trend is evidenced by the recent launch of US startup Rabbit’s LAM-powered consumer tech device R1 and investments by industry giants like Microsoft and Google in AI agents.
