We are familiar with AI working with Large Language Models (LLMs). Chat-GPT and the infamous Deep-Seek are LLMs based AIs. These are generative AI which produces texts, images and codes. But generative AI can do more. We are moving towards generative AI working with Large Action Models (LAMs) . Generative AI based on Large Action Models can act on content produced by AI working with LLMs. LAMs produce action oriented and decision making AI. LLMs like Chat-GPT are trained on vast amounts of text data and they generate coherent and contextually relevant language based outputs. Thus, these forms of AI can write emails, generate articles, reports, summarize , translate language, transcribe scripts , assist with coding tasks etc. LLMs work by predicting the next word in a sequence, producing human-like text response arising out of inputs that they receive. They automate and accelerate human tasks that might take hours of human effort. They have revolutionized marketing, customer services , content creation. LLMs are powerful but yet passive. They mostly rely on prompts and do not make any decision independently. This is why LAMs come into play. They are certainly more advance and their output goes beyond language generation. LAMs still use language generation abilities of LLMs but have ability to autonomously execute tasks, make decisions , and adapt to changing environment. This means they can handle a series of tasks seamlessly integrating tools, data sources and decision making logic.
LAMs can chiefly bring about automation of decision making by adapting to changing condition around them. This dynamic responses are the main strength of LAMs. The automation integrates by connecting tools , work-flows and data sources to manage the entire process. the automation brought about by LAMs communicates with other LAMs and bring about coordination of large scale activities. Thus, LLMs for instance in a marketing campaign can write an email but LAMs takes thing further . They gather customer data , customize messages, schedule campaigns , track results and even adjust campaigns based on performance-all autonomously. Generative AI has automated content creation but we need more than that. We need AI to act, execute and decide. LAMs is all set to take our future by storm. LAMs run tasks by themselves and undertake physical actions, control devises, retrieve data, process and manipulate information. This means LAMs can interact with the physical world, converse with people, adapt its working to changing conditions and even work in tandem with other LAMs. LAMs, therefore, become functional collaborator who gets work done in real time.
LAMs have several applications. In the medical field they can diagnose and execute tailored treatment strategies. In the financial field, they can assist in risk measurement, fraud discovery and algorithmic transactions. In automotive sector , they can produce self-driving vehicles and increase travel safety on the roads. LAMs, therefore, are all set to shape the future. They will populate our homes as personal assistants or real time action companion. They allow us to interface with them and our digital needs as well as connect with the physical world via Internet of Things. Thus , for instance, rabbit r1 is a devise based on LAM. It works at hair-splitting speed. It can order food or make appointments with the doctor etc. It has an eye. It is called rabbit eye. It can make video calls . It can look at our emails w and read them for us as well as write email , look into the refrigerator and look for what is available and suggest an recipe that we can cook . We can train rabbit r1 to generate the image or images that we want. Rabbit-ri works as a standalone devise with WiFi and cellular connectivity. It claims are yet to be checked in the domains of privacy, responsibility and sustainability. Indeed , AI is now on a new frontier on the wings of LAMs. We are living in exciting time. Moving from LLMs to LAMs , AI is all set to reboot our life in many ways.
