The rise of LAMs (Large Action Models)

Dear Fellow Learners,

Welcome to the first edition of Tech Bytes 2024, the newsletter that keeps you updated on the latest trends and point of view in the world of technology.

In this issue, we will take a look at some of the most exciting and impactful current trend - LAMs (Large Action Models).

The Buzz

Mobile apps are ubiquitous in our daily lives, but are they essential for our convenience? Developing and hosting these apps on platforms like Google Play or Apple’s App Store can be costly and time-consuming. Moreover, most mobile apps rely on APIs to function, which raises the question: can we achieve the same outcomes with API-based interactions alone? Alternatively, a new trend of LAM-based apps, which use a Large Action Model to understand and replicate human actions on various computer interfaces, may challenge the dominance of traditional mobile apps. The recent launch of the r1 device, a custom-built AI gadget that runs on LAM, sparked a heated debate in the tech communities.

At the Consumer Electronic Show (CES) 2024 in Las Vegas, Santa Monica-based AI startup Rabbit has unveiled its latest innovation, the r1, a custom-built AI device offers a natural-language operating system, Rabbit OS, designed to further the online experience by eliminating the need for multiple apps. The Rabbit OS is an operating system powered by the proprietary Large Action Model (LAM), a foundation model developed by Rabbit. For instance, you can link Spotify with the R1 to not only play music using voice commands, but you can also ask the AI to perform contextual tasks like playing other songs from the same album1. You can ask R1 to tell you who wrote the lyrics of the song, who composed the music, who sampled the song, and more. For linked apps like Uber, you can ask the R1 to book you a ride from your office to your home, and it will complete the task in a single step compared to the multiple steps you would otherwise take to book a cab on your own through a smartphone app.

The announcement created buzz in the consumer AI landscape, they keynote was compared with the iPhone launch by Steve Jobs in 2007. Rabbit's r1 device relies on the groundbreaking Large Action Model (LAM) to take human-machine interactions forward.

What is Large Action Model (LAM)

Large Actionable Models (LAMs) are an extension of LLMs, providing them with the capability to proficiently handle various modalities. Just as LLMs made it possible to automate the generation of text, LAMs may soon make it possible to automate entire processes. LAMs are being developed to understand and create content using different types like text, images, and audio. This evolving frontier focuses on enhancing the capacity of generative AI to process and produce diverse outputs, spanning text, images, audio, video, and other modalities.

Comparison between LLMs and LAMs

While we know that foundational models are more general-purpose and less data-intensive, while LLMs are more specialized and data-intensive, LAMs learns by demonstrations.The best model to use for a particular task will depend on the specific task requirements.

LAMs footprints

Research paper from Cornell University shows the potential of LAMs and how it can help with intelligent task automation. Salesforce AI Research also shared their views on actionable generative AI. Few emerging examples on these models can be found on GitHub. After Rabbit Inc. announcement, the LAMs footprints are going to grow with large players expected to join growing demand of automation in consumer landscape.

Future of LAMs

The potential of LAMs is vast. They can automate entire processes, intelligently interact with the world, communicate with people, adapt as circumstances change, and even interact with other LAMs. They can take over repetitive tasks and other busywork, freeing up humans for more meaningful, high-value endeavors.

However, it’s important to note that large language models, including LAMs, require vast amounts of energy to train and run. This has led to concerns about their carbon footprint. AI startup Hugging Face has attempted to estimate the broader carbon footprint of a large language model by considering emissions produced during the model’s whole life cycle rather than just during training. This includes the energy used to train the model, the energy needed to manufacture the hardware, maintain the computing infrastructure, and the energy used to run the model once it had been deployed.

I hope you enjoyed the first edition of Tech Trends 2024. Please subscribe to receive the regular updates on the Tech Trends.

If you have any suggestions to make this newsletter better, just drop me a email

Wish you all a happy new year 🙏