The Future Of AI

Whatever is in vogue is likely not the future. Gen AI is likely overhyped right now. In my opinion, the expansion of sensory and motor mediums will likely be what takes AI to the next level. AI is in the tadepole form, and for it to grow, it is going to need to change bodies.

Foundation Models in Robotics: Applications, Challenges, and the Future - Dec 2023

A Survey on Robotics with Foundation Models: toward Embodied AI - Feb 2024

Toward General-Purpose Robots via Foundation Models: A Survey and Meta-Analysis - Dec 2023

RT-2: Vision-Language-Action Models Transfer Web Knowledge to Robotic Control

A Path Towards Autonomous Machine Intelligence


RH20T: A Comprehensive Robotic Dataset for Learning Diverse Skills in One-Shot

Open X-Embodiment: Robotic Learning Datasets and RT-X Models

A Survey on Evaluation of Large Language Models

On the Opportunities and Risks of Foundation Models - Jul 2022

Useful Definitions:

Generalist Agent

A generalist agent refers to an artificial intelligence system or agent designed to perform well across a broad range of tasks or domains. Unlike specialized or narrow AI systems that excel in a specific task, a generalist agent is trained to exhibit competence and adaptability in various scenarios.

Foundation Model

A foundation model refers to a pre-trained model that serves as a starting point for various downstream tasks. These models are typically large, powerful neural networks that have been trained on vast amounts of data to learn general features and patterns. Once trained, a foundation model can be fine-tuned or adapted for specific tasks, allowing for efficient and effective transfer learning.

Embodied Artificial Intelligence

Embodied Artificial Intelligence (AI) refers to a paradigm in AI research and development where intelligence is not confined to a disembodied system but is integrated with a physical body or form. In traditional AI, intelligence is often associated with software algorithms and models, while in embodied AI, the focus is on systems that interact with and perceive the world through a physical entity, such as a robot.

Generalist Agent Positive Transfer learning

Positive transfer learning in the context of a generalist agent refers to the scenario where knowledge or skills learned in one task positively influence the performance of the agent on a different but related task. In other words, the agent can leverage its prior experience to improve its performance on a new and possibly more complex task.

Foundation Model Positive Transfer learning

Positive transfer learning in the context of a foundation model refers to the situation where a pre-trained model, often referred to as a foundation model, demonstrates improved performance on a new task compared to training a model from scratch. The knowledge gained during the pre-training phase positively influences the model’s ability to learn and generalize on a different but related task.