Robot Rundown

Robotic Innovations: From Kitchen Helpers to Universal AI Models – Progress, Promise, and Persistent Challenges

Written by Andrew Ashur | Sep 20, 2024 12:00:00 PM

TL;DR: Google ends its robotics project due to data challenges, Chipotle’s Autocado robot speeds up avocado prep for guacamole, and UC Berkeley's CrossFormer AI model aims to control multiple robot types, but real-world application hurdles remain.

 

1. Inside Google’s 7-Year Mission to Give AI a Robot Body

Google’s seven-year project aimed to equip AI with a robot body capable of operating in real-world environments. Despite progress, including improvements in robot arms learning tasks through end-to-end training, the challenge of gathering enough data for robots to perform complex tasks autonomously proved too great. Large-scale simulations with millions of virtual robots helped, but real-world applications remained elusive, leading to the discontinuation of the Everyday Robots initiative in 2023.

 

Major takeaway: While the project advanced AI and robotics integration, it highlighted how far we still are from creating robots that can function reliably in everyday life.

 

2. Chipotle’s testing an avocado-peeling robot and an automated bowl assembly line

Chipotle is testing a new robot called Autocado to streamline avocado preparation, a key ingredient for its popular guacamole. The robot, currently being tested in California, can cut, core, and peel an avocado in about 26 seconds, significantly reducing the time employees spend on this labor-intensive task. Once the avocados are processed, they are still hand-mashed to maintain the texture customers love. This move is part of Chipotle's broader effort to improve efficiency with collaborative robots (cobots), including another system called the Augmented Makeline, which automates bowl and salad preparation.

These automation efforts aim to help workers focus on higher-value tasks while keeping up with Chipotle's growing demand, especially for online orders. Chipotle's focus on tech innovation, developed through its $100 million venture fund, represents a growing trend in the fast-food industry to leverage robotics for operational efficiency.

Major takeaway: Chipotle's robots like Autocado could help the chain keep up with demand while easing the workload for its employees.

 

3. One AI Model to Rule All Robots - A new model can operate virtually any robot design, including arms, quadrupeds, and drones

Researchers at UC Berkeley and Carnegie Mellon have developed CrossFormer, a general-purpose AI model capable of controlling various robot types, from arms to quadrupeds and drones. Unlike previous models, it handles diverse data without manual adjustments, achieving comparable performance across different robots. This innovation could enable more adaptable robotics systems. However, the model requires external servers due to its size, and further work is needed to demonstrate improvements in real-world tasks and explore "positive transfer" between different robotic tasks.

Major takeaway: A single AI model for all robots is promising but still faces challenges in real-time application.

 

About Lucid Bots: 

Founded in 2018, Lucid Bots Inc. is an AI robotics company that is committed to uplifting humanity by building the world's most productive and responsible robots that can do dangerous and demanding tasks. Headquartered in Charlotte, the company engineers, manufactures, and supports its products domestically, which include the Sherpa, a cleaning drone, and the Lavo Bot, a pressure-washing robot. Lucid Bots' products are elevating safety and efficiency for a growing number of customers around the world. Lucid is a Y Combinator-backed company, with investments from Cubit Capital, Idea Fund Partners, Danu Ventures, and others. Lucid Bots raised $9M in Series A funding in May and was recently recognized as the 4th fastest growing manufacturing company in the United States.