The Future of Robotics

The Future of Robotics: The AI-Powered Revolution

Will we be the last generation to live without humanoid robots?

Interstellar navigators, protocol droids fluent in six million forms of communication, or bounty hunters hard on a trail, droids played crucial roles in the sci-fi world of Star Wars. What seemed incredibly far-fetched in the 1970’s is now closer. The big-screen magic of Star Wars could become reality, with humanoid robots brimming with common sense and ready to assist us.

Major artificial intelligence (AI) tech firms like Nvidia, Meta, and Alphabet want to bring adaptable learning algorithms, known as foundation models, to robotics. These foundational models are the tools that power chatbots and other applications. Now engineers want to adapt them to allow robots to tackle a wide range of tasks.

AI and Robotics: A Symbiotic Evolution

Robots don’t just benefit from AI, they also help advance it. Researchers hope that by giving AI a physical form, they can move closer to achieving ‘artificial general intelligence’. That’s the term for AI that surpasses human capabilities across a wide range of cognitive tasks. “The last step to true intelligence has to be physical intelligence,” says Akshara Rai, an AI research scientist at Meta.

Robot Challenges

Several hurdles remain on the path to fully autonomous robots. They need vast amounts of the right data, they must overcome hardware limitations, and address safety concerns.

The term ‘robot’ encompasses a wide range of automated devices: industrial robotic arms, self-driving cars, drones, and surgical assistants. Most of these incorporate some form of AI, such as object recognition, but they are typically designed for specific tasks and environments. And they usually require some level of human supervision. 

Even advanced robots like Atlas, from Boston Dynamics, known for its impressive parkour abilities, rely on pre-programmed templates to navigate their environment.

However, learning robots are evolving rapidly. Take the soccer-playing robots developed by DeepMind. These mini-Messi’s use a trial-and-error approach called deep reinforcement learning. That allows them to develop skills, agility, and techniques to improve their play—all at a higher level than could be manually programmed.

The goal for many AI researchers in robotics is to create machines that are more autonomous and adaptable. This vision started with robot arms capable of handling any factory product evolving into humanoid robots that assist the elderly or perform complex tasks around the house.

The Promise of Foundation Models

The current approach uses AI foundation models to control robots. That’s the code that underlies things like ChatGPT. These models use neural networks to learn from vast amounts of data. They can build associations between elements of their training data and generate appropriate actions or responses.

A robot foundation model, trained on text and images from the Internet, learns about objects and their contexts. It also learns from examples of robotic operations, such as videos of trial-and-error robot movements or remotely operated robots. This training allows the model to predict the best actions in new scenarios based on its learned associations.

However, to truly master physical movements, robots need extensive physical data. This requirement poses a significant challenge.

Is the world ready?

The convergence of AI and robotics will bring about profound changes across multiple sectors. They will enhance efficiency, improve quality of life, and address complex challenges. However, while the potential benefits are immense, there will be downside. It will be essential to navigate the ethical, social, and economic implications of these advancements carefully. As AI and robotics continue to evolve, they hold the promise of a future where technology works harmoniously with humanity to create a better world.

In the words of Stephen Hawking, “Success in creating effective AI could be the biggest event in the history of our civilization. Or the worst. We just don’t know. So, we cannot know if we will be infinitely helped by AI, or ignored by it and sidelined, or conceivably destroyed by it.”
For The Good

Michael Nichols

Numbers You Need to Know

10 Billion

Putting aside all of the TV spinoffs, theme parks, video games, toys, and beyond, the live-action Star Wars films are the real Jedi Masters of earnings — grossing more than $10 billion worldwide globally at the box office across four decades and counting. (Entertainment Weekly)

1954

The first modern robot. George Devol began realizing the value of factory automation while working on magnetic recording technology. In 1954, he filed a patent for a robotic arm that could move with six degrees of freedom and store step-by-step digital commands on a drum or other medium. (National Inventors Hall of Fame)

3.4 Million

There are over 3.4 million industrial robots in the world today. The global robot-to-human ratio in the manufacturing industry is 1 to 71. (Exploring Topics)

What’s New in Sustainable Investing

Smart Robot Market to Reach USD 51.2 Billion by 2031, Growing at 16.4% CAGR

According to SkQuest, global Smart Robots Market size was valued at USD 10.3 Billion in 2022 and is poised to grow from USD 13.5 Billion in 2023 to USD 51.2 Billion by 2031, at a CAGR of 16.4% during the forecast period (2024-2031)(PR Newswire)

Robotics investments are gaining speed after post-pandemic slowdown

New numbers out of Crunchbase this week see robotics investments once again trending in a positive direction. The previous two years presented a steady drop in overall numbers, following a record 2021 driven by pandemic-fueled job loss. As we head into the second half of the year, 2024 is on track to beat last year’s numbers (Tech Crunch)

Video Of The Week