A century of competition in robotics technology: the awakening of embodied intel

The trajectory of technological development was hard to predict 100 years ago, but looking back today, everything is crystal clear. Any disruptive technology in human history was not achieved overnight. Looking back at today from a century in the future, what will we see? How will today's technology change the world?

The underlying logic of the next era already showed signs in this era. An intelligent entity historian from the year 2100 discovered that for artificial intelligence and robotics, the year 2024 was a significant milestone. The following video may provide you with a complete answer.

Industry: Changing the Rules of Manufacturing

A giant scroll runs parallel with three paths, and many keywords are faintly visible, like a tree. The first keyword to light up is industry. I start from the industry of 1950.

This is the starting point of the robot's contention. The inspiration for the first robot came from Asimov's novel "I, Robot." It was called Unimate, derived from Devol's 1954 patent "Programmable Device for Moving Objects." Unimate was precise in its movements on the assembly line, tireless, and unafraid of high temperatures and pollution.

Advertisement

Subsequently, in 1969, Schiebel at Stanford University invented the Stanford Arm, a standard arm with six axes of motion that pushed the application of robots into a broader field and established the basic form and motion characteristics of future industrial robots.

Afterward, with the advent of microprocessors, force feedback control, and other technologies, robots such as KR FAMULUS and ASEA IRB 6 emerged, and industrial robots achieved standardization, completely changing the rules of production. However, until the first half of the 21st century, it remained just an arm.

In the surging technological river, is there also another technological path about robots that is also exerting force at the same time?

Mobility: Setting Out to Explore the WorldThe tech tree once again guided me back to the starting point, where I will begin with mobile exploration robots.

In the latter half of the 20th century, a robot named Shakey, with a body that was somewhat different, equipped with cameras, rangefinders, and collision sensors, helped it establish vision and spatial cognition, allowing it to pace around the room. All of its judgments were processed by a PEP-10 model computer that took up an entire laboratory space. It turns out that since the 1960s, mobile robots have already begun to explore the world with human-like planning.

This was just the beginning. In the subsequent 1980s, technologies such as sensor fusion and planning intelligent agents emerged. And with the maturation of integrated circuits, in the four decades following Shakey, more and more robots were able to move more easily and precisely, but they were also following artificially preset instructions to accomplish tasks that are relatively difficult for humans.

Such machines were merely more advanced tools, not becoming human. Where exactly is the dormant path leading to intelligence?

The decisive moment of neural networks

Following the tech tree, the focus shifts to the year 1950, where intelligence and neurons meet. An electronic turtle, without instructions, had already begun to explore the world on its own. Neuroscientist Walter built it to understand the workings of the brain, capable of moving towards the light and changing direction upon hitting an obstacle. Walter used the most primitive simulated light sensors and collision sensors to make the electronic turtle exhibit biological-like intelligence.

The tech tree extends to 1987, when Nobel laureate Edelman published Neural Darwinism. The Darwin machine he constructed, controlled by a computing system designed to mimic brain neurons, could predict the environment based on the optic flow signals generated by its own movement, navigating through winding paths without touching the traffic cones on either side. Each step of trial and error feedback in this process iterated to form new neural circuits, advancing from biological intelligence to creating machine behavior that self-acquires experience.

Neural connections link the senses of living beings, and the complexity of neurons determines the level of biological intelligence. The above insights about neurons, under the name of intelligence, will write a new chapter in the history of robotics.

In 2016, AlphaGo defeated the human Go champion, marking a decisive moment for the maturity of deep learning neural networks, like a key that freed artificial intelligence from the shackles of control. Humans finally began to use the term "intelligent agent." An intelligent agent needs to make decisions independently, without external control. In 2024, the intelligent agent was still a brand-new concept, but it had been proposed for nearly a century.The Long-Awaited Triumph of Embodied Intelligence

The technology tree once again takes me back to the starting point of a hundred years ago. This time, it begins with the history of intelligent exploration robots.

As artificial intelligence matures, humans have finally been able to simulate how machines can act like a human. In the subsequent human history, various wonders have been created. Disembodied intelligence took a step ahead. Within computers, it could learn to walk autonomously—by 2017, it had learned to walk from completely random movements. It could also learn to cooperate in teams—during the hide-and-seek task of 2020, four independent disembodied intelligences within a computer developed the ability to communicate and collaborate.

Humans also began to explore the infinite combinations between the nervous system and the physical body. Thus, after a half-century-long game of technological chess involving industrial, mobile, and neural aspects, a convergence finally occurred in the face of artificial intelligence. Embodied intelligence, which had been dormant for nearly a century, welcomed a long-awaited victory.

In this ancient "cat cognition experiment" from 1963, the key to the awakening of embodied intelligence was found. A kitten that could walk actively pulled a kitten that could not walk, making it turn. After ten days, only the cat that had walked itself developed a deep perception of the environment.

This indicates that the most important part of constructing perception is not just sensation and cognition, but experience. A cat without personal experience is like disembodied intelligence in software; only embodied experience can achieve embodied intelligence.

As the secrets of intelligence were unveiled, the Morningstar legged robot GS, developed by Lenovo Group in 2024, appeared before us. It possesses a robust body and precise perception, representing the culmination of control technology that has evolved in the field of robotics to this day.

The Morningstar robot can navigate through complex terrains such as forests and wilderness, as well as through precise and strict factories or energy stations, with precise steps, swiftly and safely, performing inspections, emergency rescues, and field exploration tasks that are beyond human capabilities. This is thanks to the "perception" ability endowed by the visual large model, evolving from relying on 3D semantic maps to mapless navigation with generalized capabilities. The landscapes and paths it has seen and traversed are all experiences, and they are the courage for it to explore unknown scenarios.

AI Embracing the Physical World: Reshaping Productivity

All the previous technological preparations, due to this long-awaited victory, have finally and completely overturned the history of robotics, ushering in a new era of robot awakening.Starting from the perspective of industry efficiency and the essence of application, robots have transcended being mere tools and have become masters of themselves. Robots walk in the real world, entering various industries, introducing new impetus for reshaping productivity, and bringing forth greater value to the tangible economy. In the process of promoting AI towards practical applications, Lenovo Group's Morning Star robot is a pioneer in this revolution.

Throughout the century-long technological evolution, humans have competed in succession, illuminating a complete development map of robotic technology. The technological torrent of AI fundamentally endows robots with intelligence, creating wonders. In the present year of 2024, it has profoundly influenced the new industrial revolution, making the future happen now.