Robots Enabled with Artificial Intelligence are Learning to Take Over
Connected, smart robots are doing, learning, and even teaching more tasks and processes than ever, with less programming than ever, thanks to AI.
At a time when one must lament the dearth of human intelligence being applied to some of the most existential challenges humans have ever faced, artificial intelligence is beginning to have an impact on our personal lives.
Loosely defined, artificial intelligence (AI) is a mechanism by which a task that would normally require the use of human intelligence can be completed by a program or algorithm. AI systems may exhibit multiple human intelligence characteristics including planning, learning, logical reasoning, comparing, problem solving, perception, and in advanced systems, limited creativity. In some markets the terms AI and machine learning are being used interchangeably but machine learning is generally recognized as a subset of AI. The algorithms that enable AI are capable of recognizing relationships in a large set of data using a structure that mimics connections between neurons in a human brain. AI can be thought of as a limited form of thinking.
The science of AI has made tremendous progress over the past five years and evolved into multiple levels of capability, including deep learning, machine learning, and convolutional and recurrent neural networks (CRNN). Narrow AI, a type of AI in which a specific algorithm performs a single function, can be used to teach a machine to perform specific tasks without the laborious programming process. General AI more closely mimics human intelligence with a broadly adaptable form of intelligence that can learn a wide range of unrelated tasks. It utilizes accumulated experience and repetition to gain expertise and increase accuracy.
AI is being deployed today in diverse applications that range from analyzing purchasing preferences, chatbot support and facial recognition, to optimizing aircraft design and identifying potential adverse drug reactions. A team at the University of Liverpool is using AI’s ability to identify patterns in data at scale beyond human abilities to discover new useful chemical compounds. The algorithm is fed unlabeled data and infers relationships among the properties of several hundred thousand known materials. Results are ranked by usefulness.
Robotics is an enabler of Industry 4.0 automation that transfers manufacturing processes from traditional human labor to “smart” robotic machines.
Humanoid robots have been with us for years. One of the first was Elektro, displayed at the 1939 New York World’s Fair. Among other questionable talents, he could smoke a cigarette.
Over the years, robotic machines were built to take on dangerous or repetitive tasks that exceed human capabilities. Automated spot welding, heavy material handling, and noxious spray paint operations have long been assigned to automotive assembly robots. These devices require detailed programming and are limited to very specific tasks.
Robots today include a broad class of programmable machines that can carry out a series of actions, semi or fully autonomously. They have evolved in flexibility to enable quick adaptation to new tasks. They range in size from those capable of moving automotive engines to table-top machines with a small robotic arm designed for simple assembly or picking operations.
Early robots were often located in protective enclosures to prevent unintended interactions that could result in injuries to human workers. A newer generation of cobots designed to work collaboratively and in close proximity to humans rely on multiple sensors to ensure safety.
The entire field of robotic devices continues to expand into such diverse categories as search and rescue, checking inventory, meal delivery in hospitals, or bringing an extra towel to a hotel guest.
Facebook is developing the Bombyx, a robot that crawls along power lines to string fiber optic cable.
Amazon’s Astro robot operates in consumer’s homes, where it uses voice commands to initiate tasks. It can take photos, maintain lists, play music, recognize different people, as well as patrol the home for threats when residents are away.
Boston Dynamics has been a leader in developing advanced humanoid robots with remarkable physical skills, including jumping, walking over uneven terrain, and even dancing.
Once seen primarily as a threat to blue collar workers, robots have broken out of the factory floor to disrupt the service sector and are being deployed to address the ongoing shortage of skilled workers. With large numbers of employees choosing to leave the work force, robots are in the position to take over more routine tasks. The current supply chain disruption has highlighted the need to reduce Americas’ dependance on foreign manufacturing. Robots will play a major role as a cost-effective alternative to human labor.
The merger of robotics and AI has spawned the current crop of artificially intelligent robots. Using AI, a robot can “learn” a new task without the labor and time-intensive process of creating detailed programs. In some cases, a database consisting of multiple examples of a desired action is fed into an AI robotic device which quickly learns what is to be done and then replicates that action.
The Orcado Group has become a leader in automation of on-line grocery fulfillment. They built a fully automated distribution system using hundreds of robots that utilize vision systems and sophisticated artificial intelligence that can differentiate between a carton of eggs and a sack of potatoes. The robots pick products of different shapes and sizes and safely deliver them to a human packer who loads them into boxes for shipment.
This is where it gets a little concerning. The process by which AI algorithms are created raises the possibility of faulty or insufficient information as well as the inclusion of unintended biases, making the output difficult to predict with absolute certainty. This may not be a major problem when a robot is flipping hamburgers, but it becomes a significant issue when an autonomous weaponized drone is dispatched to take out a terrorist in a residential neighborhood.
A robotic dog outfitted with a gun and thermal camera has already been shown at a military equipment expo. Current U.S. military policy requires human intervention in such missions, but if left to their own discretion, fully autonomous weapon systems will raise some serious ethical and political concerns.
Industrial and commercial robotic devices are highly complex electromechanical machines that may contain multiple computers, dozens of sensors, power distribution systems, multiple motors, hydraulic pumps, actuators, and connections to a network. They typically consist of many subassemblies which offer prime opportunities for cable assemblies and small, reliable, and rugged interfaces. Connectors in these applications may be exposed to high temperatures, industrial fumes, dust, moisture, and high-voltage spikes as well as excessive shock and vibration. Electromagnetic shielding may be required to protect sensitive signals. Environmental sealing may be required in some applications.
As robotic devices have evolved from relatively few highly specialized devices to broad market applications, leading connector manufacturers have recognized the growth potential. Suppliers such as Amphenol Industrial Products Group, Belden, Bulgin, Fischer Connectors, Harwin, LEMO Connectors, Omnetics Connector Corporation, TE Connectivity, Molex, and ODU USA feature connector and sensor product families applicable to robotic equipment. Suppliers including HARTING, Hirose, Phoenix Contact, Weidmüller, Anderson Power Products, and Hirschmann offer market-specific brochures or websites highlighting interconnects for robotic machines. Smart robotic devices have become a rich new market for power as well as low voltage signal connectors.
Robots interact with the physical world via sensors and actuators. Sensors provide the eyes, ears, and sense of touch that allow a robot to detect, measure, and act in the physical world.
Advanced robotic devices with integrated artificial intelligence continue to become smarter while simplifying the task-learning process, but we are only at an early stage of achieving their full potential. Nvidia recently demonstrated the ability of a robot to use a vision system that simply watches a trainer complete a task one time and can then duplicate the process. This opens the possibility of simple verbal training as well as direct machine-to-machine communication in which robots share information and teach each other.
History has demonstrated that highly complex systems that are not fully understood introduce the potential for unintended results. Just something to think about if you’re considering turning over babysitting responsibilities to a domestic AI robot.