Seniors will surely remember the Spanish group radio of the future, and how Santiago Aucelon and his bandmates proclaimed in 1980 that “the future is already here.” Indeed, that future is already our present day, and robotics technology seems to have a lot to say about it. , are part of our daily lives. In fact, you may be reading this article on your smartphone. face, text, or voice recognition system…artificial intelligence is already entrenched in our daily lives.
IBM’s Watson Computer, American Quiz Winner Danger!, is just another example of this. It’s now a system that uses super-advanced learning technology that’s being trained to help doctors make better decisions.
A constant trend appears to be around improving training algorithms so that computers can learn themselves to eventually behave like humans. This is a concept that is discussed throughout this article. technology reinforcement learningappears: a set of algorithms that allow robots to operate in increasingly autonomous ways. let’s see…
Machine learning systems are getting less and less overseen
How do machines learn? The foundation of this technology is supervised learninga kind of machine learning that relies on A set of examples for which the machine knows the answers in advanceIn other words, it attempts to predict the future based on already well-known behavioral patterns stored in the data history.
of Incredible increase in computing power In the face of much more advanced artificial intelligence training concepts have allowed this learning process to become increasingly precise. deep learningUnlike the former, it requires far less direct human supervision.
Deep learning relies on large-scale simulated neural networks, be able to recognize behavioral patterns in their stored dataThe most famous examples of deep learning in action are Google’s Now, Apple’s Siri, or Microsoft’s Cortana speech recognition system.
Combining deep learning and reinforcement learning could be the key to achieving true human behavior on the part of machines.
Experience as a basis for technical reinforcement learning
So can machines learn on their own? The short answer is yes. The long answer makes us understand what. reinforcement learning When it comes to boosting automation processes for robots and computers, it is possible.
technology reinforcement learning Systems powered by artificial intelligence make decisions based on their own experienceThat is, when faced with a situation, it should be possible to take the optimal action through an interactive trial-and-error process based on positive reinforcement each time a certain goal is achieved.
As such, machines can make decisions through reinforcement learning and successfully perform higher abstract problems without storing a priori knowledge of the environment and occurring variables.
Implementations of this learning process are already capable of recognizing faces, sequencing DNA sequences, driving vehicles, and making medical diagnoses. Current, Big companies like Google, Apple and IBM are investing in research To train robots to perform simple tasks using this technology.
The idea is to allow robots to autoprogram themselves while learning on the fly, and share this learning process with other robots to accelerate the entire procedure. The results obtained are fed to some kind of central server, a large neural network with all the learned patterns, which feeds them back to the robot to create a new learning cycle.
Technology Reinforcement Learning Applied to Driving
It’s the dream of those behind the wheel for hours. A car that can drive itself. Inspired by behavioral psychology, Reinforcement learning is the key to realizing autonomous driving: A vehicle that can enter right-lane roundabouts, enter highways safely, and behave appropriately during tremendous traffic jams.
We’ve been reading the news about self-driving cars for years now, and even though its implementation seems imminent, driverless vehicles can’t help other human drivers, such as intruding into lanes. still hesitant in the face of complex situations involving If you don’t want to get caught in unnecessary risks and traffic, they need to learn more precise driving skillsProper performance when surrounded by many vehicles, etc.
mobile eyeis an Israeli company specializing in providing safety systems to various automotive companies and currently A platform that allows manufacturers to share data collected by their autonomous vehicles, and therefore learns from a whole set of environmental factors, whether humans or self-driving cars. This software enhances the process and is much more efficient than a bunch of programmers trying to decipher all those decisions.
Collaboration as a working methodology between robots and humans
For many disciplines, now is not the time to replace humans with machines, but the time to collaborate and improve.
Canadian startup Kindred AI It teaches machines to perform complex tasks with the help of human ‘pilots’. Those who assist them through virtual reality. Robots attempt to perform tasks such as grabbing objects and achieve desired outcomes through machine learning. But when none of the algorithms provide a solution, the robot seeks human assistance. At this stage, humans utilize virtual reality hardware to visualize the task and temporarily assume control of the action. Thanks to reinforcement learning, robots can learn from what their human assistants are doing and accumulate that experience next time.
Even if the future was already heralded in 1980, its rapid pace is clearly unstoppable today. Machines will be our allies in the ever-approaching future. But ultimately it’s up to us to draw the line how far we want to go.
Fuente: MIT Technology Review, Advanced Tech, Fernando Sancho Caparrini, Clever Data
