Just consider this, a poor and incapacitated quadruped is being kicked over and over again everywhere on the body, pushed to the ground ruthlessly, and even now the rogue, as appears from a distance, is full of contempt and rage and in no mood of stopping the cruelty, as he picks up a stick and shoves the impoverished creature. But the dog seems to bear incredible strength and shows a great resilience as it manages to get up and stares at his attacker right in latter’s eyes, as if testing his energy level and endurance to such atrocity.
Now dear readers, we are not supposed to get saddened by such event, as it is a robotic dog which is in its testing session in sprawling lawns of Edinburg University and the attacker is none other than Mr. Alex Li, works at Advanced Robotic Lab as Head and is pious and sensible enough to devote free time in animal rights campaigns as well.
Now, this is the unadulterated potential initialized in and unleashed by AI technology that directs the dog and enables it to cope with any of such type of difficult situations, that it may never have been to before, like on stairs, roads that could be slippery, where it is likely to be sent in future, while on rescue mission.
The robotic dog is called Jue-Ying while the way Mr. Li prepares it, fills us with great excitement and interest as the central role played by AI in dealing with the complex circuits arranged thereby. In his mind, he observes robotic dog training as something similar to how kids learn to play soccer and lick up skills such as passing, dribbling and shooting.
Now, on gaining the mastery over such leggy skills, they are promoted to play matches and to display their techniques of handling football in grounds with the goal of scoring more goals and to seal the tournament in their name.
At this point, drawing inspiration from such ways which are but natural to learn skills in any game, tech corporations, in-line with their researchers’ consent, endeavour to implement similar techniques in machines as well, such as, this robotic dog.
In its case, a couple of skill-set is purposefully instructed to him à walking and getting up from falling down as well as trotting and each of such actions has been diligently worked upon in distinct artificial neural network.
However, to define a neural network, such is a bunch of high-class algorithms by which express efforts are made to uncover the basic association among a given data-set, by applying a process that acts on the grounds similar to operations that take place in palatial human brains. As such, by neural networks, we admit the framework of neurons that can either by organic or crafted by man, in essence.
Now, the initial skill-sets are used as basis on which subsequent skill-sets are to be developed and there are a total of 8 neural networks.
If we suppose, such 8 are the players in football team, then, the following necessity would be to appoint, i.e. develop a coach and which is exactly what AI is, as it binds such skills to attempt problem-solving such as standing up when grounded in any difficult positions to continue to walking towards the target.
Certainly, the sheer amazement and the superior objective of the technique, is to let the robotic dog loose in complex environments and to enable him to accomplish different tasks, more difficult than before, such as climbing stairs or walking rocky grounds and it is thereby programmed to make adjustments quickly to get back to basic structure, i.e. up and about and to move towards the goal it is asked in the program.
Other than thins, Mr. Li is also hopeful that such tried and tested method can be utilized and refined further so as to enable robots accomplish a multitude of other tasks that can be of higher difficulty and complex. He discloses, “Of course, locomotion is cool, you can see the robots running around getting kicked in and getting up. But by the end of the day you want the robot to do something useful for you”.
Clearly, such an expectation underlines certain other features, as vision systems and robotic hands, which is simply an addition of complexity to the tune of multiple levels.
As for Mr. Li’s project, such is firmly grounded on research carried out by DeepMind Technologies, which is a branch of Alphabet, dealing strictly in AI and is operational in London.
Neural networks are designed to learn from experience and this technique is known as deep reinforcement learning which DeepMind Technologies are a great stalwarts of.
Exploiting the technique, AI technology was distilled in such a way, it managed to beat chess champions who are humans and went on to become a prominent player at the crowd-pleasing computer game StarCraft.
At this corporation, Raia Hadsell sites as a director and opines the challenging case when AI and movement were mingled.
She uncovers, ”your actions change the world”. Now, we have an AI technology, designed to play chess and become victorious and a robot that is taking care of house chores, clearly, such can be attributed to changing environment.
Still, if such a procedure is managed properly and success is due, the outcome is bound to be tremendous.
In her words, “I think that you will start to see robots being used more with humans in a safe way, because you’ll be able to interact with these robots a little bit more. So they start to be more capable with doing tasks in the home.”
She continues, “But probably more significantly, used in parts of industry, agriculture, construction. Imagine being able to enable a farmer with a robot that has general purpose, and could imitate different types of behaviour.”
Now, returning back to Mr. Li’s robotic dog, it perceives the world on the basis of alerts that motors and joints provide to him, which is termed as a string of inputs pretty straightforward, while the output is equally exclusive, i.e. type of movement dog takes up and pursues the target.
Wrapping up the article, on the other side of the spectrum, at Bristol Robotics Laboratory, we have Mr. Nathan Lepora, who’s also working on AI, in the capacity of a seasoned professor, to enable hand movement but not any animal structure.
AI technology, he plans to offer by him has inherent potential to identify objects, the feature is triggered by artificial sense of touch while the AI experiment is still in its infancy.
He maintains, “The AI opens up much more general ways of learning how to control rather than, if you like, handcrafting simple controllers. That’s the difference. And that’s what the deep reinforcement learning opens up.”
“And deep reinforcement learning also gives the capability to use much more complex sensory inputs as well, for that control.”
Even though, there are voices that claim that refining AI to manage a humanoid robot is an uphill task, needing a large variety of sensors.
Prof Lepora reiterates, “The level of mechanical engineering [involved in] building these robots has kind of gone past our capability to control them, because they’re so complicated. And that’s the problem that’s getting cracked at the moment”.