Machine learning is a method by which machines can acquire knowledge and improve themselves without the need for human intervention. Scientists are exploring approaches that would help machines develop their own common sense, similar to how humans learn primarily through observation or trial and error. Supervised learning, which is limited to relatively narrow domains, has been successful in many practical applications, from autonomous cars to language translation. However, scientists at the forefront of artificial intelligence research have once again focused their attention on less supervised methods. Reinforcement learning is a powerful approach that works by setting a goal and allowing the system to work towards it through trial and error until it receives a consistent reward.
Pieter Abbeel, who runs the robot learning laboratory at Berkeley in California, uses reinforcement learning systems that compete against themselves to learn faster using a method called autoplay. Dr. LeCun believes that other forms of machine learning are more critical to general intelligence. After a self-monitoring computer system “watches millions of YouTube videos”, it can extract some representation of the world from them and use it to teach itself when asked to perform a particular task.
Cox, from the Watson AI laboratory at MIT-IBM, works in a similar way but combines more traditional forms of artificial intelligence with deep networks in what his laboratory calls neurosymbolic AI. The goal is to build AI systems that can acquire a basic level of knowledge based on common sense similar to that of humans. Robots will eventually embody artificial intelligence and act freely in the world, but it will take more than just supervised learning to achieve it. Currently, robots can only operate in well-defined environments with little variation. It is using a form of self-supervised learning in which robots explore their environment and accumulate the basic knowledge that Dr.
Abbeel believes will eventually be combined with other methods. Deep learning is a type of machine learning and artificial intelligence (AI) that mimics the way humans obtain certain types of knowledge. It's extremely beneficial for data scientists who are tasked with collecting, analyzing and interpreting large amounts of data; deep learning makes this process faster and easier. Explacable AI (XAI), or interpretable AI, or explicable machine learning (XML), is artificial intelligence (AI) in which humans can understand the decisions or predictions made by AI. Artificial General Intelligence (AGI), also known as strong AI or deep AI, is the idea of a computer with general intelligence that can learn and use its knowledge to solve any problem. Could we ever build machines that are as intelligent as humans? “Of course, there's no question” said Dr. Craig S.
Smith.
