Artificial intelligence (AI) machines will be soon able to sustain a conversation with humans. This is one of the oldest goals in artificial intelligence and soon it might become a reality. Facebook has a chance to be the first company able to achieve this goal.
According to Yann LeCun, the head of Facebook's artificial intelligence lab, the company made progress in revolutionizing artificial intelligence research. After the recent successes in speech recognition and face recognition, now AI researchers are focusing their efforts on deep learning. This field has become a battleground between the high-tech giants, such as Google, Microsoft, IBM and Facebook, in their efforts to bring new AI applications on the consumer market.
Facebook's artificial intelligence research group, called FAIR, has currently employing 50 scientists, but is soon to grow to 100. This is Facebook's first significant investment in fundamental research. LeCun's lab could play a crucial role in company's attempts to reshape consumers' expectations of what computers can do.
Companies such as Microsoft, Google, Facebook and IBM have made significant progress in the field of artificial intelligence the past few years. The reason is that deep learning is an innovative technology, far better than previous AI techniques. Deep learning allows computers to pick up challenging skills, like understanding text content and photos. Deep learning software can make sense of data for itself, without any supplementary programming. There are some AI systems today that can recognize faces or images almost as accurately as humans.
Facebook's AI lab under LeCun's management is aiming now for something even more powerful. The researchers work to develop software with common sense and language skills needed for basic conversation. This will allow users in the near future to communicate naturally with machines Instead of having to enter search terms or click buttons. According to LeCun's comments this will bring a revolution in our "relationship with the digital world." Deep learning software will allow machines to understand our sentences and clarify questions, respond with appropriate answers, or make suggestions of their own.
AI agents that book restaurants for us or answer our factual questions are among the potential applications. Improved online learning and more stimulating video-game characters might also be among the applications of deep learning software.
However, according to LeCun, more provocative applications could be designed, such as AI systems that could get to know their users well enough to recommend what's good for them. AI systems of the near future would be capable to understand not just what their users prefer for entertainment but, more important, what content they need to access "regardless of whether they will enjoy it or not," according to LeCun.Today's virtual assistants like Apple's Siri are not capable of these features since they function based on statistics and matching keywords. "They don't really understand the text", according to LeCun.
Other systems that can perform complex language tasks, like IBM Watson, are super-specialized to a particular format. Unlike today's applications, deep learning software will be able to make sense of language almost in the way humans do. Researchers at Facebook, in competition with those at Google, are developing a software that is able to understanding what words mean. LeCun's research team has designed an AI system capable of logical deduction in reading simple stories and answering questions about them.
