SOLVVE continues to discuss the key concepts and events in the field of machine learning and artificial intelligence systems. And today we move on to the second part of the short historical overview of its milestones. In this article, we will take a look at what happened since the 1980s after the first Artificial winter up to the present day. Let’s dive right in.
The 1980s – expert systems and embodied mind
Despite the drastic exhaustion of interest towards artificial intelligence and the lack of funding during the AI winter that has started in the late 1970s, experts in machine learning continued their research.
By the end of the 1970s scientists focused on reasoning. One of the major discoveries was that intelligent behavior required a very extensive and detailed knowledge within a certain topic to demonstrate intelligent behavior. Deeper research within the field gave birth to expert systems, programs that could answer questions, and find solutions to very specific problems within defined domains by following rules for decision making like the human experts would do (remember DENDRAL and MYCIN from the 1970s?).
Knowledge engineering became the key direction in AI for the whole decade. Success in building intelligent systems, even if they supported reasoning within strictly defined field, reignited interest in funding from external sources. Such systems presented great interest to corporations in many industries who willingly provided financial support to researchers. For example, the Japanese government invested about USD 400 million into the Fifth Generation Computer Project from 1982 to 1990 in an attempt to push artificial intelligence forward and revolutionize logic programming.
Moreover, during this decade John Hopfield with the help of David Rumelhart brought “deep learning” techniques into the light. Hopfield proved that replication of a neural network (called the Hopfield net) could help machines learn in a new way. It was a huge step for the first time since the birth of cybernetics machines not only performed commands but became capable of learning from their own experience just like people do.
Nevertheless, the idea of learning from experience gave birth to a new debate: can AI learn like humans do if they do not experience the world the way living creatures do? Some researchers argued that artificial intelligence needs a body in order to live and interact with the world like humans. Only then they will be able to think, learn and reason like people do because only then artificial intelligence will be able to show commonsense reasoning. To understand the world abstract symbols are not enough. One must also live the world through senses and interact with it often enough to make sense of it (embodied mind theory).
The 1990s – the second AI winter
Despite the success of the previous decade, AI as a field again entered a harsh period of funding deprivation. On the one hand, expert systems were novel and promising. On the other hand, they were difficult to implement, required a lot of money to deliver, and most importantly such systems had so-called “black boxes” – no one could really explain how the machine was reasoning to reach conclusions. Other less sophisticated systems could do the same work faster and cheaper.
Mentioning artificial intelligence in the scientific world became despicable once again and many researchers called their studies advanced computing, informatics, cognitive systems, etc. Still, AI received wide publicity and the absence of money from scholar of governmental institutions did not turn attention off the topic. Engineers kept on showing amazing results, like IBM’s Deep Blue winning a chess competition against grandmaster Gary Kasparov in 1997, followed by the demonstration of Kismet, a robot that could understand and display emotions putting the first stones on the road of training an artificial intelligence in emotional intelligence.
Finally, this decade helped scientists to agree upon at least some definitions, and artificial intelligence was defined as an intelligent agent, or such a system that perceives the surrounding world, analyzes it, and behaves so that it can maximize the chances of success. If you are to explain what artificial intelligence is in simple words, you should start from this definition.
Since the 2000s – data explosion
The new millennium brought a new wave of interest to artificial intelligence. Firstly, scientists kept on demonstrating what AI could do to the public. For example, once again AI performed better than humans when IBM’s Watson won Jeopardy! show. Secondly, computing power grew and finally became sufficient to keep up with the demands of AI systems. Powerful processors were rare and costly. Thirdly, and most importantly, people started to produce and store data that can be used to train AI. If today one can go on to Google images or any photo stock and compile a sample set quickly, previously collecting enough images of high enough quality was a challenge in itself.
As technologies advanced computers became cheaper and smaller. All sorts of portable personal electronic devices became a part of our life. Smartphones, smartwatches and fitness trackers, GPS navigators, voice recognizing assistants, and other wonders of modern tech are constantly collecting, generating, transferring, and storing great amounts of data that can be accessed and used for AI training. The explosion of available data advanced deep learning, computer vision, text and speech recognition. Currently, we have so much information about so many things around us that big data became a discipline of its own with a wide range of applications across domains.
Data scientists and machine learning engineers kept on demonstrating successful AI solutions so often that more and more researchers were paying attention to this field. The stigma of AI dissolved quickly and once again all sorts of industries became interested in the implementation of ML-backed solutions. Today, it is hard to find an area that could not be improved with artificial intelligence system: banking, security, education, gaming, healthcare, marketing, eCommerce, social platforms, you name it. Driverless and autonomous cars are being tested all the time. Computer-assisted translation systems keep on improving dramatically and maybe one can hope for a real time interpretation in a not so distant future.
Although artificial intelligence is not yet as smart as humans, an on-spot implementation of ML-boosted systems is commonplace. And SOLVVE is well equipped with the experts in the fields of machine learning, cognitive computing, natural language processing, and chatbots to assist the needs of your business or project. If you have any questions or ideas about AI-solutions do not hesitate to contact us. Let us make it happen!