|
Versatile artificial intelligence
Definition and History of Artificial Intelligence
By the late 1940s and early 1950s, the possibility of an artificial brain was discussed by scientists in various fields, including mathematics, philosophy, engineering, and economics. Elon Turing conducted a Turing test to determine machine intelligence based on how similar a machine can communicate with humans, and it was the first philosophical proposal for artificial intelligence. In 1956, 10 scientists gathered at a Dartmouth conference held by Professor John McCarthy at Dartmouth University in the U.S. to define the concept of artificial intelligence and to use artificial intelligence terms for the first time. In other words, artificial intelligence was being studied even before the term artificial intelligence was coined. The core of artificial intelligence research at that time was inference and search. The Dartmouth Conference was a moment that included the name ai, the goal, the first success and the people who achieved it, and the birth of ai in a broad sense. The program created during this period surprised many people, who solved algebra problems, defined the definition of geometry, and learned English. Seeing the intelligent behavior of such a machine, I believed that everything would be possible with ai. Through search engines, it has been able to collect unprecedented amounts of data, and as it has evolved into machine learning, where artificial intelligence systems learn by themselves by analyzing numerous big data, artificial intelligence has entered a revival period with the Internet. As deep blue, a type of artificial intelligence, beat Gari Kasparv, the world chess champion, after six battles, it was revealed that humans could see 10 moves, but deep blue, an artificial intelligence, could see up to 12 moves. The current Ai is developing enough to recognize even the face of a cat, or animal. Let's find out where Ai, which has endless potential for development, is being used.
Artificial Intelligence in Mechanical Engineering
Since machine artificial intelligence is a convergence field, it is necessary to study fields that look different from each other, and artificial intelligence itself is developing so fast that it is not easy to adapt. Understanding artificial intelligence algorithms requires mathematical knowledge such as linear algebra, probabilities, and statistics.
- Unsupervised and supervised learning in machine learning
Machine learning is generally divided into two categories according to the type of learning data.Through unsupervised learning, hidden variables and structures that exist behind the data are mainly found, or representation is given through clustering. The supervised learning extracts a pattern capable of determining an output for a new input based on the data given the input and output.
-Deep learning
Since images contain a lot of visual information, efforts to extract necessary information from images have long been made. In particular, in the field of image processing, access has been attempted through extraction of characteristic factors such as vertices and contours that can represent objects. However, the recent emergence of convolutional neural networks during deep learning has changed the paradigm of image recognition. This succeeded in classifying objects at a level similar to humans, with less than 5% errors in 100,000 images divided into 1000 categories.
Artificial Intelligence in Architecture
-Basic Principles of Architecture in Artificial Intelligence Technology
If you know the principle that artificial intelligence technology at the current level derives architectural plans, you can infer the possibilities and limitations of artificial intelligence. The most basic method is to create numerous alternatives using the computing power of computers to find an appropriate architectural design plan for the land. Through artificial intelligence technology, various alternatives that can be built on a specific site are created and scores are set. You can choose the proposal with the highest score in the result. However, the number of alternatives is very large. Quantification of artificial intelligence technology is suitable for architectural design if you want to find the most scoring alternative among them.
- Genetic algorithm
It is a search algorithm that imitates the theory of evolution that finds better results with a combination of inputs that produce good results based on the genetics of nature. Since the early 1990s, it has been tried as a way to find a building plan, and it has been used in the field of building planning with the recent development of technology.
– Deep reinforcement learning
It is similar to the deep blue mentioned earlier, but it is a technology applied to AlphaGo. Deep reinforcement learning does not learn or refer to people's notation. Instead, if you tell us the rules, we calculate the number of cases on our own and learn the situations and patterns of how effective we are when we act. In the case of Landbook, Spacework's architectural design automation service, artificial intelligence technology is set up with parking lot planning laws and trained through in-depth reinforcement learning. Artificial intelligence technology learns by itself the patterns in numerous parking lot plans within a given environment. If a new land shape is presented after about 8 days of learning, it can be confirmed that a parking lot plan is derived within a few seconds. In addition to rental returns, high appropriation rates, and efficient construction costs, indicators such as mining, view, straight thread shape, and short movement lines can also be quantified.
- The Future Direction of Artificial Intelligence
Artificial intelligence technology is already coming deep into our lives through various services and products, and in the future, almost all data-based industries will be automated and optimized by artificial intelligence technology. In addition, it is not unfamiliar to see artificial intelligence replace or collaborate with experts, not only innovation in traditional industries, but also areas of experts that require high intelligence and field experience. However, it is appropriate to say that this rapid spread does not mean the technical completion of artificial intelligence, but rather that most of the intellectual actions performed by humans could be implemented using large amounts of data, sufficient computation, and efficient machine learning models. Artificial general intelligence, which is pursued by superintelligence or strong artificial intelligence or deep mind that threatens the survival of mankind warned by Ellen Musk or Stephen Hawking, is still far from being achieved, and current weak artificial intelligence or narrow artificial intelligence has much room to improve in terms of computational efficiency or learning. For now, the most efficient way to overcome these limitations of artificial intelligence is to share and collaborate with experts in the field, and building a user environment that enables this collaboration efficiently will be a key competitive edge in artificial intelligence products or services.
I feel it…
The impact of artificial intelligence on me was not small. Artificial intelligence was essential to me in my daily life. Therefore, when I had to write a newspaper on the subject of artificial intelligence, I think there was a little feeling of welcome. Furthermore, it is because mechanical engineering, architecture, and electronic engineering, that is, artificial intelligence in the departments I want to hope for, have been mentioned recently. It was a time to find out the role of artificial intelligence in the department I want and remind myself of the necessity in life.