Phobos. An experiment in creating AI in architecture. Part 1.

        "I think our mind is a program, while the brain is an analogue of a computer. It is theoretically possible to copy the contents of the brain to a computer and thus create a form of eternal life. Today, however, it is not in our power"
                                                                                                                                     Stephen Hawking.

        Reading this phrase, you can think about it, and after all, perhaps all of this really naked theory is not applicable in practice!? This question can not be answered by giving an example from everyday practice. When I started developing the CUBE application, for the most part I simply decided to test my strength, but the further I went beyond my experience, the more my interest in artificial intelligence grew. And today (01/01/2019) I decided to try my hand at a more difficult task, which I had been working from last May. The task itself is quite simple and at the same time incredibly difficult, since it is not limited to managing points of space, or solving a system of equations, or decomposing into simple factors, it is creative in its structure. The solution of the problem is reduced to the creation of the layout of the room / apartment / floor from scratch and this task will be solved by my new program called Phobos. While the work is being done on the layout of the premises and the work is enormous, the preparatory period was more than 2 months, during which I was simply making a plan, the slightest mistake would throw me back a few months ago.
        Phobos is a deity denoting fear in Greek mythology, and he accompanied his father Ares, the god of war, in the battles, terrifying his enemies. Phobos is an experiment to create a creative mind that may or may not replace people, it is the fear of the unknown in the confrontation between artificial intelligence and man.

        The problem is how to impose the creative component of the program in the classical way? After all, creativity is not a mathematical function, creativity is not possible to describe in numbers! But let's see if this is in fact true, can it be quite possible to describe the work mathematically ?! How do you like this idea?


         For starters, let's go over history. Within the framework of European architects, there are many architects of theorists, Ernst Neufert and Peter Neufert are among the most prominent. Their works belong to the 40th years of the last century in which they reduced the accumulated experience of mankind into various patterns. For example, a person's comfortable step on the steps, the width of the aisles, the size of sofas, dressers, bedside tables and much more. For an example, let's see how he saw the comfortable placement of furniture in the bedroom, Ernst Neufert:

 And now Peter Neufert:

      One bedroom has a size of 4.5 * 3.62 = 16.29 m2, the other has 3.6 * 3.5 = 12.6 m2. In fact, the solutions are no different except the size of one of the nightstands with the lamp. That is, we see a clear relationship with the location of the furniture in the bedroom. The same applies to bathrooms and hallway and a common room, etc.

       If there are dependencies, then there are mathematical laws describing them. For example, it is impossible to install a double bed in a bedroom of 8 m2:

         The area of ​​this room is dictated by the furniture that can be installed in it based on the purpose of the room: a desk, a bed, a chair, a wardrobe. Each element has close to standard dimensions, it is also necessary to take into account a convenient passage for the average person. In addition, the layout is strongly influenced by the location of windows and doors, their mutual arrangement and much more.

        Phobos is not a BIM program, this is the next level of programs that use the information that the BIM model stores and uses for training; after training, the BIM models will no longer be needed .....

        In the next part we will talk about an algorithm based on a convolutional neural network that parses websites (in simple terms, it extracts all the necessary graphical information in the most cruel way) of developers and forms the basis of planning, as well as the crisis of planning decisions of developers. A total of about 30 parts of the article are planned and this is a kind of developer diary. If the experiment is successful, then in the hands of the person there will be a tool capable of offering thousands of layouts.