The databases of the various systems in the construction business – with their inevitably decaying and increasingly complex infrastructure – are becoming a breeding ground for future solutions. Company servers, like a forest, are rich with a biomass of important information, often hidden underground in the bowels of folders and servers. The masses of data from the various systems being created today – after use, after falling to the bottom of the server and after years of fossilization – will fuel machine learning and language models in the future. Internal company chat rooms (e.g., a separate instance of locally configured ChatGPT, LlaMa, Mistral, DeepSeek) will be built on these in-house models using centralized storage to quickly and conveniently retrieve information and generate the necessary graphs, dashboards, and documents.
Fossilization of plant mass in combination with pressure and temperature creates a homogeneous and uniquely structured homogeneous mass of trees of different species that lived at different times – charcoal (“Conditions Required for Plant Fossil Preservation,” 2024). In the same way, information recorded on hard disks in different formats and at different times under the pressure ofanalytics departments and temperature of quality management eventually forms a homogeneous structured mass of valuable information (Fig. 9.2-1).
These layers (or more often isolated nuggets) of information are created through painstaking data organization by experienced analysts who begin to gradually extract valuable information from seemingly long irrelevant data.
The moment these mature layers of data are no longer just “burned” in reports, but begin to circulate in business processes, enriching decisions and improving processes, the company becomes ready for the next step- the transition to machine learning and artificial intelligence (Fig. 9.2-2).
Machine learning (ML – Machine learning) is a class of methods for solving artificial intelligence problems. Machine learning algorithms recognize patterns in large data sets and use them to learn themselves. Each new data set allows the mathematical algorithms to improve and adapt according to the information obtained, which allows to constantly improve the accuracy of recommendations and predictions.
As the influential CEO of the world’s largest investment fund (which owns key stakes in almost all of the largest construction software companies, as well as the companies that own the most real estate in the world (А. Boiko, “Lobbying wars and BIM development. Part 5: BlackRock is the master of all technologies. How corporations control open source code,” 2024) said in an interview with 2023) – machine learning will change the world of construction.
AIhas enormous potential. It will change the way we work, the way we live. AI and robotics will change the way we work and the way we build, and we will be able to use AI and robotics as a means to create much greater productivity (“BlackRock’s Fink on bonds, mergers and acquisitions, the U.S. recession, and the election: Full Interview,” 2023).
– CEO of the world’s largest investment fund, interview, September 2023.
Machine Learning (ML) works by processing large amounts of data, using statistical techniques to mimic aspects of human thinking. However, most companies do not have such datasets, and if they do, they are often not sufficiently labeled. This is where semantic technologies and transfer learning, a technique that allows ML to be more effective when dealing with small amounts of data, the feasibility of which has been discussed in previous chapters of this part, can help.
The essence of transfer learning is that instead of processing each task from scratch, you can use knowledge gained in related fields. It is necessary to realize that patterns and discoveries from other industries can be adapted and applied in the construction industry. For example, methods of optimizing logistics processes developed in retail help to improve the efficiency of construction supply chain management. Big data analysis, which is actively used in finance, can be applied to cost forecasting and risk management in construction projects. And computer vision and robotics technologies being developed in industry are already finding application in automated quality control, safety monitoring and construction site facilities management.
Transfer learning allows not only to accelerate the introduction of innovations, but also to reduce the cost of their development, using the already accumulated experience of other industries.
Human thinking is organized on a similar principle: we rely on previously acquired knowledge to solve new problems (Fig. 4.4-19, Fig. 4.4-20, Fig. 4.4-21). In machine learning, this approach works too – by simplifying the data model and making it more elegant, we can reduce the complexity of the problem for ML algorithms. This in turn reduces the need for large amounts of data and reduces computational cost.
