Daniel Medina

Daniel Medina

Future Researcher

Understanding ChatGPT and Its Application in Construction

5 minute read

Published:

Introduction

Pre-trained Transformer models, commonly known as GPT, are advanced natural language prediction systems. Their goal is to mimic the process of human speech to facilitate fluid conversations, a task that is inherently complex due to the nuances of human language, which cannot be easily programmed through strict rules. To address these challenges, machine learning is employed, where the machine adjusts itself automatically based on provided examples.

For instance, ChatGPT has been trained with a vast amount of data, including texts from web pages, highly rated Reddit posts, compilations from the Universal Web Library, and the entirety of Wikipedia, covering the period from 2016 to 2019. Despite its capability to generate coherent text, it is crucial to clarify that ChatGPT does not genuinely comprehend the topics, lacks mathematical or abstract thinking, and has limited memory. Although these aspects are continually being improved, the most significant barrier remains human understanding. So, how does ChatGPT work?

History of AI

Figure 1: A brief history of Artificial Intelligence.

How ChatGPT Works

The process begins with the tokenization of the input text, where the text is converted into lists of vectors with multiple dimensions (parameters). In GPT-1, 117 million parameters were used, while GPT-2 employed 1.5 billion, and GPT-3 increased this figure to 175 billion parameters. These tokens are placed on a language map, where words are organized according to their semantic proximity. For example, the words "cat" and "dog" may be considered close due to their relationship with pets. This mapping is combined with techniques like Long Short-Term Memory (LSTM), which allows recalling what was previously mentioned in the conversation, and the Transformer model, which enables contextualization—in other words, it can determine the language map area that best fits the requested topic. Finally, the model generates text based on the calculated probabilities for the next token, producing a coherent response.

ChatGPT Operation Diagram

Figure 2: ChatGPT Operation Diagram.

Application in the Construction Sector

In the construction sector, one of the most appealing uses of ChatGPT is its application as a virtual assistant to facilitate information on projects, management, optimization, and report generation. However, for these applications to be effective, a large amount of verified data is required. In construction, the quality and availability of data can vary, limiting the implementation of these virtual assistants. Nevertheless, the use of Building Information Modeling (BIM) offers hope, as the information from the Common Data Environment (CDE) of each country could allow the creation of virtual assistants capable of managing projects efficiently.

GPT models in the construction industry

Figure 3: GPT models in the construction industry.

Recommendations for Using ChatGPT

  1. Interaction with the Machine: Remember that you are interacting with a machine that does not understand human context. Avoid complex manners and literary resources.
  2. Logical Language: Use clear and logical language, although it may seem obvious, this skill is being lost, especially in Latin American countries with low reading rates and deficient mathematical abstraction skills.
  3. Specificity and Detail: Be specific and detailed in your requests. Avoid redundancy and ambiguity.
  4. Technical Limitations: Consider technical limitations, such as the maximum number of input characters and the maximum output length. Remember that the machine has a short memory and may forget details after prolonged conversations.
  5. Use as an Information Source: Do not use ChatGPT as a definitive source of information. The hallucination phenomenon, along with detailed writing, can lead to misinformation.
  6. Mathematical Calculations: Do not use ChatGPT for mathematical calculations. Although it can solve various problems requiring mathematical abstraction and is constantly improving, it is not a calculator.

Conclusion

In conclusion, ChatGPT has great potential to improve efficiency in the construction sector, provided that its limitations are understood and respected and that it is used appropriately. However, its application also presents new challenges. It is necessary to have professionals with skills that go beyond the traditional training in the sector, as well as a standardization in project production that includes not only construction standards but also writing and report presentation guidelines.

Moreover, it would be highly beneficial if large companies in the sector were willing to release part or all of the information they possess, as this high-quality information could represent a significant advance for the construction industry in an entire country. Collaboration and access to verified and quality data are essential to fully exploit the capabilities of ChatGPT and other artificial intelligence tools in the construction sector.

Bibliography

  • Hochreiter, S., & Schmidhuber, J. (1997). Long short-term memory. Neural computation, 9(8), 1735-1780.
  • LeCun, Y., Bottou, L., Bengio, Y., & Haffner, P. (1998). Gradient-based learning applied to document recognition. Proceedings of the IEEE, 86(11), 2278-2324.
  • Krizhevsky, A., Sutskever, I., & Hinton, G. E. (2017). ImageNet classification with deep convolutional neural networks. Communications of the ACM, 60(6), 84-90.
  • Vaswani, A., Shazeer, N., Parmar, N., Uszkoreit, J., Jones, L., Gomez, A. N., … & Polosukhin, I. (2017). Attention is all you need. Advances in neural information processing systems, 30.
  • Goularas, D., & Kamis, S. (2019, August). Evaluation of deep learning techniques in sentiment analysis from twitter data. In 2019 International Conference on Deep Learning and Machine Learning in Emerging Applications (Deep-ML) (pp. 12-17). IEEE.

You May Also Enjoy

Do You Know the Finite Element Method?

16 minute read

Published:

Introduction

The history of the Finite Element Method (FEM) dates back to the early 20th century, with its roots traceable to methods for solving partial differential equations developed by mathematicians like Richard Courant and Raymond D. Mindlin. In the 1940s, Courant utilized discretization techniques to solve torsion problems, laying the theoretical groundwork for what would later become FEM. However, the formal development of FEM began in the 1950s with the pioneering work of aerospace engineers like Jon Turner and Ray W. Clough. Turner and his colleagues at Boeing developed the concept of finite elements for analyzing aerospace structures, while Clough coined the term "Finite Element Method" in his influential 1956 paper.

From Stone to Self-Healing Bacterial Concrete

14 minute read

Published:

Introduction

Throughout history, construction materials have evolved significantly, adapting to the needs for durability, strength, and efficiency of each era. From the early civilizations using natural materials like stone and clay, to contemporary advancements in composite materials and nanotechnology, the quest to improve infrastructure has been constant. In ancient times, the earliest stone structures, such as the pyramids of Egypt and the temples of Mesopotamia, demonstrated advanced construction techniques for their time. With the advent of the Roman Empire, concrete was introduced—a mixture of lime, water, and volcanic stone—which enabled the construction of more complex and durable structures, such as the Pantheon and Roman aqueducts.

During the Middle Ages and the Renaissance, construction materials continued to improve with the use of more sophisticated bricks and mortars. However, it was during the Industrial Revolution that significant advancements were made with the introduction of steel and reinforced concrete, allowing for the construction of taller buildings and longer bridges. In the 20th century, the development of new materials such as prestressed concrete and advanced metal alloys revolutionized civil engineering. However, despite these advancements, issues related to the durability and maintenance of concrete structures persisted. The need for new materials that could offer solutions to these problems led to the research and development of self-healing concretes.

Self-healing concretes are designed to autonomously repair cracks and minor damages without human intervention. These materials can be classified into several types based on the self-healing mechanism employed.

Is the E.030 Standard Enough?

12 minute read

Published:

Introduction

On the morning of November 1, 1755, during All Saints' Day, a catastrophic event shook Lisbon. It was one of the most devastating earthquakes in modern history, classified as an XI intensity on the Mercalli scale (extreme earthquake). This disaster was not isolated; it was accompanied by a tsunami and a fire. Reports from the time indicate that the earthquake lasted between three and six minutes, causing cracks up to five meters wide in the city's center. Survivors who fled to the docks saw the water recede, revealing the seabed. Forty minutes later, three giant waves, ranging from 6 to 20 meters high, swept through the port and the city center, advancing up the Tagus River. In areas unaffected by the tsunami, fires erupted, devastating the city for five days, mostly ignited by candles lit in churches in honor of the deceased.

Why Should We Learn to Program?

8 minute read

Published:

Introduction

In today's environment, technology and programming are fundamental across numerous industries, particularly in construction. According to a 2023 Stack Overflow report, most professional developers use languages like JavaScript, Python, and Java, highlighting the growing importance of these skills. It is anticipated that by 2025, the global software market will reach 650 billion dollars, reflecting the continuous expansion of technology in various sectors.