Software + Tech
GPT Goes Beyond Document Management Previous installments of this series delved deep into the transformative impact of Generative Pre-trained Transformer (GPT) technology on the AECO industry. It has explored how GPT's natural language processing capabilities can revolutionize document management and empower sustainability initiatives, making information more accessible and processes more efficient. While GPT has proven itself indispensable in handling documentation and language-based queries, the AECO industry is not solely about documents–data plays an equally crucial role. Transitioning into discussing data analytics, it becomes apparent that while GPT excels in language, its capabilities in data analytics present a different set of challenges and opportunities. This article explores how GPT's language model nature influences its handling of data and discusses the potential for synergy between it and advanced analytics modules. The Strength of GPT in Handling Documentation GPT's core strength lies in its ability to understand and generate human- like text, making it exceptionally good at handling documentation. As a language model, GPT has been trained on vast amounts of text data, enabling it to understand context, infer meaning, and generate coherent responses. This allows users to ask natural language questions, whether objective or subjective, and receive accurate and relevant answers. For instance, questions like, “How can I implement a specific protocol?” or “What should I consider when drafting a contract?” are easily handled by GPT, as it aggregates knowledge from various sources to provide comprehensive answers. By Dr. Jeff Chen, Director of Digital Transformation and George Broadbent, VP of Asset Management, Symetri Synergizing GPT and Advanced Analytics: Unlocking New Opportunities in the AECO Industry
GPT’s Limitations in Data Analytics Despite its competency in language understanding and response generation, GPT is not inherently designed for data analytics. When faced with a question like, “What is 1+1?,” GPT will respond with “2”, but this answer is more a result of language experience rather than real calculation. Just as humans instantly know the answer is “2” without going through a conscious calculation process, GPT generates its response based on patterns in the text data it was trained on. While this process often results in correct answers, it can sometimes lead to obvious calculation errors, thus highlighting the limitation of relying on GPT for data analytics tasks that require precision and reliability. Bridging the Gap with Data Analytics Modules Recognizing GPT’s data analytics shortcomings opens the door for synergies with advanced analytics modules that excel in this area. Integrating GPT with specialized data analytics tools and packages can create a powerful combination, where GPT handles natural language queries, and the analytics modules perform precise calculations and data analysis. Some of the data analytics modules that can work seamlessly downstream of GPT include: • Pandas and NumPy: These Python libraries provide extensive functionalities for data manipulation and numerical computations. GPT can be used to interpret user queries and generate appropriate code snippets to be executed in Pandas or NumPy, ensuring accurate and efficient data analysis. • SciPy and Scikit-learn: For more advanced statistical analysis and machine learning tasks, these offer a wide array of algorithms and tools. GPT can guide users in selecting the right methods and parameters, while the libraries carry out the computations.
40
csengineermag.com
December 2023
Made with FlippingBook Annual report