ChatGPT and its Impact on Transit Asset Management

ChatGPT is gaining tremendous popularity and showcasing potential to disrupt most, if not all, industries. Google, the dominant leader of the $149 billion search engine business, has declared a “code red” situation and re-engaged the company’s founders for the first time since they left their daily roles. Considering how Google has impacted our every day life, our imagination knows no bounds in assessing the potential impact of ChatGPT.

What is ChatGPT?

ChatGPT is a chatbot powered by artificial intelligence (AI) that generates human-like responses based on text prompt of user. It has been trained on a diverse range of Internet text to generate human-like responses to a given prompt. The GPT (Generative Pre-training Transformer) architecture enables various language tasks such as question answering, language translation, and text summarization.

ChatGPT aims to offer quick and direct responses to specific problems, even if the questions are complex. The large size of the model and its performance on a variety of language tasks has made it a popular choice. The chatbot can automate and simplify specific tasks. Instead of filtering through search engines, users get relevant information with a simple text input.

OpenAI, the developer of ChatGPT, is a San Francisco-based startup that has received another $10 billion in funding from Microsoft in the last few days, which is in addition to the $1 billion that the tech giant invested in the startup in 2019. Meanwhile, Google has invested nearly $400 million in Anthropic, which is testing a rival to OpenAi’s chatbot.

OpenAI offers a cloud-based API service for a fee that allows users to integrate the model into their applications and use it for a variety of natural language processing tasks. The cost of using the API is based on the number of API requests and the amount of processing power required.

ChatGPT is a free service in its research stage. In February 2023, OpenAI announced launch of ChatGPT Plus, a paid subscription plan that offers faster response times and reliability during peak hours.

Use with spreadsheets and documents

ChatGPT uses natural language processing techniques to understand and respond to text input, but it does not have the ability to directly read and process spreadsheets (e.g., Excel) or document (e.g., MS Word, PDF, etc.) data. The following methods can be deployed to use ChatGPT with spreadsheets and documents:

• Integrate ChatGPT with a chatbot platform to extract data from Excel spreadsheets or extract text from a document and use it within ChatGPT's responses.
• Convert any Excel data into a readable format, such as JSON or CSV, and any document into plain text or HTML, which can be processed by ChatGPT.
• Use a data scraping tool to extract the data from Excel or use optical character recognition (OCR) technology to extract the text from the document and feed as input into ChatGPT.

Impact on TAM

Transit Asset Management (TAM) provides a framework to monitor the condition and performance of assets over time, prioritize investments to achieve state of good repair, and allocate resources to deliver service in compliance with agency goals (such as on-time performance, equity and sustainability) and regulatory requirements (such as safety and accessibility).

Transit agencies are increasing the use of data and analytics in TAM to:

• Develop plans that are comprehensive and effective;
• Manage assets that are aging and require major overhaul and frequent maintenance;
• Deploy staff and resources to maximize output in available work window; and
• Prioritize investments in an environment of limited funding (significantly increased since the pandemic but still inadequate to clear state of good repair backlogs).

Data collection and analysis

ChatGPT can collect, process, interpret and store asset data from a variety of sources (including sensors, video cameras, fleet data, etc.) for efficient and accurate data entry into the Asset Registry, the centralized database that manages information about all assets and is a critical component of a TAM system.

The information stored in an asset registry typically includes details such as the asset's make and model, location, age, condition, maintenance history, operational data, and costs of renewal or replacement.

ChatGPT can analyze large amounts of data and identify patterns to monitor asset condition, predict equipment failure and security threats, inform decision making and investment prioritization, optimize asset utilization, increase operations efficiency, manage mixed fleets and energy consumption, improve data visualization and enhance passenger experience.

EAMs and PIS

AI and machine learning (ML) have enabled development of asset management systems that learn, diagnose, predict failures and trigger maintenance alerts by processing data from multiple sources. With its natural language processing interface, ChatGPT can be integrated into cloud-based Enterprise Asset Management Systems (EAMS) to provide a conversational experience for users who access data. These users could be from Maintenance, Operations, Finance or Executive departments.

If use of data from the EAM is easy then it is more likely to be developed as a single source of data within the organization. This can provide immense benefits such as high data quality, improved accuracy and consistency, enhanced collaboration and reduced costs.

In addition, Chat GPT can be integrated into passenger information systems (PIS) to provide real-time information and customer support by answering questions about routes, fares, vehicle arrival and service disruptions.

Business intelligence

ChatGPT can analyze Big data and identify patterns, trends, and correlations that may not be easily apparent to human analysts. This can provide insights to improve the effectiveness of TAM plans, recommend strategies to manage risks, and provide new business intelligence to meet the goals of cybersecurity, climate resilience, equity, and customer engagement.

With Internet of Things (IoT) interconnecting physical devices, vehicles, buildings, other items embedded with sensors, social media, and online transactions, massive volume of structured and unstructured data is generated. This data is characterized by its volume, velocity, variety, and veracity, and is typically too large and complex to be managed and analyzed using traditional data processing tools and techniques. ChatGPT can process this data in real-time, identify patterns and anomalies, and make predictions for maintenance or alerts for frauds/crime. This facilitates automation and development of connected systems.

Blockchain can be used to store and manage the decision-making process of AI algorithms, providing an immutable record of the inputs, outputs, and intermediate steps. This helps understand the basis for AI-generated decisions, increasing transparency and accountability. The decentralized nature of blockchain provides resistance to single point failures and reduces the risk of data breaches and hacking.

Maintenance

The transit industry has been relatively slow in adopting advanced technologies such as AI and ML for maintenance. More mature strategies such as predictive and condition-based maintenance are typically adopted by transit agencies with big investment programs. For the usual mid- and small-size agency with smaller budgets, maintenance has traditionally been performed by a combination of corrective and preventive actions based on experience and understanding of asset performance.

ChatGPT can be trained to analyze sensor data, maintenance records, and historical performance data, to provide predictive maintenance recommendations, reducing downtime and increasing availability. This reduces costs associated with equipment failures and unplanned maintenance.

Workforce

Many transit agencies have a significant portion of their workforce reaching retirement age, and are facing challenges in attracting and retaining younger workers to replace them. This raises concerns regarding preservation of institutional knowledge and expertise to adequately maintain equipment and infrastructure. ChatGPT can automate routine tasks such as report generation, form completion and asset record update to save time and reduce workload of existing staff. It can also record history of maintenance activities to assess plan for the future.

Issues and challenges

In its own words, ChatGPT has the following capabilities and limitations:

Capabilities:

• Remembers what user said earlier in the conversation;
• Allows user to provide follow-up corrections; and
• Trained to decline inappropriate requests.

Limitations:

• May occasionally generate incorrect information;
• May occasionally produce harmful instructions or biased content; and
• Limited knowledge of world and events after 2021.

The rapid growth of ChatGPT has not allowed time to address legal, ethical, data privacy and security concerns. As development and deployment progresses, regulations and standards will emerge to facilitate fair use and mass adoption. 

Transit agencies require experts with specialized technical skills and knowledge to fully benefit from advanced tools such as ChatGPT. Training and development, including online courses, workshops, and certifications, are needed to upskill existing employees, foster innovation, and reduce resistance to change.

Adoption of AI in transit

Several transit agencies in the United States and Canada are using AI to improve the efficiency, safety, and reliability of their services. These include:

• New York City Transit (NYCT), Société de transport de Montréal (STM), and Edmonton Transit Service to monitor rolling stock and/or track infrastructure condition as well as predict maintenance needs.
• Seattle Streetcar to monitor energy use and asset condition, as well as predict maintenance needs.
• San Francisco Municipal Transportation Agency (SFMTA) to improve accuracy of real-time passenger information systems.
• Massachusetts Bay Transportation Authority (MBTA) to analyze operations data from bus and train for improved service delivery.
• Los Angeles Metropolitan Transportation Authority (LA Metro) and OC Transpo to optimize bus routes.
• Vancouver TransLink and Toronto Transit Commission (TTC) to optimize bus and train schedules.

It is likely that more transit agencies will adopt AI as technology continues to advance; funding is made available; legal, ethical and security issues are addressed; and standards are developed.

Conclusion

The use of AI and ML is expected to increase in TAM because it can help transit agencies better understand asset condition and make data-driven decisions about maintenance, repair, renewal and replacement. The ability of ChatGPT to generate human-like text easily makes it a valuable tool for chatbots, language generation, and question answering, all of which have relevance for different tasks in TAM. The early adopters of this new technology must embrace a try-and-fail or fail-safe culture. As the technology develops further and its deployment increases, new use cases will emerge and assessments will be made regarding benefits and costs. We cannot fully imagine how and what will change with the arrival of ChatGPT, but we need to be aware of its significance and learn more about it to use it to our advantage.