Artificial intelligence (AI) uses computer algorithms to simulate human intelligence. With the advent of Large Language Models like ChatGPT (released in November 2022), AI has taken the business, professional, and academic worlds by storm. In basic terms, ChatGPT is trained on vast amounts of text from various sources, which it uses to generate responses by predicting the most likely sequence of words based on an initial prompt. A separate set of AI tools (for example, DALL-E, which I used to generate the images included in this piece) similarly create graphics based on prompts from users.
Throughout human history, technological advancements have reshaped the nature of work. However, the promise—and potential perils—of generative AI appear to represent a fundamentally different magnitude of change. Until recently, the most tangible impacts of technology on the workforce were seen in automation, where routine tasks performed by humans are replaced by machines. The impacts of AI, however, extend into creative and professional domains. On the one hand, it can enhance productivity, enabling professionals to work more efficiently and creatively. (Full disclosure: I called on ChatGPT as an editorial assistant in writing this piece.) On the other hand, AI may pose a threat to creative and professional jobs, potentially displacing roles that were once thought to be immune to automation. Furthermore, AI “levels the playing field” between highly skilled and less skilled professionals using AI-powered tools, both increasing productivity and raising concerns about job displacement and the devaluation of expertise. These considerations raise a critical question for the future of fields like planning: how will AI impact professional practice?
Living and Working with AI
Earlier this year, Ethan Mollick published Co-Intelligence: Living and Working with AI – a book I highly recommend. A professor at the University of Pennsylvania’s Wharton School, Mollick defines co-intelligence as the potential for AI to augment, or even replace, human intelligence. Part 1 of the book identifies four principles for working with AI:
- Always invite AI to the table: This means regularly using AI not only for help, but also to understand its capabilities, limitations, and implications for your job.
- Be the human in the loop: At least for the present, AI works best when paired with human judgment and expertise. This underscores the importance of collaborating with AI to leverage its strengths and account for its limitations.
- Treat AI like a person: This means creating a specific persona or role for AI and interacting with it as you would with a colleague or intern to achieve optimal results.
- Assume this is the worst AI you will ever use: Given AI’s exponential pace of development, current applications will rapidly be replaced by far more sophisticated ones. This means that the potential for AI to transform work – and life – is just beginning (Mollick 2024, Chapter 3: Four Rules for Co-Intelligence).
In Part 2 (the main body of the book), Mollick explores how to use AI in various roles (for example, as a co-worker, tutor, and to spur creativity and innovation). He also provides examples of how to construct prompts to maximize AI’s value in these roles.
ChatGPT on AI and Planning
Returning to the question on the implications of AI for planning, and armed with what I learned from Co-Intelligence, I went right to the source by posing the following prompt to ChatGPT:
I am an urban planner seeking to understand the implications of artificial intelligence (AI) for the city and regional planning profession. Please identify the top three impacts you think that AI will have on the profession over the next 5-10 years. Please explain your reasoning, including why you picked these three and not others.
In a matter of seconds, ChatGPT responded with its “top three ways AI might influence the profession”:
Data-driven decision making and predictive analytics: enhanced modeling and simulation, predictive maintenance and infrastructure management, and tailored community engagement
Automation of routine tasks: permit processing and compliance checking, automated reporting, and geospatial analysis
Equity and inclusion in planning: identifying disparities, scenario planning, and accessibility improvements
The program provided further explanation of the subtopics identified for each of the three impacts. Predictive maintenance and infrastructure management, for example, means “predict(ing) when infrastructure like roads or utilities might fail, allowing for proactive maintenance and resource allocation.”
In reviewing ChatGPT’s response, I found the first two impacts – data-driven decision making and automation of routine tasks – interesting but unsurprising, as they reflect current (human) thinking about the implications of AI for planning. (For an example, see Plan GPT: Enhancing Urban Planning with Tailored Language Model and Efficient Retrieval, He Zhu et al. 2024). The third impact – equity and inclusion – was more intriguing, as it represents an implicit value judgment derived from its training data (why not, for example, economic growth?). ChatGPT explained its rationale as follows:
These impacts were chosen because they represent a combination of immediate, practical benefits (data-driven decision making and automation) and a broader, values-driven shift (equity and inclusion) that reflects the evolving goals of the planning profession. Other potential impacts, such as changes in employment or the ethical considerations of AI, are also important but are likely to emerge as secondary effects of these primary changes.
Pretty impressive “reasoning” for a technology still early in its development! Which raised another question in my mind: what is the possibility of AI not just making urban planners and related design professionals more productive, but eventually replacing them altogether? A 2023 study on the potential job impacts of Large Language Models identified language and literature teachers, writers, legal services, finance, insurance, and accounting as occupations potentially vulnerable to replacement by AI (Felten et al. 2023). Could AI’s rapid development make so-called “creative professionals” similarly vulnerable?
What About Architecture?
To test the above possibility for the architectural profession, I posed the following question to ChatGPT:
Do you foresee AI having the capability to design buildings that surpass human architects in functionality, performance, AND beauty?
This question arose in the final reviews for a spring semester 2024 Master of Urban Design studio I taught at Thomas Jefferson University, in which students used computational software to generate and evaluate the performance of literally thousands of design alternatives. While this process yielded alternatives that optimized factors related to functionality and performance (for example, economic feasibility, affordability, and solar energy potential), it could not address whether the final result would be an attractive place for people to be. Here’s how ChatGPT responded:
AI is advancing rapidly in fields like architecture, and it’s possible that it could eventually design buildings that surpass human architects in functionality, performance, and even beauty, but with some caveats…AI might generate designs that are considered beautiful by many, but whether it can consistently create beauty that resonates deeply on a human level is still an open question.
The program went on to describe what it called a “Collaborative Future” (which is quite similar to Ethan Mollick’s concept of co-intelligence):
Rather than fully surpassing human architects, AI is more likely to serve as a powerful tool that enhances their capabilities. Architects could use AI to explore a broader range of design possibilities, optimize for performance, and even create new forms of beauty that combine human creativity with AI’s data-driven insights. The synergy between AI and human architects might lead to buildings that are more functional, perform better, and are considered beautiful in ways that neither could achieve alone. In summary, AI might eventually surpass human architects in certain aspects of design, but the most impactful and beautiful buildings will likely come from a collaboration between human creativity and AI-driven optimization.
In response to my follow-up prompt, DALL-E (which is integrated into ChatGPT Plus) generated an image it described as “a futuristic building that embodies functionality, performance, and beauty, envisioned as a result of collaboration between human creativity and AI optimization.”
Concluding Thoughts
ChatGPT’s description of a collaborative future sounds promising and (at the risk of anthropomorphizing a computer program) perhaps diplomatic given the trajectory of AI’s development. Consider, for example, that Mollick identifies “The Machine God” as one of four possible scenarios for the future of AI. In this scenario, machines not only match but exceed human intelligence, superintelligence emerges, and human supremacy ends (Mollick 2024, Chapter 9: AI as Our Future). Is there a role for human designers (and planners) in this scenario?
Given the unpredictability of technological evolution, it is undoubtedly more productive to consider the short-term impacts of AI on the planning and design professions than to speculate about an unknowable future. The top three ways AI might influence planning identified by ChatGPT address potentially positive impacts. Potential negatives include privacy concerns, bias and discrimination, and quality/availability of data, among others. Large Language Models are, at present, poor research tools – they are both dependent on the incomplete databases they were trained on and have an unfortunate propensity to “make stuff up” (often called hallucinations).
It is also reasonable to question whether the “futuristic building” generated by DALL-E is “beautiful” by objective human standards. In response to a prompt, the program asserted that “beauty is embodied through (five) architectural elements and design choices,” which I will not describe here. I did not find the response to be wholly convincing and will leave it up to readers to make their own judgments.
Returning again to my original question on the implications of AI for the planning profession, I asked ChatGPT to make concluding recommendations for planners. It responded with three:
- Continuous Learning: Keep abreast of technological advancements to understand and leverage AI tools effectively.
- Ethical Considerations: Advocate for and design guidelines that ensure AI applications in planning enhance societal well-being and do not exacerbate inequalities.
- Collaborative Design: Engage with technologists, policymakers, and the community to create urban spaces that reflect collective needs and values.
These certainly sound reasonable, and I can’t disagree with any of them. If I were to add anything, it would be to encourage planners to experiment with tools like ChatGPT to spur innovation and (hopefully) achieve better outcomes than we could achieve on our own. And to remember the ultimate goal of planning: to make communities better for people, irrespective of utopian or dystopian visions of what future technologies may bring.
References
Felten, Ed, Manav Raj, and Robert Seamans (2023). How Will Language Modelers like ChatGPT Affect Occupations and Industries?
Zhu, He et al (2024). PlanGPT: Enhancing Urban Planning with Tailored Language Model and Efficient Retrieval.
Mollick, Ethan (2024). Co-Intelligence: Living and Working with AI. New York: Portfolio/Penguin