Alice laughed. “There’s no use trying,” she said. “One can’t believe impossible things.”
“I daresay you haven’t had much practice,” said the Queen. “When I was your age, I always did it for half an hour a day.
Why, sometimes, I’ve believed as many as six impossible things before breakfast.”
Imagining the impossible
In Lewis Carroll’s “Through the LookingGlass,” the White Queen’s whimsical
directive for Alice to contemplate “six
impossible things before breakfast”
invites exploration beyond rationality,
sparking enduring wonder and possibility
in readers across generations.
Image: iStock (1866 print of “Alice’s Adventures
in Wonderland”)
We all want to use AI for good. But using it for bad is just too tempting.
Take the students we teach. It is all too easy for them to use AI to generate mindless answers to assignments. Even where the answers are correct (e.g., no hallucinations), mindlessness means students tussled less with the materials. They learn little as a result.
That is why our research — spanning the future of work, education, trust, and innovation — shows how to use AI for good.
We go beyond prompt engineering and personalisation (e.g. asking ChatGPT to be a teacher giving personalised student feedback). We jump several steps ahead to creating custom GPTs to perform a suite of tasks to elevate everyone’s potential.
We show six such possibilities — once thought impossible — for educators. They span how we:
POSS01
Assemble teams of research assistants to stay current
Educators must stay current in their fields. This is challenging because advances are accelerating. Educators have limited time and limited funds to hire research assistants.
Enter custom GPTs that help educators distil the latest advances into insights relevant to their fields.
For example, in our master’s-level Smart Cities class, we teach strategy so that urban leaders learn to use smart tech to create a competitive advantage for their cities. We built Strategy GPTs to review and transform recent research, publications, and long-form analysis (e.g., in reputable newspapers) into micro-case studies for discussions. Students and teachers thus stay current and cutting-edge on trends, ideas, and insights.
How much does this cost? The price of one large pizza every month. For educators, this possibility was once impossible.
By combining the strengths of AI and humans, teachers and students will ideate, iterate, and innovate more, pushing the boundaries of their learning and growth as creators.
POSS02
Assemble teams of teaching assistants to guide every student
Growing workloads make it hard for educators to give every student close guidance.
Enter the custom GPT “Innovating/Failure Secret Manual” (our current working name), which we created for a class in design innovation and designing deliberate failures.
It provides personalised, targeted, and 24-hour grading and feedback for an entire cohort of students from the perspectives of multiple teaching assistants across different design domains (e.g., engineer, architect, designer, citizen, customer, etc.). The instructors determine these assistants’ makeup to complement their domain expertise and fill in the gaps between their interactions with students.
Hiring an army of assistants to guide the students was previously impossibly unaffordable for educators. However, it is now a possibility — again, for just the price of a large pizza every month.
POSS03
Elevate everyone to be AI x Humans-Powered Creators
When we can assemble AI assistants to help teachers and students, what do we nurture them to become when AI now disrupts all skills?
As Stanford professor Erik Brynjolfsson concedes, “We had a hierarchy of things that technology could do, and we felt comfortable saying things like creative work… would be hard for machines to ever do. Now that’s all been upended.”
We can elevate them to be AI x Humans-Powered Creators.
On the surface, this contradicts the quote. But dig deeper, and sublime possibilities emerge.
AI has indeed outperformed humans in creativity tests. For example, it generates more initial ideas more quickly than humans. But AI has a fatal flaw: its ideas tend to be similar. Humans, however, have more diverse ideas, including higher-quality ones.
Combining AI with humans will give us more initial, diverse, high-quality ideas. Couple that combo with Poss01 & 02, and we elevate everyone to be AI x Humans-Powered Creators.
Take the same custom GPT “Innovating/Failure Secret Manual” we mentioned in Poss02. By combining the strengths of AI and humans, teachers and students will ideate, iterate, and innovate more, pushing the boundaries of their learning and growth as creators.
In the past, elevating everyone was impossible — it cost too much money and time. Now, powered by AI x Humans, it is possible.
POSS04
Design endless combos of AI x Humans interactions
To elevate everyone to be creators, we must elevate their capacity to design endless combos of AI x Humans interactions.
Why endless? Because AI is general-purpose — it can perform many tasks (unlike most technologies built for specific tasks). The possibilities are only limited by the creators’ imagination and domain knowledge when designing the AI to perform the tasks well.
A good example is a recent one-week Gen AI hackathon between our university and a large Singapore company. Sixty students worked with the company’s executives in teams on 19 live use cases. Through the interactions between the students’ use of AI, the executives’ domain knowledge, and their collective imagination, they created 19 solutions. In stark contrast to how most hackathons end, what was remarkable was that the executives wanted to adopt all 19 for immediate use or development.
Our research calls these interactions between combinations of AI x Humans Distributed Mastery. Distributed Mastery describes the phenomenon where expertise is no longer isolated within individuals but dispersed across a broad network of people and AI technologies. This setup allows for a collective intelligence that blends human domain knowledge with AI’s vast data processing capabilities. As a result, expertise becomes a shared asset, enhancing our ability to leverage rapid knowledge growth and innovate collaboratively
[Distributed Mastery] allows for a collective intelligence that blends human domain knowledge with AI’s vast data processing capabilities.
POSS05
Rapidly prototype cross-domain lessons to deepen domain knowledge
Domain knowledge is thus foundational. We must deepen it. This must be increasingly interdisciplinary, crossing domains.
However, lessons remain siloed within disciplines, and cross-domain endeavours remain hard.
The lesson plan is a case in point. Several years ago, when we prototyped a high school lesson plan linking climate challenges to school subjects, showing how different topics in different subjects were related, it took a team of eight researchers/ teachers/students six months.
Today, our recent prototype with AI took six minutes. It still needs refining by all involved, but the pace of rapid prototyping of the first and subsequent versions is unprecedented.
Imagine what we could pursue at that pace. Then, imagine what new possibilities we can create when we stack these with Poss01 to 04.
With each new AI advance, we can update our comparisons of where humans can continue to outperform AI and remain resilient.
POSS06
Measure resilience concretely: Task outperformance instead of skill learned (“Tasks first, skills second”)
One new possibility is this: we can measure our students’ resilience when they enter the workforce.
Previously, resilience was a fuzzy idea. Many presumed that being more highly skilled meant higher resilience. But concrete measures were missing.
Al changes all that. First, being highly skilled no longer means having higher resilience — that has been “upended” (see Poss03 quote). Second, we can use AI’s performance as a concrete comparison because AI now performs many tasks (see Poss04).
We can thus measure and compare the tasks where:
- Humans outperform AI
- Humans x AI outperform AI
- AI outperforms humans/Humans x AI
In tasks where humans/Humans x AI outperform AI, we equip humans with the skills, interactions, and domain knowledge needed. Where AI outperforms, we delegate those tasks to AI and equip humans with the smarts to verify the AI’s work.
Using task outperformance also fixes the perennial problem of assessing students’ application of skills learned. Grades were convenient but poor proxies. Now, we can accurately and directly measure students’ task outperformance vis-à-vis AI to assess how well they apply skills learned.
Moreover, with each new AI advance, we can update our comparisons of where humans can continue to outperform AI and remain resilient. Hence, with “Tasks First, Skills Second”, we can build concrete roadmaps to resilience.
MASS FLOURISHING
What is our cumulative aspiration for these six possibilities?
Beyond minimising AI for bad and maximising AI for good, our research centres on an aspiration to mass flourishing.
Nobel-prize-winning economist Edmund Phelps writes that “mass flourishing” occurs when many enjoy a combination of “material wealth… meaningful work, self-expression, and personal growth.” Historically, mass flourishing happened when many innovated instead of an isolated few.
Our six possibilities elevate everyone’s potential to “create, explore, and meet challenges” so that many can innovate. We need more innovators than ever because our world is beset by seemingly impossible-to-solve socio-economic and climate challenges. What better way to turn these around than by pursuing six possibilities once thought impossible? Our six possibilities lay the foundations for mass flourishing. ∞
The Singapore University of Technology and Design (SUTD) was one of the first universities in the world to incorporate the art and science of design and technology into a holistic interdisciplinary education and research experience.
Conceptualised as a niche trailblazing publicly-funded university as part of Singapore’s economic transformation, faculty and students are nurtured to be innovators to solve societal needs. For example, students are required to take humanities, arts, and social science classes in seven of their eight semesters, and would have completed at least twenty design projects by the time they graduate. In 2018, SUTD topped a list of emerging engineering schools in the world in an MIT-commissioned study.
SUTD students thrive in a design-steeped environment and are encouraged to think creatively and innovatively to solve real-world problems. We leverage on the power of deep tech like artificial intelligence to bring to life human-centric solutions using design and technology.
On 11 March 2024, SUTD unveiled a new growth strategy called SUTD Leap. By redesigning higher education with an even greater focus on design, AI and technology, SUTD Leap aims to propel the university to the forefront of the design x tech space whilst nurturing the next generation of design x tech innovators and innovator leaders.
REFERENCES
- Burn-Murdoch, John. “Here’s What We Know About Generative AI’s Impact On White-Collar Work,” Financial Times, 9 Nov 2023. https:// www.ft.com/content/b2928076-5c52-43e9-8872-08fda2aa2fcf.
- Huang, Qian, et al. “A Pedagogical Approach of ‘Learning from Failure’ For Engineering Students: Observation And Reflection on a Robotics Competition (RoboRoarZ-Edition 2).” 2023 IEEE International Conference on Teaching, Assessment and Learning for Engineering (TALE), Auckland, New Zealand, 2023, pp 1–5. https://doi.org/10.1109/ TALE56641.2023.10398407.
- Huang, Qian, et al. “How To Understand ‘Failure’ in Engineering Design Thinking Course.” The 10th HERA Conference, Taipei, 5–6 Jun 2024.
- Kaur, Ameek, et al. “Pedagogy Based On ‘Deliberate Learning from Failure’ and ‘Culture Setting’ Approach.’” Teaching and Learning Conference of Academy of Management Conference 2024.
- Kaur, Ameek, et al. “Teaching Design Thinking to a Large Cohort, A Process Perspective.” 2023 IEEE International Conference on Teaching, Assessment and Learning for Engineering (TALE), Auckland, New Zealand, 2023, pp 1–4. https://doi.org/10.1109/ TALE56641.2023.10398367.
- Miller, Claire Cain, and Courtney Cox. “In Reversal Because of AI, Office Jobs Are Now More at Risk.” The New York Times, 30 Aug 2023. https://www. nytimes.com/2023/08/24/upshot/artificial-intelligence-jobs.html.
- Mollick, Ethan. Co-Intelligence: Living and Working with AI. Penguin Publishing Group, 2024.
- Phelps, Edmund. Mass Flourishing: How Grassroots Innovation Created Jobs, Challenge, and Change. Princeton University Press, 2013.
- Poon, King Wang, et al. Living Digital 2040: Future of Work, Education and Healthcare. World Scientific, 2017. https://doi.org/10.1142/10725.
- Poon, King Wang, et al. “The Future of Expertise: From Stepwise Domain Upskilling to Multifaceted Mastery.” International Handbook on Education Development in Asia-Pacific. Springer, Singapore, 2023, pp 1-19. https://doi. org/10.1007/978-981-16-2327-1_42-1
- Willems, Thijs, et al. “Embracing Failure in Engineering Education: A Comparative Study of Design Thinking Approaches.” The 20th International CDIO Conference, Tunis, Tunisia, 10–13 Jun, 2024.
- Willems, Thijs, et al. “Assessing Employment and Labour Issues Implicated By Using AI.” AI Impact Assessment, Oxford University Press, forthcoming.
POON KING WANG
Poon King Wang is the Chief Strategy Officer of the Singapore University of Technology and Design (SUTD). He is also concurrently the Director of the Lee Kuan Yew Centre for Innovative Cities.
DR THIJS WILLEMS
Dr Thijs Willems is a Research Fellow at the Lee Kuan Yew Centre for Innovative Cities at SUTD. His main research interests are in the Future of Work, and he conducts ethnographic research to study how new technologies such as AI impact the practices and expertise of workers.
DR HUANG QIAN (CATHY)
Dr Huang Qian (Cathy) is a Research Fellow at the Lee Kuan Yew Centre for Innovative Cities at SUTD. Her main research interests are in educational research covering engineering education, AI in education, and blended learning.
JULY 2024 | ISSUE 12
NAVIGATING THE AI TERRAIN
