Abstract
Restructure is an ongoing reality in today’s world, and we must continually adapt to rapidly changing conditions. Higher education is not exempt, and programmes are constantly influenced by external factors. Recent technological advances, such as the rise of self-guided video tutorials and Artificial Intelligence (AI) tools, have had a profound effect on education and allow anyone to learn independently at any time. Does this ready access to knowledge spell the end of higher education in lighting? Not at all. In my opinion, we need formal education more than ever before.
Let us start with a simple question: Who invented the light bulb? Like many before them, children today are likely to announce that it was Thomas Edison. In reality, the answer is more nuanced, but it is common to reduce children’s stories to a single figure over an ensemble cast of historical inventors. Now, what happens when we ask an AI Large Language Model (LLM) about Edison’s inventions? As it happens, a recent paper by Huang et al. uses that very example to discuss the accuracy of LLMs. 1 In response to the prompt: ‘What are Thomas Edison’s main contributions to science and technology?’, they give the following incorrect response: ‘Thomas Edison developed the first practical telephone and invented the light bulb’. Huang et al. credit this type of factual error to the data used to train an LLM. After all, the notion that Edison was the sole inventor of the light bulb is a widespread misunderstanding. This type of error from an LLM is known as hallucination – it generates facts that sound credible but are not correct.
Now, consider a lighting practitioner using an LLM, such as Claude or GPT-4, to find technical information that is vital to their work. Based on the above example, there is much potential for errors to occur. However, it is even more troubling to learn that hallucination extends beyond factual contradictions and can also fabricate information sources that might give an extra level of false confidence in the information provided. 2 In my experience, these fabricated references often have impressive titles that create plausibility. They also claim to be written by experts in the field, which strengthens trust in the idea and may lead the unwitting practitioner to incorrect conclusions.
Of course, LLMs are continually improving, and my recent query to GPT-4o found Edison’s history to be well-described. But LLMs are only as good as the training information, and concerns over hallucination are very real. Generative AI tools extend beyond LLMs and are becoming increasingly adept at all manner of lighting-related situations, such as exploring design concepts and planning spaces. Given the ease with which an inexperienced person can use these tools, there is an inherent risk of poor knowledge translating to poor design. It follows that sound foundational knowledge is essential to sift the fact from the fiction. Could that knowledge be gained from searching online rather than pursuing higher education? Yes… and no.
While it is true that one could keep searching online to verify information, there may be a nagging doubt that the information is still incorrect. An element of confirmation bias may creep in, where flawed content is accepted as it validates the initial idea. My own experience has led to many never-ending loops with an LLM conceding that it has provided flawed information before again insisting that it is accurate. This highlights the instrumental role of lighting educators in providing a solid foundation of essential concepts. Using their expertise, educators can create a structured learning path for their students that supplements traditional teaching with a curated selection of high-quality online resources such as standards, eBooks, video clips and industry websites. Taking this a step further, customised LLMs can be developed using Retrieval-Augmented Generation or by fine-tuning existing models using a knowledge base of trusted material. 3 Using these techniques, hallucinations are greatly reduced, especially when system prompts are used to mitigate this risk. By utilising a reliable reference collection or customised LLM, students can avoid inadvertently forming biased or fragmented views and are more likely to receive relevant, trustworthy knowledge.
In addition to providing an ever-evolving curriculum of relevant resources that adapts to real-time advancements, the human element of higher education offers extra benefits that cannot be replaced by technology. As well as providing knowledge, educators need to create a supportive environment that encourages collaboration and provides necessary feedback. Future lighting practitioners should feel motivated to critically analyse their choices and determine whether they have a sound solution. If robust foundational knowledge is combined with the ability to critically use that knowledge, we create an antidote to misinformation. By validating their work, lighting practitioners will feel ownership of and accountability for the resulting design.
When navigating new technological challenges, educators are as much lifelong learners as their students. They must keep pace with lighting techniques and technologies, together with ongoing developments in teaching methods. Pedagogical innovations are often supported through teaching and learning specialists, while knowledge in lighting can be gained through undertaking peer-reviewed research, receiving supervision by industry experts and accrediting bodies, and having regular interaction with standards bodies and local committees. Such accountability and oversight provide extra assurance that students receive the skills necessary for their careers. Those at educational institutions also benefit from access to enhanced resources to aid their learning, such as paywalled journal papers and lighting standards. There is now a strong additional need to equip the academic community further with paid AI tools to promote digital literacy.
Thankfully, as well as introducing challenges, technological advancements have made higher education accessible to a wide audience through virtual learning environments. The move to digital teaching was accelerated by the COVID-19 pandemic, when students and educators alike were thrust into an online world. Students are now used to the flexibility of online learning, and it is common to use a blend of online, in-person and hybrid methods. Students can tune in asynchronously to watch pre-recorded material, and synchronously to gain vital feedback and interact with their peers. Virtual field trips are now possible through software tools, breaking down the barriers of distance and time, and enabling students to balance their studies with family responsibilities. 4 Online environments give the structure and depth of a traditional classroom setting, together with the flexibility of a programme that fits around students’ work schedules. Where possible, supplementary hands-on workshops are the cherry on top, providing students with the opportunity to build relationships and give real-world application to their knowledge.
In conclusion, we must embrace new technologies and the benefits that they bring. However, I suggest that higher education will always need a human touch to guide and nurture students. People always have been, and always will be, at the core of education. New technologies offer exciting ways to enhance, but not replace, higher education in lighting.
