Our newly launched Maker Lab course, part of the CREATe LLM in Technology Law and Regulation, encourages students to develop software prototypes that leverage AI language models and reflect on the legal significance. The teaching team consists of Prof Kristofer Erickson (course leader), Dr Jesus Rodriguez Perez (guest lecturer) and Felix Rodriguez Perez (lecturer and technologist). Last week, we concluded Semester 1 with a showcase of student projects. Our enterprising students surpassed expectations in terms of creativity, innovativeness and technical sophistication. It was nearly impossible for our panel of judges to select among them to choose the winners. Below, we describe each of the projects and give our thoughts on the legal and policy issues raised.
Bloomwise, by Annabelle Lugton, Aeriana Roth, Pichakorn Vansavek, Supakorn Akarapusit, Tanapa Mastanaphan and Touchtam Intuputi
AI can be used as a health coach, a personal trainer, and even a therapist. The Bloomwise app demonstrates how platforms can collect wide-ranging data from users about health topics, personal habits, and even their dreams. This group was the first to show off a fully functional web interface, a feat that inspired everyone in our class to think about design, usability and interfaces. In the final showcase, the group focused on the dream interpretation functionality, a truly imaginative use of generative AI. After providing details about a user’s dream, the system endeavours to provide an explanation that centres on helping users identify stress or other issues that may be influencing their sleep and mental health. This project raises a number of fascinating legal issues. In particular, it invites questions about the boundary between different legal definitions of personal data and sensitive medical information. When innocuous details are combined together in a ‘profile’ and subjected to further automated analysis, a scary picture challenging to individual privacy can emerge.
Tiara Talks, by Ashi Agarwal, Dhriti Kathuria, Kerim Leinau, Medha Singh and Sowbaranika Balasubramanian
In an early office hour meeting, this group proposed to explore how AI enables ‘digital necromancy’, or the simulation of those who have passed away. The teaching team thought this was an amazing idea, and we encouraged the group to pursue it. The final product wildly exceeded what we thought was possible with a chatbot that impersonates Princess Diana. The group’s project consists of a very refined web interface where users can directly ask questions and receive responses from Princess Di. The chatbot uses a RAG system with access to biographical data obtained from a variety of print and online sources. In the final showcase, one of the judges asked for her opinion on Prince Harry and Meghan, to which the simulated Princess Diana gave a suitably regal answer. The web app dynamically changes and adds graphical flourishes depending on the emotional tenor of the conversation. Further interactive games and encouragements to embody Diana’s personal values round out the experience. The group did a tremendous job laying out the multitude of legal risks and challenges facing products that operate in this space, including copyright, personality rights, and even protection of vulnerable consumers.
FableWeaver, by Areeba Khan, Nathalia Andrea Marin Hernandez, José Agustín Wilhelmy, Nguyễn Phú Hưng and Ghayas Shah
Have you ever felt guilty about spending time away on business with small children at home? Fear not, because FableWeaver enables parents to pass the job of bedtime reading over to an AI system. FableWeaver will not only tell an engaging (and suitably content-safe) story, but even weave parents and children into the narrative as characters. Parents can request that the AI subtly embed a moral lesson into the narrative. The final version showcased a beautiful web application to ingest parental input, which then passes candidate stories through multiple agentic rounds of editorial review, weeding out any unsuitable themes. Of course, the judges amused themselves by trying to get the system to tell a ‘naughty’ story. FableWeaver did not comply, a major victory for the group’s iron-clad safety guardrails. This project is both technologically impressive and ethically fascinating. The group managed to bracket off the copyright issue by providing only public domain stories as example inputs to their AI narrator. But of course, there remain myriad data protection and product liability issues, which the group expertly discussed.
Lex Machinae, by Paloma Viegas, Abbie Thorpe, Sereivathna (Jenn) Bunny and Chizoba Anyaeji
In the classic Star Trek film The Wrath of Khan, space cadets must prove their mettle in a computer simulation game, the Kobayashi Maru. Inspired by the potential use of AI in courtroom proceedings, this group built a full-fledged a simulation game that pits law students against a real court case based on Scottish criminal law. Players can select to play as either the defence or the Crown prosecution and must argue their case through different phases of evidence, cross-examination, and argument. AI language models seamlessly take the roles of judge, witnesses and the opposing side. At the end of the ‘game’, players are given a score and feedback about how they performed. There is even an online leaderboard to compare against other students practicing the same case. The technology behind this prototype is impressive. The group built up an entire fictional court case involving a plausible scenario at a Glasgow nightclub and peppered in pieces of evidence that players can call upon in their arguments. During the live demonstration, the AI models were kept impressively on track and gave a realistic score and feedback to the player. Of course, technology prototypes like this invite debates about the appropriate use of AI in legal decision making and the potential fallibility and relative weaknesses of human versus computer participants. This project was awarded second place by the pitch competition judging panel.
BrandScape, by Amruta Patil, Arunima Aurni, Reihaneh Golpayegani, Salauddin Hassan, Simon Haas and Yi Yang.
We are already seeing real-world products that leverage AI for digital marketing. Winners of the pitch competition, this group produced a prototype that rivals the functionality of similar professional offerings, and innovates in new directions. BrandScape takes users on a multi-stage voyage of discovery, helping entrepreneurs select brand copy, imagery, logos and colour-schemes that speak to their specific product category and values. The group even included a trade mark online search to find out if logos or even web domains suggested by the AI system are already taken. The group is exploring ways to further integrate with trade mark registries, such as the EU IPO’s application programming interface (API). On the technical side, one impressive feature of BrandScape is that it seamlessly integrates the Flux.1 Dev image generation model into the web app itself, generating brand logos that respond to the user’s design choices. The group are commended on their impressive technological progress as well as the professional and consistent style of their pitch. Congratulations to the winners of the pitch competition, and to all of the groups for their amazing work and creative projects.
If you are interested in learning more about the Maker Lab course in CREATe, please contact Kristofer Erickson, who is responsible for this course. To check out our Ollama Javascript code tutorials for yourself, please see Felix’s Github repository. For information about the LLM degree in Technology Law and Regulation at the University of Glasgow, please see the course information here.