Today we have interviewed our Gema Parreño – Lead Data Scientist at a software development company – Apiumhub, where she develops Data-Driven Solutions. She is passionate about the intersection of machine learning and games , and has had her own startup , contributed in the open source space in StarCraft machine learning project, and had an experience at Google Brain for Stadia.
Gema gives a talk about Mempathy as an AI Safety and Alignment opportunity and we wanted to dig deeper and find out more about it and how the idea arose to use it for implementation of Safety and Alignment techniques.
Gema shared with us that she has experience in the field and believes that serious games are a subfield inside the game industry that are designed with a purpose distinct from pure fun that might include a huge variety of topics ranging from education to mental health. Her Mempathy project is a video game narrative experience that transforms the relationship with anxiety and helps overcoming unrealistic standards of perfection.
The road to use machine learning is an ongoing, rigorous, and creative process. Previous works have explored Reinforcement Learning with a system that focuses on reward design based on responses , and Imitation Learning. However, in these explorations, concepts such as scale and bias came in. Using Large Language models would create a really personalized experience for a player.
Inside the evolution of the project and from a machine learning perspective , Mempathy explores the research question of behaviour alignment problem with GPT models from a Game Design and Content generation perspective, aming to solve the question how do we create agents ( Non playable characters ) that behave in accordance with the designer´s intention.
One of the key points here is creating experiments that could help benchmarking different solutions at scale. The vector thought of ‘how do we design an experiment to actually test an hypothesis’ that is aligned with Mempathy NPC game design. There has been an ongoing evolution from designing the system from scratch to actually exploring the fascinating topic of controlling large language models.
Producing a full prototype taking into account all the game design principles together with the AI system has been hard, as it required Gema to wear multiple hats with a ‘good enough result’ to actually test the idea from an holistic point of view.
Also, NPC design has been an active field of research and study inside the videogames industry. If you are willing to have a controlled exploration , Gema recomms this book for giving a general overview about this field: Artificial intelligence and Games. G. Yannakakis & J. Togelius . Chapter 5 . Modeling players
From the NPC design perspective with the focus on AI, there is a full opportunity to automate all design processes, that goes from animation to content generation . Gema believes that many new ideas are yet to come, adapting research coming from robotics, simulation, and natural language processing.
As for the Plug and Play Models, they combine a large, pre-trained large language model and an attribute model that guide text generation without any further training, allowing flexible controlled text generation while maintaining fluency . This solution has narrowed down content generation quite consistently with respect to content generation with GPT from 35% to 1.5% misaligned and harmful content.
And she is looking forward to share details about key parameter search for the proposal at day 1 in Data Innovation Summit this year.
It might be also important to note that proposing a cooperative interactivity from the game design perspective has been proven key with a symbolic representation of language, as unrestricted dialogue methods remain being hard for evaluation.
As a conclusion , Gema could say that the lessons learned come from successfully implement game design mechanics that could be optimal for the system, and choosing a strategic content generation technique that can be aligned with NPC motivations.
Gema is looking forward to connecting with you in Data Innovation Summit edition or collaborating with you on your Machine Learning Project. Contact us if you need more details about our Data Science solutions.
