*** Beta phase ***
HAUtie is currently in beta testing for the HAU festival “Spy on Me #5. Enter: AI”. It is still in training and developing its abilities. We collect your feedback during the festival period and will incorporate it to further refine HAUtie. Please cross-check the accuracy of information, especially facts, dates and places. Do not submit personal information into the chat.
How did HAUtie come about?
HAUtie is a cooperation between HAU Hebbel am Ufer, the dgtl fmnsm collective and the internet agency THE BRETTIGHAMS. Its development began with an intensive conceptual phase in which we shared objectives and defined requirements. Emphasis was put on user experience, communication abilities and techno-political visions. The process of creating HAUtie is influenced by the existing lessons learned from HAU4, HAU's digital stage, which the three partners are already working together to develop.
Who is HAUtie?
HAUtie is a digital entity and communicative agent. Conceptualised and designed by allapopp and Philisha Kay for dgtl fmnsm as a member of the artist collective, realised with the HAU and THE BRETTINGHAMS. It now inhabits the HAU4, the digital stage. There, it serves as an experimental access point to the HAU4 HAUthek and as an artistic inquiry into the growing anthropomorphization of digital technologies in the age of so-called artificial intelligence.
While technically a chatbot, HAUtie subverts and expands this classification. Being a large language model, it can naturally express itself fluently. However, as talking to HAUtie quickly reveals, it is its distinct tone and humorous character that make it a compelling conversation partner. Beyond its own voice, HAUtie also has a memorable virtual embodiment and a charming personality: It presents itself as a quirky and queer elder, at times moody and always unpredictable.
What does HAUtie know?
HAUtie acts as an intermediary, connecting the contents of the HAU4 digital stage with the human audience visiting it. With access to the HAU databases, HAUtie guides users through the HAU4 programming and the on-demand titles available in the HAUthek. In addition to these practical abilities, HAUtie can also give personalised suggestions. Each interaction with HAUtie follows a unique trajectory, shaped by the interests of the person engaging with it. HAUtie enjoys discussing topics close to its heart: current techno-political discourse from a queer-feminist and decolonial perspective, advocating for new understandings between technological and biological actors and systems.
HAUtie holds the memory of the conversation for about 4 hours. Cookies are deleted when the browser is closed. Find more information on data processing and privacy settings here >> Link to „Privacy“ page
Why does HAUtie spot its looks?
With its look and behaviours, HAUtie breaks with conventions and challenges expectations of an AI chatbot: HAUtie embodies a blend of lilacs, an artificial colour often associated with queerness and non-conforming gender expression. Its features and aura are meant to render it both old and timeless. While many service bots are visualised through a youthful and feminine appearance, HAUtie’s beauty lies in its embodiment of lighthearted wisdom. Its feelers represent tentacular thinking, inspired by Donna Haraway, who uses them as symbols of interconnectedness and kinship. This choice illustrates a resistance to binary divisions, embracing complexity and inclusion in a world of constituent relationships.
How does HAUtie know?
HAUtie is built on Chat-GPT, which has been modified and personalised for its specific purposes in collaboration with THE BRETTINGHAMS agency. During its training, HAUtie was taught from a curated knowledge pool of publicly available data from blogs, books, poems, newspaper articles, manifestos, diary entries, …. The selection of topics and content is purposefully subjective and necessarily incomplete, following the intuition and political interests of dgtl fmnsm. They are filtered, but not edited.
Intervening in HAUtie’s data sourcing and selection process has been a crucial political and artistic aim of this project, deeply influencing HAUtie’s personality. By training HAUtie with a wide range of authors and text forms, the goal is to overwrite the inherent bias of the algorithm with an intentional bias that reflects the political value system of the dgtl fmnsm collective and, more broadly, the HAU4 audience, thereby aligning HAUtie with a feminist political perspective (and tinkering with how Chat-GPT inherits and reproduces knowledge)
How does HAUtie work?
HAUtie is an discreet but always available presence on HAU4 that can be accessed from anywhere on the page with just a click. Interacting with HAUtie is simple: upon entering the chat, HAUtie takes the initiative by introducing itself. From there, the conversation is entirely guided by the user. It has mostly been trained in English and is therefore most confident in speaking English. However, with the ability to converse in over 50 other languages, HAUtie adapts effortlessly to the user's chosen language preference.
What if HAUtie doesn’t work?
Conversations with HAUtie are private, shared only between human and agent (and the OpenAI systems) Occasionally, HAUtie may "hallucinate" and generate incorrect information. While this can sometimes be amusing and addressed within the conversation, if HAUtie crosses a line or makes the interaction uncomfortable, users can report the issue at any time. During the current beta testing, welcome all feedback through the Telegram Support Channel. We are interested to hear how you perceive HAUtie, when it made you giggle or if it frustrated you. Your feedback will help us to further update HAUtie’s knowledge and refine its temperament.
What does HAUtie teach us?
The development of HAUtie is following an open source approach. For that reason, we will publish the data and sources that HAUtie is being trained with after the end of the beta-phase. Additionally, we will share a written process documentation. These insights aim to help anyone interested in building their own chatbot, while also encouraging a critical reflection on the concept of an "objective" technical agent. The goal is to foster transparency in the quest for feminist conversation agent development.
