How we use artificial intelligence (AI) including large language models (LLM) in our work.

No AI-Assisted Decision Making

Large Language Models (LLM) appear to be extremely knowledgeable, but we choose to rely on humans, preferably domain experts for all of our executive decision making. When it comes to hiring, firing, pricing, marketing, and other business decisions, we choose to use human input and allow human mistakes. While learning to repair bikes is the primary skill volunteers learn at Imaginal Cycles, running a nonprofit, dealing with customers, managing inventory, and other day to day decisions are also valueable skills to learn.

No AI-Art

While people who use our services are welcome to decorate their bikes however they like, we don't use AI generated art in our website, promotional materials, or branding. Imaginal Cycles was founded by artists, and we hope to attract more artists to our project. Even when the art shows signs of novice work, we prefer the DIY aesthetic as it matches the unique and personal aesthetic of our bikes.

Limited AI-Site Copy

LLMs shine as an auto-complete tool, and allow us to quickly generate draft copy for our website. Where entire sections of copy were generated by LLM, we hope to eventually replace that content with text that has a more human voice. Producing copy in this manner is fast, and allows us to focus on more important aspects of our work, like getting bikes into the hands of people who need them. Incidentally, the text on this page was written by hand, and not generated by an LLM.

We also use LLMs to do fast language translation into Spanish to make our services more accessible to our community. Some of us do speak Spanish, but don't have the time to translate all of our content and keep up with site changes. We apologize for any errors in translation, and welcome corrections from the Spanish-speaking community.

AI-Assisted Software Development

In order for our bikes and our prices to be community accessible, and for our donors to get the most value for their contribution, we use several software tools to assist our volunteers and partners in managing the shop. While AI tools have been contreversial in art communities, they are significantly less so among computer programmers. Much of computer programming is the combination of standard patterns in novel ways; there's already a strong tradition of cut and paste with minor modification in software development circles. When LLMs are used to perform this function, there are two concerns.

• Programmers are pushed to produce code unreasonably fast, and aren't given time by management to properly test and debug the resulting code. This results in unusually insecure and buggy code, data loss, and programmer burnout.
• People who are not programmers and do not understand the underlying code producing software that appears to work, but has deep underlying issues that require a deeper knowledge of software design to solve.

Programmers commonly complain about having to deal with a flood poorly-written vibe-coded additions to their codebase, or being pressured by management to approve LLM generated code at a rate faster than they can properly review it. Many of these programmers still use LLMs to assist them in their work, as an assistant rather than the primary source of code.

Environmental Concerns

We don't want to minimize the environmental impact of AI technology. We also don't want to treat it differently than other existing technologies that also harm the environment and carry other moral hazards.

As ecologically concious people, we are concerned about the environmental impact of petrochemicals. We still use cars, though we would prefer we had more opportunities to use bicycles and that cars polluted less. We use cars to transport bikes and parts, and to help volunteers to get to the shop. We take some comfort in the fact that our work is helping to reduce the number of car trips taken, and hope that we have a net positive environmental impact.

Similarly, we use some LLM technology to assist us in our work. Similar to car use, our relative contribution to the total environmental impact of the technology is small, and we are using it for the purpose of furthering an activity that has a positive effect on the environment. We don't endorse every use of LLM technology, and our opinion of what constitutes responsible use may change as it becomes better understood. The use of this technology by a small bike nonprofit does not justify the environmental harm of the technology, and we are investigating ways to reduce the impact our use of LLMs has on the environment.

We are part of a conversation

We are aware that our use of LLM technology is potentially controversial. We hope that is is a small enough part of our work that it isn't noticed, but we are committed to transparency .

Come see the place

The fastest way to understand what we are is to come and see it for yourself. Bring a bike, or do not.

Plan a visit Volunteer with us