Generative AI: Home

There are many ethical concerns at the core of generative AI use and development. In this guide you will find the following topics addressed in more detail:

Copyright: Generative AI tools are trained on data sets. Data sets are comprised of language, images, and other forms of data. Where does all the material used to create a data set come from? For some generative AI tools, it is harvested from the internet and other public sources - without the consent of the people who originally posted it. Further, when the generative AI tool creates new images and text from these data sets, the creators of the source material are not credited.

Bias: As mentioned above, the data sets used to train AI tools are often harvested from the internet and other public sources where people have posted their thoughts, creations, and ideas. Humans are inherently biased. (See the "Bias" tab for a more expanded exploration of these ideas.) When biased information is used to create new material, it recycles - and compounds - the biases that are threaded throughout the original materials. These biases manifest in many ways - including racism, misogyny, ableism, and other forms of discrimination.

Labor Concerns: AI is also changing the way we work. While the use of it may lighten some workloads, it has already been shown to replace people completely at their jobs. It has also been used to bar people from employment when used during the recruitment stage. AI companies raise labor concerns because they pay some of the lowest wages in the world in jobs that are intermittent and not guaranteed.

Environmental Concerns: The energy used by AI is so massive that its usage is often measured in countries rather than kilowatt hours. This is making the transition to renewable energy much slower. The data centers that power AI need so much water to run that they have drained whole towns of their water tables. Read more about this topic on the "Environmental Concerns" tab.

Access: Access to different AI tools come with different price points, or, one tool may have different subscription levels for different fees. This creates a disparity in access for people who can and cannot afford different levels - and/or qualities - of generative AI tools. Further, consider a future in which people have ingrained generative AI into all elements of their lives, their actions, and their thinking - and then are beholden to pay whatever fees are instituted to maintain their daily lives. Consider also that even though something might be advertised as free - monetarily speaking - that tools can harvest other information from you that can then be monetized. For example, selling your shopping habits to a retailer who can use your information to better target you as a potential customer.

The further consolidation of power: Imagine that people have fully absorbed generative AI into their lives. This could happen in many ways - intentionally or not - some examples being:

• by ceasing to diversify their information sources;
• by outsourcing their thinking to generative AI tools, rather than doing the sometimes uncomfortable or challenging work of thinking for themselves;
• by forgetting - or not learning - skills that they allow generative AI tools to do for them instead.

They could become dependent on an even smaller array of tools created by an even smaller group of people. This consolidates power (and the derivative wealth) into the hands of the even fewer people - who are in control of developing the tools people live by and are dependent on.

This guide was originally created in Fall 2024 by SVA Librarians Shea'la Finch & Caitie Moore with contributions from David Pemberton.