When considering using Generative AI tools, it is worth being mindful of the potential harms involved. While they may provide convenience there may be consumptive and predatory aspects as well. Some students and faculty may reject participating in the use of Gen AI on account of one or more of these issues. The categories below are not exhaustive but may provide some introduction to the considerations necessary. While you may employ Gen AI tools it is worth thinking about the output and applying these considerations as critical lenses when evaluating the output.
Many modern technologies take advantage of data and meta-data a user shares directly or inadvertently with its host company. This data may be used in the training of other AI tools, may be sold to additional third parties, or more. While it is good to be mindful of what you pieces of your personal information are shared with these tools and companies it is also important to respect the rights of others and not share their personal data.
The European Union is introducing regulation to limit some uses of AI and what data it is allowed to collect. If you are conducting research that may involve participants who may be European citizens it is worth understanding if some tools or approaches may be limited.
Large GenAI tools rely on a vast number of processors. These processors can consume a lot of energy and consume water to prevent hardware from overheating. Below are some articles explaining how some Gen AI tools impact the Earth.
AI is accelerating the loss of our scarcest natural resource: water - Forbes
AI power consumption: rapidly becoming mission-critical - Forbes
Software tools may position themselves to appear neutral and objective but they are not immune from bias. Generative AI tools are influenced both by the bias of their developers and the data upon which they have been trained.
Generative AI takes stereotypes and bias from bad to worse - Bloomberg
Generative AI tools required human moderation to set boundaries on content that is produced. The work that these moderators have participated in was often traumatizing as well exploitive. While these tools may out-source future moderation to other AI systems or employ more ethical practices there is no separating the fact that the groundwork for these systems was laid upon human exploitation.
'It's destroyed me completely': Kenyan moderators decry toll of training AI models - The Guardian
Large AIs have been trained on wide swathes of material, including copyrighted materials. The CEO of OpenAI has stated that limiting Gen AI to the public domain would undermine the technology completely. Copyright may be a legal quagmire until courts decide a number of questions. When using services you may want to see if Terms of Service agreements allow the service to use your work to train AI or sell your data to train AI. Additionally using AI may risk your ability to copyright your work or AI-generated parts of your work.
"Impossible": OpenAI admits ChatGPT can't exist without pinching copyrighted work - Salon