This post is the second section of a 3 part series, created by members of the Mayor’s Youth Council, featuring discussions and ideas regarding the upcoming technologies of AI used in academic settings.
Introduction: Why is it Important to Regard the Ethics of AI Usage?
With issues around AI such as job loss of individuals as well as over-reliance on AI tools in academic settings, how does this affect society as a whole? It is inevitable that AI tools such as generative AI, artificial general intelligence, and artificial narrow intelligence assist us in our daily lives, whether through studying, brainstorming, or working. Whether affecting the environment, user data and privacy, or even the created bias through shallow level AI training, it is important that users are informed about the decisions they make regarding AI usage.
The Environment
To work, generative AI models require building and training, requiring the use of significant amounts of energy, contributing to carbon emissions. Additionally, the use of AI requires significant amounts of water for cooling. Powerful AI chips generate great amounts of heat, and require cooling. To cool AI data centers, water is used in cooling towers, and evaporates during the process, making it not able to be reused. According to studies, a single AI prompt (depending on model, prompt length, server efficiency) consumes small amounts of water between 0.26 – 500 mL. However, when tech giants such as Google report staggering amounts of up to 5 billion searches per year, the small amounts of water slowly add up. Although researchers as well as companies are finding ways to make AI more viable for the future, it is important for you as the user to consider whether the use of AI is worth the environmental impact, and whether you can use tools as efficiently as you can.
The Use of AI – Data / Privacy
When using AI tools, it is important to think about how the information you submit to the AI, whether on purpose or not, can infringe on user privacy. AI is trained through the use of data whether being copyright-protected material, or licensed materials with rights the user has acquired (or needs to be acquired)
By submitting content to AI platforms, the user essentially grants the AI tool the right to reuse, and distribute the content, which may be a breach of copyright, and privacy for the original owners of the content. Similar to other digital tools such as Microsoft 365, or Google Workspace, generative AI stores and collects data about users based on the inputted information. Signing in and making an account with AI tools such as ChatGPT allows the companies to collect personal information relevant to the user, used to make changes to tools to keep the user engaged, using the tools. User data stored within AI tools may also be sold, or granted to third parties for marketing purposes. To ensure user privacy, users should exercise caution when giving sensitive information to AI tools.
Bias Within AI
Bias within the model training can create one-sided generated information. If the training data used to instruct the AI is biased, or the model design is not designed to respond to diverse inputs, the AI model may produce biased information. Optimization techniques used when training AI tools may favour popular predictions over minority views. If the data collection used to train the AI favours a particular view on topics that are not diverse, resulting generated information will reflect the biases the AI was trained with.
Questions to think about when I use generative AI:
- How can I be sure that the information presented to me is reliable, and credible?
- How can I address the possible bias towards one factor in the information I generated?
- How can the information I shared with AI be used and stored after the generation of content?
Despite the many drawbacks brought up by this article, AI has the potential to greatly influence and positively affect society. The next article, “AI as a Research Tool” will cover the frequently asked question: So what can I do about it?
Sources
https://professional.dce.harvard.edu/blog/ethics-in-ai-why-it-matters/#The-Importance-of-AI-Ethics
https://katiecouric.com/news/how-much-water-does-chatgpt-use/
https://searchengineland.com/google-5-trillion-searches-per-year-452928