Skip to Main Content

FLEET LIBRARY | Research Guides

Rhode Island School of Design

Generative Artificial Intelligence

An overview for students and faculty

Ethical Considerations of Using Generative AI

There are several main ethical concerns regarding generative AI to keep in mind before using any software.

  1. Misinformation and deepfakes - generative AI software is known to produce false information and sources that sound authoritative at times. These instances are known as "hallucinations." It is important to double check any information you receive from a chatbot, especially sources. Generative AI that produces images may incorrectly interpret prompts and styles, and is capable of producing deepfakes - digitally altered images or videos of real individuals.
  2. Bias and discrimination - because generative AI is trained on existing information created by humans, it often reproduces the biases present in the scope and the text of its training data. This was a huge source of criticism when chatbots and image generators first emerged, and since then, companies have prioritized combating bias in a number of ways, including Reinforcement Learning from Human Feedback (RLHF).
  3. Copyright and intellectual property - in closely mirroring or appropriating existing artists' work, generative AI software risks infringing on artists' intellectual property. This is a murky issue that will undoubtedly continue to be contested. Additionally, many artists and other individuals have expressed concern, and started litigation, over generative AI companies using copyrighted material available on the internet in their training data.
  4. Accountability - many generative AI companies have not been forthcoming about their training data or policies. This makes it difficult to hold these companies accountable when ethical concerns such as the ones outlined arise. There is also the matter of user accountability when relying on generative AI. If you are using generative AI for a project or assignment, are you disclosing this use?
Image: Thinking Person Illustrations by Parth Atara is licensed under CC BY 4.0
References and Further Reading

Environmental Considerations of Using Generative AI

Training generative AI requires a significant amount of power. According to Business Insider, AI servers use anywhere from four to six times as much electricity as cloud servers. The projected demand for AI has led to the growth and construction of new data centers, mainly in rural areas of the United States. In light of environmental concerns about AI and data centers, members of Congress introduced a bill in February 2024 called the Artificial Intelligence Environmental Impacts Act of 2024 to determine and measure the impact the development of AI will have on the environment.

Power

To address the power needs of such data centers, some tech companies are turning to nuclear energy as a climate-friendly solution. In September 2024, Constellation Energy announced it would reopen its nuclear plant at Three Mile Island in 2028 and sell all of the power generated to Microsoft for the following twenty years. Amazon Web Services bought a data center next to the Susquehanna nuclear power plant back in March 2024 with an agreement to purchase power from the plant for the next ten years. Nuclear energy is reliable and carbon-neutral, but expensive and haunted by past disasters, such as those at Three Mile Island in 1979 and Chernobyl in 1986.

Tech companies have also invested in renewable (and non-renewable) energy sources in addition to nuclear energy, but there are still concerns that the amount of power used by these data centers could strain the electric grid and even threaten residential supply in the future.

Further Reading
Water

Another major environmental concern with data centers is the amount of water they require to cool their servers. Because training AI requires so much electricity, it also generates a lot of heat. The temperature of the environment external to data centers also contributes to this issue. Most data centers use air-cooling systems, which are themselves energy-intensive, but when temperatures rise above 85 degrees, a liquid-cooling system is needed. This becomes a particular problem in states like Arizona, where Meta, Microsoft, and Google have all built or are planning to build data centers, and where summer temperatures can reach 115 degrees. As perhaps an omen of future water woes, in June 2023, the Governor of Arizona had to halt new constructions in Phoenix that relied on groundwater because proposed projects would outstrip the projected groundwater supply.

Further Reading