Ethical Considerations
Copyright Infringement and Inherent Biases
- GAI derives its output from previously human-created texts and images from other sources that their models are trained on. Although it may cite sources, it uses that mass of training material – the work of others – without permission or acknowledgement, in many cases violating copyright laws and privacy expectations. It may also replicate the biases of that material.
- ChatGPT may give you an output that looks and reads okay but can perpetuate social biases (gender, race, etc.).
Environmental Impacts
- Every chat contributes to global warming. According to Goldman-Sachs, a single ChatGPT interaction uses ten times the electricity of a single Google search.
- According to the International Energy Agency, electricity demands from the data centers that power GAI will double in the next two years, consuming as much energy as Japan.
- The U.S. Dept. of Energy estimated that in 2014, U.S. data centers alone consumed 14% of daily freshwater use in the country, where they are often located in areas already facing water shortages.
Social Justice
- Almost as soon as ChatGPT-3.5 was released in November 2022, articles began to appear around its controversial use of labor. In January of 2023, Time magazine ran an article about Open AI's contract with a company in Kenya that paid workers $2.00/hour to screen highly objectionable material for the company. Many of the workers reported feeling traumatized by the viewings. Workers in other parts of the globe also have been exploited by big tech companies.
- Another fear is that the digital divide will expand even more between those who have access to computers and AI and those who don't. As Open AI continues to monetize ChatGPT, some will be able to afford it while others will not.
