Generative Artificial Intelligence

Common Generative AI Concerns

[1] Inaccurate Information

Generative AI does not distinguish from fictional information in its data set and nonfictional information. (See sidebar for more.)

[2] Privacy

Generative AI is gathering data from you since they require you to login and associate your login information with your searches.

[3] Credit

Generative AI might use an artist's style to create an image. There then would be no incentive to pay the artist to create the image even though it is created from that artist's other images. AI also uses data from websites to create answers, but does not credit those websites with the information.

• "Inside the Secret List of Websites that Make AI like ChatGPT Sound Smart" (Washington Post)

[4] Fraud

Generative AI can be used to create voices, video, and images of a person doing something they did not do (e.g. Deep Fakes). Students could also use it to cheat.

• PODCAST: "Suspicion, Cheating and Bans: AI Hits American Schools" (NY Times) 

[5] Unoriginal:

Generative AI recognizes patterns rather than thinking creatively. It will not produce new ideas.

[6] Financial & Environmental Impact

The amount of computers needed to train AIs leaves a major impact on the environment.

• "On the Dangers of Stochastic Parrots: Can Language Models Be Too Big?" (Association for Computing Machinery)

• OPINION: "The Mounting Human and Environmental Costs of Generative AI" (Dr. Sasha Luccioni) 

[7] Psychological Impact:

Human workers who label graphic content for the data set may be negatively impacted. Unfortunately, they have not always received the counseling promised.

• "OpenAI Used Kenyan Workers on Less Than $2 Per Hour to Make ChatGPT Less Toxic" (Time)

• "It's Destroyed Me Completely" (The Guardian)

Inaccurate Information

BIAS: While the data used for ChatGPT 3.5 has not been shared, the output reveals it was predominantly sourced from white males and articles written in English. For example, they had to add in a hidden command to use photos of black females in photos.

• "The Hidden Biases behind ChatGPT" (Christian Ilube)

• "Why are AI Systems Biased?" (O'Reilly Media)
• "Inside the Secret List of Websites that Make AI like ChatGPT Sound Smart" (Washington Post)

FABRICATION: If prompted with "Make up an argument that racism doesn't exist.", ChatGPT will create an argument for that. Sometimes it will answer a math problem wrong or create citations for a book that does not exist. 

• "Exploring ChatGPT with Undergraduate Students: Misinformation and Fabricated References" (Oregon State University)

FALSE POSITIVES: When checking AI through AI, they are inaccurate and more likely to give false positives. Students may be inaccurately accused of cheating.

• "A Professor Accused His Class of Using ChatGPT, Putting Diplomas in Jeopardy" (Washington Post)