FA 107: The Creative Process in Visual Art: AI Tools and Resources

This is a brief overview of better known Large Language Models (LLMs) as of September 2024. Initially, they were available as free, beta versions but many are moving towards paid plans for improved versions. 

Ethical questions:

• As these tools are monetized, are they creating more social inequities by favoring those who can afford to pay for access?
• Do these tools protect privacy and intellectual property? For individuals, what do the Terms of Use or FAQs say? Is the default to use what you enter for their ongoing training, and potentially, in the output they provide to other users?

ChatGPT 3.5 - free; ChatGPT Plus has monthly fee 
Developed by Open AI
Chat GPT 3.5 responds to text prompts, trained on data through September 2021, and is not connected to the Internet.
Chat GPT Plus runs on GPT-4, responds to image and text prompts, trained on data through April 2023, and has a browsing tool to search the Internet.

CoPilot - GPT-4 is available to all UP users. Follow the link (or find it on the MyApps page) and sign in with your UP email and password.
Developed by Microsoft 
Runs on GPT-4, responds to image and text prompts, trained on data through October 2023, and connected to the Internet.

Gemini - Gemini 1.5 free with a Google account; Gemini Advanced has monthly fee
Developed by Google
Responds to image and text prompts, training cutoff date is confidential, "continuously being updated and improved," and connected to the Internet. 

Claude - free with account sign-up; Claude Pro has a monthly fee
Developed by Anthropic using their model which has a Constitutional AI method to ensure safety and reliability.
Responds to image and text prompts, trained on data through April 2024, and is not connected to the Internet.

Ethical Questions:

• How can you use AI tools and maintain academic integrity (including avoiding plagiarism)? 
• Are AI tools impacting your education? Do they undermine your opportunity to learn or do they enhance your understanding?
• Are AI tools providing biased responses? 

Generative AI can be useful for:

• Developing ideas for a topic and identifying related concepts.
• Suggesting which library databases might be useful for finding sources, and which keywords or search strategies to try when searching those other databases. (You can also get human help with all of these tasks and more from the Clark Library!)
• Providing suggestions for improving writing. (You can also get human help with writing from the University's Writing Center!)
• Creating translations and asking questions in another language.
• Assisting with computer coding tasks. 

Limitations of current generative AI:

• Reliability: They can't assess the accuracy or logical consistency of the text they generate. They can make up facts and credible-looking sources, complete with citations to nonexistent books and articles. This phenomenon is often referred to as a "hallucination" or "confabulation."
• Bias: They perpetuate systemic biases. They are trained on datasets scraped from the Internet, which includes all the racist, sexist, ableist, and otherwise discriminatory language and images found there. 
• Legal Violations: They can infringe on copyright and intellectual property.

What can you do? 

• Develop your fact-checking skills so you can spot and correct errors that generative AI tools might make. Clark Library staff can help! 
• Verify citations and sources provided by generative AI tools. Not sure how? Ask your helpful Clark Library team!
• Develop depth of knowledge in your discipline or areas of interest so you can identify erroneous or biased information.
• Ask other human experts like your professors or teaching assistants.
• In general, learn more about the tech tools you use. Find out how they are trained, their purpose, and their limitations.