Generative AI

I'm concerned about being falsely accused of using AI to plagiarize an assignment. How can I prove that my work is my own?

Document your writing process. The quickest and easiest way to prove that your writing is your own is to "show your work" - or, in other words, to provide evidence of the steps you took to complete the assignment. More complex assignments like writing a research paper tend to involve a lot of steps and may naturally create a "paper trail" of documentation that can be used for this purpose - brainstorming pages, research notes, multiple drafts, an outline, a folder of sources you found while researching, etc. - but it may be helpful to take additional steps to document your work, especially for simpler assignments that don't require much planning or outside research.

Here are a few tips:

• Consider writing your assignments in Google Docs (which automatically keeps a record of changes to each document) or turning on the "Track Changes" feature when writing in Word. 
• Save each new draft as a separate file. 
• Hold on to any hand-written notes or drafts until your assignment has been graded and returned to you.
• Be prepared to discuss your assignment in detail. Make sure you're able to explain your reasoning, the sources you used (and why you chose them), and your grasp of the core concepts being assessed.

Check out the video below for more tips on how to handle false accusations of cheating with AI.

What privacy and security concerns should I consider regarding AI?

Data collection
When users input personal or confidential data into generative AI models, this data is often stored, processed, and used to further train the model. This could inadvertently expose sensitive information that users did not intend to share publicly. It's important to always be careful about the information you share and to be mindful about how it may be used.

Even if you don't share any personal data with AI tools, they may already have your private information! LLMs are trained on large datasets, often obtained through web scraping, which can include a wide range of personal information such as names, addresses, birthdays, and other contact details. There is a risk that they may share any sensitive data gathered this way with users or that the data may be viewed by developers.

Generative AI platforms are also susceptible to data breaches, which could expose private data collected through user input or web scraping to unauthorized individuals.

Read the Terms & Conditions for your preferred AI tool(s) to find out what user data is being stored, how it may be used, and whether you can opt out of participating in the collection of certain private information.

Deepfakes
Generative AI allows users to create realistic videos and/or audio designed to impersonate a specific individual, which can be used to gain unauthorized access to accounts, perpetuate scams, spread disinformation, or damage reputations.

What ethical considerations should I make when considering AI?

Bias and Discrimination

Generative AI reflects the biases of its developers, of users who are contributing to training the AI model, and of the documents on which the model was trained. This can lead to the perpetuation and even amplification of existing societal biases. It's crucial to be on the look out for bias and discrimination in AI outputs.

Environmental Impact

Training and using generative AI requires substantial energy consumption and water usage. The high electricity usage of generative AI and the data centers necessary to operate these tools results in increased carbon dioxide emissions. AI models are also reliant on high-performance computing hardware, and the manufacturing and disposal of this hardware have an additional resource cost and contribute to electronic waste, which can pollute the environment if not managed properly.

Exploitative Labor Practices

Generative AI models require massive amounts of data to learn patterns, and that data needs to be meticulously categorized, labeled, and verified by humans. This work is often outsourced to workers in the Global South or to gig economy workers who often receive low wages and lack benefits, job security, and labor protections.

Copyright and Intellectual Property

Many AI models have been trained on massive amounts of text, image, audio, and video content scraped from the internet, often without the knowledge or permission of the original creators of this content. This raises questions about consent, copyright, and compensation for the original creators whose work is used to train these models.