GenAI – Limitations and Considerations

GenAI can be an invaluable tool for enhancing education, offering numerous benefits. However, it is important to be aware of the limitations and considerations when using GenAI in your studies.

Protecting your and others’ information and rights

• Read and Adhere to Terms of Use: Before using GenAI tools, familiarize yourself with and follow their terms of use.
• Data Privacy:

• • Be aware that your input can be used to train the AI, meaning your data might be stored, analysed, and used in ways you might not expect.
  • Your data might be at risk of breaches or unauthorized access, especially when transmitted over unsecured networks.
  • Many GenAI tools integrate with other services, each with its own privacy policies and data handling practices, adding complexity to data privacy.
• Avoid Sharing Sensitive Data: Do not provide GenAI tools with:

• • Personally identifiable information such as names and addresses
  • Unpublished research data and results
  • Biometric, health, and medical information
  • Geolocation data
  • Government-issued personal identifiers
  • Confidential or commercially sensitive material
  • Security credentials
• Copyright Compliance: Use copyrighted material only with explicit permission from the copyright owner; and do not upload material from university library eResources.
• Opt-Out of Data Collection: Where possible, choose to opt out of data collection features in AI tools to prevent your data from being stored or used for model training or human review.

Consider the ethical implications of using GenAI

• Outsourcing of Content Labelling: In January 2023, Time reported that OpenAI outsourced labelling tasks involving sensitive content (sexual abuse, hate speech, violence) to a Kenyan firm.
• Worker Exposure to Traumatic Content: The outsourced workers were exposed to disturbing and traumatic content during their duties.
• Highlighting Hidden Human Labour in AI: The article underlines the often-unseen human labour involved in AI development, which can be exploitative and harmful.
• Ethical AI Principles: Emphasize the importance of adhering to ethical principles in AI use, including:
  • Well-being
  • Human-centred values
  • Fairness
  • Privacy
  • Reliability
  • Transparency
  • Contestability
  • Accountability
• Building Trust and Societal Benefits: Following these ethical principles can help build trust, foster loyalty, influence positive outcomes, and ensure AI brings benefits to society.

GenAI’s limitations

• If you are not specific enough, the AI will make assumptions. If unsure, you can ask the AI to see if any assumptions were made in the process of responding to your prompt.
• Try to use a neutral tone and framing in the way you present information or ask questions in your prompt, to avoid introducing bias or leading the AI model towards a specific response.

Hallucinations and Inaccuracies

• Generative AI is prone to hallucination: GenAI tools can “hallucinate”, meaning they fabricate facts.
• Questioning Authority: Just because the output from a GenAI tool sounds authoritative, it does not guarantee its accuracy.
• Impact of Inaccuracies: The consequences of factual inaccuracies vary based on context:
  • Low Impact: In brainstorming sessions, inaccuracies might be less problematic.
  • High Impact: In critical applications, like creating a medical technology user manual, inaccuracies could have serious consequences.

Equitable Access

• Cost: Cost of tools poses a barrier for many students in accessing generative AI tools. With many tools currently available for free, some have paid tiers with significant improvements in functionality and performance for paid subscribers. Those students who can afford to pay for paid tiers may be disproportionality advantaged in assignments that incorporate the use of generative AI.
• Regulation and censorship: Students may be studying remotely from a location where government regulation and/or censorship may restrict access to GenAI tools.

Environmental Impact

• High Energy Consumption: GenAI requires substantial energy for training and running large models, leading to a significant carbon footprint.
• Computational Resource Intensity: The process involves millions or billions of parameters, necessitating vast computational resources, often in energy-intensive data centres.
• Escalating Demand vs. Environmental Impact: Growing demand and scale of AI systems pose ongoing environmental challenges, despite advancements in efficiency and renewable energy usage.
• Hardware Production Footprint: Production of AI hardware components, such as GPUs and specialized accelerators, contributes to resource depletion and electronic waste.

Attribution: Generative Artificial Intelligence in Teaching and Learning Copyright © 2023 by Centre for Faculty Development and Teaching Innovation, Centennial College is licensed under a Creative Commons Attribution 4.0 International License; 

Attribution: Adapted from A Generative AI Primer JISC National Centre for AI Version 1.3 – 2nd Jan 2024. License: CC BY-NC-SA.