✋ Hallucinations Are Only One Potential Problem with AI
Strengthen your AI literacy by understanding these additional AI reliability pitfalls:
- Poor quality of source data (garbage in, garbage out)
- Seemingly small changes can cause models to drifting away from the relevant content
- Too many or too few humans in the process
- Using prompts that don’t include enough context for the model to work with
- Working with a model that someone else has tweaked or updated
Learn More: Murray, S. (2025) Top 5 AI reliability pitfalls.
3️⃣ Three Types of Faculty AI Anxiety
Higher levels of academic integrity among academics correspond to lower use of ChatGPT in their work. Researchers identified three types of tech-related anxiety impacting behavioral intent to use AI:
- Fear of harming student learning (most influential),
- Fear of misusing the tool themselves, and
- Fear of professional displacement (least influential).
💰 OpenAI Begins Monetization in Search
Open AI is framing their new search and shopping model as “search enhancement” but this means that, like Google, retailers can pay to have ChatGPT show users their products. Will this negatively impact search features or model behavior? All signs point to YES.
Learn More: Improved Shopping Results from ChatGPT Search
👍 Beware the AI sycophant
Many AI chatbots are rewarded when the user rates an answer with a thumbs-up. This is RLHF at work (Reinforcement Learning from Human Feedback). When an AI says you are wonderful, it might be a hallucination! It’s long as you continue to “Like” the compliments, it will pile them on to try to keep you longer in the interaction.
Learn More: Sycophancy is the first LLM “dark pattern”