Generative AI, or GenAI, is a type of computer program that creates new content, such as text, images, audio, video, or code, by recognizing patterns in the data it was trained on. It responds to user prompts based on what it has learned, but its output depends on the quality of its training data and the prompt. Human feedback helps improve accuracy and reduce bias, especially when correcting errors or unrealistic results (also called hallucinations).

Though there's a lot more to it, GenAI works like this:

1. The AI model is trained on human-created data.
   Examples include writing, images, code, and more. These inputs help the model learn how people express ideas.

2. The system looks for patterns.
   It analyzes the data to understand how words, visuals, or sounds typically appear together.

3. You interact with the model by giving it a prompt.
   A prompt is a task or question—like “Summarize this article” or “Suggest a title.”

4. Generative AI creates a response.
   It uses what it learned during training, plus your prompt, to produce new content.

5. You evaluate the result.
   The response might be helpful—or it might miss the mark. If the training data was biased or your input unclear, the output could be wrong. This is called a hallucination.

6. Your feedback helps improve future results.
   Giving a thumbs up or trying again helps fine-tune how the AI responds over time.

AI models learn by studying patterns, similar to how humans get better at recognizing things through examples and repetition. For instance, you might learn to identify different types of flowers by seeing them in pictures or at the store.

• They're computer programs designed to find patterns in data.
• They learn from examples instead of being explicitly programmed with rules.
• Once they’ve been trained, they can make predictions or decisions when given new information.

Just like humans improve with practice, AI models get better with data and training.

Generative AI tools are not sentient and rely on data provided by humans. Remember:

• An AI is only as good as the data it’s being trained on.
• It is dependent on humans to “train” it by:
  • Reviewing and providing feedback on its output.
  • Ensuring the data the AI has access to is up to date.

Start simple.

1. Pick a task you do often (like summarizing notes).
2. Choose a tool from CMU’s approved list.
3. Ask it a clear question or assign it a task.

That happens—it's called a hallucination! If one occurs, you can:

• Rephrase your question.
• Upload supporting sources of information to provide more context.
• Enter a chain-of thought prompt where you ask the AI to explain its reasoning. 
  Explain your reasoning step by step, including complete links to sources. Outline how you reached your conclusion.

Yes—if you're using CMU-approved AI tools while logged in with your Andrew userID.
These tools:

• Meet CMU’s privacy and security standards

• Are FERPA-compliant

• Prevent unauthorized access to your personal or academic data

• Does the response answer your question or complete the task you gave it?
• Does it back up claims with facts, examples, or relevant and trustworthy references when needed?
• Can you tell how or why the AI came to that answer? Is the reasoning clear?
• Does it go beyond the obvious and offer helpful ideas, summaries, or patterns?