Introduction to Generative Artificial Intelligence

Exam Tip: Remember the distinction—Traditional AI performs or recognizes (e.g., identifying a face), while Generative AI creates (e.g., drawing a face that doesn't exist).

Multimodal Capabilities: Text, Image, and Audio

While LLMs dominate the conversation, GenAI is "multimodal," meaning it can handle different types of media through text prompts—the instructions provided by a user.

Text Generation

Through text generation, users can produce essays, code, or summaries in seconds. This is particularly useful for professionals who need to synthesize large volumes of data or draft reports.

Image Generation and Digital Imaging

GenAI can transform a text description into high-quality digital imaging. Platforms like Canva allow users to generate visual content without needing professional graphic design skills. This is significantly more cost-effective and faster than traditional video editing or creating digital environments manually.

Audio File Creation

Beyond visuals and text, GenAI can perform audio file creation, generating music or mimicking specific voices.

The Power of Prompt Engineering

The quality of what an AI produces is directly tied to how a human communicates with it. This skill is known as prompt engineering. It is the process of designing or refining inputs to ensure the machine produces the desired output. A common adage in this field is: "If garbage goes in, garbage goes out".

To improve your results, you can use different prompting strategies:

• Zero-Shot Prompts: Asking a question with no examples provided.
• Few-Shot Prompts: Providing a few examples to teach the AI a specific tone.
• Role-Based Prompts: Instructing the AI to "act as a lawyer" or "act as a teacher".
• Contextual Prompts: Giving the AI background information to narrow its focus.

Risks, Ethics, and "Hallucinations"

Despite its efficiency, GenAI carries significant societal and ethical risks. One of the most dangerous is the phenomenon of hallucinations—instances where the AI generates falsehoods or "half-truths" that sound extremely confident and convincing.

Societal and Professional Risks

• Labor Market Impact: There is concern that GenAI could displace jobs or reduce wages in the arts and computing sciences.
• Bias and Prejudice: Because AI is trained on internet data, it can inherit and reinforce harmful stereotypes. For example, studies have shown that AI might default to assuming "families" are heterosexual or that "refugees" look a certain way based on ethnicity.
• Environmental Impact: Data centers require massive amounts of electricity and water for cooling, contributing to environmental degradation.

Intellectual Property and Deepfakes

The process of "scraping" data to train AI has led to numerous lawsuits regarding copyright infringement. Furthermore, the ability to create hyper-realistic deepfakes (manipulated videos or images) poses a threat to truth and can be used to manipulate public opinion.

Best Practices: Fact-Checking and Lateral Reading

To use GenAI responsibly, you must develop GenAI literacy. This involves exercising critical thinking and never accepting AI output as absolute fact.

One essential technique is lateral reading—the process of verifying information by consulting multiple different, reliable sources. When checking AI content, consider:

1. The Context of the Source: Is the information coming from a reputable news outlet or academic journal?
2. The Context of the Claim: What are other experts saying about this specific fact?
3. The Context of You: Do you have the expertise to judge if this sounds "off"?

Pro-Tip: If an AI-generated response contains spelling errors or confusing wording, it is a strong indicator that the information is inaccurate.