Unit 6: Modern AI Trends and Tools - Subjective Questions

Artificial Intelligence (AI) is a branch of computing concerned with developing machines and software that can perform tasks normally requiring human intelligence.

The main capabilities of an intelligent system include:

• Learning: It can learn patterns from data and improve its performance over time.
• Reasoning: It can analyze available information and draw logical conclusions.
• Problem-solving: It can identify possible solutions and select an appropriate one.
• Perception: It can interpret images, speech, video, sensor readings, and other inputs.
• Natural language processing: It can understand and generate human language.
• Decision-making: It can choose actions based on goals, rules, predictions, or evidence.

For example, a voice assistant uses speech recognition to understand a command, natural language processing to interpret it, and AI-based decision-making to provide a response.

The three concepts differ mainly in the range and level of intelligence they describe:

• Narrow AI:

  • Designed to perform one task or a limited set of tasks.
  • It does not possess general human-like understanding.
  • Examples include recommendation systems, spam filters, face recognition, and present-day chatbots.

• General AI:

  • Refers to a hypothetical system capable of learning and performing a wide variety of intellectual tasks at a human level.
  • It would be able to transfer knowledge from one domain to another.
  • No universally accepted example of true General AI currently exists.

• Artificial Superintelligence:

  • A hypothetical form of AI that would exceed human intelligence in most or all fields.
  • It could potentially outperform humans in reasoning, creativity, planning, and scientific discovery.

Thus, most AI systems used today are examples of Narrow AI, while General AI and Artificial Superintelligence remain theoretical concepts.

AI is applied in many areas of everyday life:

1. Healthcare: AI assists in medical-image analysis, disease-risk prediction, drug discovery, and patient monitoring. It can help professionals detect problems earlier.
2. Education: Adaptive learning platforms personalize lessons, while AI tutors provide explanations and feedback based on a learner's needs.
3. Transportation: Navigation applications predict traffic and recommend routes. Driver-assistance systems can detect lanes, vehicles, and obstacles.
4. Banking and finance: AI is used for fraud detection, credit-risk analysis, customer support, and unusual-transaction monitoring.
5. Entertainment and e-commerce: Recommendation systems suggest films, music, videos, or products by analyzing user preferences.

Other applications include agriculture, language translation, cybersecurity, manufacturing, and smart-home automation. These systems can improve speed, accuracy, personalization, and efficiency, but important decisions should still involve appropriate human oversight.

AI can support both healthcare and education, but it should generally assist rather than completely replace qualified professionals.

AI in healthcare

• Analyzes X-rays, scans, and other medical images.
• Helps predict health risks and identify patterns in patient data.
• Supports appointment scheduling, documentation, and remote monitoring.
• Assists researchers in studying diseases and potential medicines.

AI in education

• Creates personalized exercises and learning paths.
• Provides immediate feedback and explanations.
• Assists teachers in preparing quizzes, summaries, and lesson resources.
• Improves accessibility through translation, transcription, and text-to-speech tools.

Benefits

• Faster processing of large amounts of information.
• More personalized services.
• Improved access and administrative efficiency.
• Support for professionals in routine tasks.

Limitations

• AI can generate incorrect or biased results.
• Sensitive health and student data may be exposed or misused.
• Excessive reliance can weaken human judgment and interaction.
• Not every user has equal access to AI technology.

Therefore, outputs should be verified, data should be protected, and final high-impact decisions should remain under human supervision.

Generative AI is a category of AI that produces new content, such as text, images, audio, computer code, video, or animation, after learning patterns from data.

A simplified generation process is:

1. Training data is collected: The model is exposed to a large quantity of relevant data, subject to the system's design and data permissions.
2. Patterns are learned: During training, the model learns statistical relationships among words, pixels, sounds, or other data elements.
3. A prompt is provided: The user gives an instruction, question, image, or other input.
4. Content is predicted: The model predicts suitable elements step by step, such as the next word in a sentence or visual features in an image.
5. An output is generated: The predictions are combined into a response that appears new.
6. The user reviews the output: The result may be refined through further prompts and should be checked for accuracy, bias, copyright concerns, and suitability.

Generative AI does not necessarily understand a topic in the same way a human does. It may produce convincing but incorrect information, commonly called a hallucination.

Generative AI can be classified according to its output modality:

• Text generation: Produces explanations, summaries, stories, emails, reports, and dialogue. ChatGPT and Gemini are examples of tools with text-generation capabilities.
• Image generation: Creates or edits images from text or image prompts. It can be used for posters, illustrations, concept art, and product mock-ups.
• Audio and music generation: Produces speech, sound effects, music, or voiceovers. It is useful in accessibility, entertainment, and media production.
• Video generation: Creates or edits video clips from text, images, or existing footage. It can support advertising, demonstrations, and storytelling.
• Animation generation: Produces motion graphics, animated characters, transitions, or movements from instructions.
• Code generation: Suggests code, tests, comments, and debugging approaches for software development.
• Multimodal generation: Accepts or produces more than one data type, such as text, images, audio, and video.

These categories may overlap because modern multimodal tools can perform several types of generation within one system.

Discriminative AI and generative AI differ mainly in their goals:

A discriminative medical model might classify a scan as high-risk or low-risk. A generative model might draft a plain-language explanation of the scan report. Some modern systems combine both approaches, so the categories are not always completely separate in practice.

ChatGPT is a conversational generative AI tool that receives prompts and generates natural-language responses.

Main features

• Maintains conversational context within a chat.
• Generates, summarizes, rewrites, and translates text.
• Explains concepts and answers follow-up questions.
• Assists with brainstorming, coding, planning, and drafting.
• May support multimodal inputs or tools depending on the available version.

Common uses

• Drafting emails, reports, stories, and presentations.
• Creating study notes, quizzes, and examples.
• Simplifying difficult passages.
• Suggesting code and debugging steps.
• Simulating interviews or practice conversations.

Limitations

• It can produce inaccurate or invented information.
• Its responses may contain bias inherited from data or instructions.
• It may misunderstand ambiguous prompts.
• It should not be trusted automatically for medical, legal, financial, or safety-critical decisions.
• Entering personal, confidential, or copyrighted material can create privacy or legal concerns.

Users should provide clear prompts, verify important claims with reliable sources, and disclose AI assistance when required.

Gemini is Google's family of generative AI models and related assistant experiences. It can support tasks such as answering questions, drafting text, summarizing information, analyzing supplied content, and working with multiple data types depending on the version and service.

Similarities with ChatGPT

• Both can generate and rewrite text.
• Both can answer questions and explain concepts.
• Both can assist with brainstorming, coding, translation, and summarization.
• Both may provide multimodal capabilities, such as working with images or files.
• Both can make mistakes and require verification.

General differences

• They are developed by different organizations and use different model families.
• Their integrations, interfaces, subscription plans, available tools, and supported context may differ.
• Gemini may integrate with supported Google services, while ChatGPT may integrate with tools and services available in the OpenAI ecosystem.
• Output quality can vary by task, model version, language, and prompt.

Neither tool is universally better for every task. A user should compare accuracy, source support, privacy settings, cost, accessibility, integrations, and output quality before selecting one.

An effective and responsible workflow includes the following stages:

1. Understand the assignment: Identify the topic, learning outcomes, word limit, format, and institutional AI policy.
2. Plan independently: Write initial ideas and prepare a basic outline before asking the tool for assistance.
3. Use AI for support: Request brainstorming, alternative structures, explanations, or feedback rather than submitting an unexamined AI-generated answer.
4. Consult reliable sources: Use books, academic databases, official reports, and credible websites for evidence.
5. Verify claims and citations: Check every important fact, quotation, statistic, and reference. AI may invent citations.
6. Write and revise: Express the ideas in one's own words, add original analysis, and ensure the report reflects personal understanding.
7. Check bias and originality: Remove unsupported claims, copied wording, and misleading content.
8. Acknowledge AI use: Disclose the tool and the nature of its assistance when required by the teacher or institution.
9. Protect data: Do not upload confidential research, personal information, or restricted documents without permission.

AI should support learning and editing, not replace the student's responsibility for the final work.

AI image-generation tools create or transform visual content based on text prompts, reference images, or editing instructions. Examples may include DALL-E, Adobe Firefly, Midjourney, and other image-generation systems.

Basic process

• The user provides a prompt describing the subject, style, composition, lighting, color, and format.
• The model interprets relationships between language and visual patterns learned during training.
• It constructs an image through a model-specific generation process.
• The user may refine the result by changing the prompt, generating variations, or editing selected regions.

Applications

• Posters, advertisements, and social-media graphics.
• Book illustrations and concept art.
• Product mock-ups and design prototypes.
• Educational diagrams and storytelling.
• Background removal, restoration, and image editing.

Limitations and risks

• Images may contain distorted objects, incorrect text, or unrealistic details.
• Training data and outputs can involve copyright or licensing concerns.
• Generated images may reproduce social stereotypes or bias.
• Deepfakes can be used for deception or harassment.
• Ownership and permitted commercial use depend on the tool's terms and applicable law.

Users should inspect outputs carefully, respect licenses and consent, and label synthetic images when they could mislead viewers.

AI video and animation tools can generate or modify moving visual content from text, images, audio, or existing footage. Depending on the platform, they may support text-to-video, image-to-video, automatic subtitles, avatars, lip synchronization, background replacement, motion generation, and scene editing.

Typical production process

1. Plan the objective: Define the audience, message, length, platform, and visual style.
2. Prepare a script or storyboard: Divide the content into scenes and specify narration and actions.
3. Enter prompts or assets: Provide text instructions, reference images, audio, or video clips.
4. Generate scenes: Create short clips, avatar presentations, or animated sequences.
5. Review and edit: Correct visual errors, timing, transitions, subtitles, and audio.
6. Export and disclose: Publish in a suitable format and label AI-generated or altered content when necessary.

Risks

• Creation of deceptive deepfakes.
• Use of a person's face or voice without consent.
• Copyright and licensing violations.
• Biased or inappropriate visual representation.
• Inconsistent characters, unnatural motion, and factual errors.

Human review, consent, source checking, and transparent labeling are therefore essential.

Perplexity AI is an AI-assisted answer and research tool that searches for information, summarizes findings, and commonly presents links or citations alongside its response.

Research uses

• Obtaining an initial overview of a topic.
• Discovering web pages, reports, and possible sources.
• Asking follow-up questions to narrow a topic.
• Comparing viewpoints or creating a preliminary research outline.
• Finding keywords for deeper academic searching.

Suggested workflow

1. Ask a focused research question.
2. Examine the linked sources rather than relying only on the summary.
3. Check the author, date, publisher, evidence, and relevance of each source.
4. Compare the information with books, academic databases, and official publications.
5. Record and cite the original sources according to the required citation style.

Verification is necessary because the AI may misinterpret a source, omit context, use a weak webpage, or attach a citation that does not fully support the claim. The presence of citations does not automatically guarantee accuracy or academic reliability.

NotebookLM is an AI-assisted research and note-taking tool designed to work primarily with sources supplied or selected by the user. Its exact features may change, but its central purpose is to help users understand and organize a collection of documents.

Major uses and features

• Summarizes uploaded or connected source material.
• Answers questions based on the selected sources.
• Identifies themes, important points, and relationships across documents.
• Helps create study guides, outlines, frequently asked questions, and other learning aids.
• Provides source references or links within the notebook so users can inspect supporting material.

Difference from a general chatbot

• A source-grounded tool is instructed to base answers mainly on a defined collection of documents.
• A general chatbot may respond from broader learned patterns and any tools or sources available to it.
• Grounding can make the evidence easier to trace and keep answers focused on the chosen material.

However, source grounding does not guarantee correctness. The tool may overlook information, misread a passage, or produce a poor summary. Users must check the original documents and avoid uploading confidential or copyrighted material without permission.

Jenni AI is an AI-assisted writing tool that can support drafting, continuation, rewriting, organization, and other parts of the writing process. Available features may differ by plan or version.

Possible uses

• Brainstorming titles, topics, and research questions.
• Developing an outline for an essay or report.
• Suggesting sentence continuations or alternative wording.
• Improving clarity, grammar, and paragraph flow.
• Summarizing notes or helping organize arguments.
• Assisting with citation-related workflows where supported.

Precautions

• Verify every citation: A suggested reference may be incorrect, incomplete, or unrelated.
• Check factual claims: AI-generated statements should be compared with original, credible sources.
• Avoid plagiarism: Students must not present copied or unexamined AI output as independent work.
• Follow institutional policy: Some assignments may restrict or prohibit AI assistance.
• Maintain authorship: The final argument, interpretation, and conclusions should reflect the student's understanding.
• Protect confidential data: Unpublished research, private records, and personal information should not be entered without authorization.
• Disclose use when required: The student should explain how the tool contributed to the work.

Jenni AI can assist writing, but responsibility for accuracy, originality, and academic honesty remains with the user.

The three tools support different stages of research and writing:

Recommended combined workflow

1. Use Perplexity AI to explore a topic and identify useful keywords and sources.
2. Obtain and evaluate reliable original documents.
3. Add permitted sources to NotebookLM to compare themes and prepare notes.
4. Use Jenni AI selectively to improve structure or expression.
5. Verify every fact and citation using the original material.
6. Add independent analysis and disclose AI use where required.

The choice should depend on the task, privacy requirements, institutional policy, and the reliability of available sources.

Prompt engineering is the process of designing, testing, and refining instructions given to an AI system so that it produces a more relevant, accurate, safe, and appropriately formatted response.

An effective prompt may include:

• Role: State the perspective the AI should adopt, such as a tutor, editor, or technical explainer.
• Task: Clearly describe what the system must do.
• Context: Supply relevant background, audience information, or source material.
• Scope: Specify what should and should not be covered.
• Constraints: Include length, tone, difficulty level, deadline, or other limits.
• Output format: Request a table, bullet list, report, JSON structure, or step-by-step explanation.
• Examples: Show a sample input or expected style when consistency is important.
• Quality criteria: Ask for clear reasoning, balanced viewpoints, source identification, or uncertainty statements.

After receiving the output, the user should evaluate it and refine the prompt. Prompt engineering is usually an iterative process, not a single command.

A poor prompt is vague, lacks context, and does not specify the expected output. A good prompt clearly communicates the task, audience, scope, constraints, and format.

Poor prompt

Explain AI.

Problems with this prompt include:

• The scope is extremely broad.
• The intended audience is unknown.
• No length or format is specified.
• It does not state which aspect of AI is important.

Improved prompt

Act as an introductory computing tutor. Explain three real-life applications of AI to first-year students in 250 words. For each application, give one benefit and one limitation. Use headings and bullet points, avoid technical jargon, and end with a two-sentence conclusion.

This is a better prompt because it includes:

• Role: Introductory computing tutor.
• Task: Explain three AI applications.
• Audience: First-year students.
• Length: 250 words.
• Content requirements: One benefit and one limitation per application.
• Style and format: Simple language, headings, bullet points, and a conclusion.

A good prompt improves relevance and consistency, although it cannot guarantee that every generated claim will be correct.

A high-quality revised prompt could be:

Act as a computing instructor preparing material for first-year students. Create an eight-slide classroom presentation on Generative AI. Include: a definition, a simple explanation of how it works, major types, two text-generation tools, image-generation tools, video and animation tools, three real-life applications, limitations, and ethical-use guidelines. For each slide, provide a title and four concise bullet points. Add one practical example and two discussion questions. Use clear language, avoid unsupported claims, and identify any information that should be independently verified.

Why the revised prompt is better

• It assigns a clear role to the AI.
• It identifies the audience as first-year students.
• It specifies the purpose as a classroom presentation.
• It defines the length as eight slides.
• It lists the required content and scope.
• It controls the format by requesting slide titles and concise bullet points.
• It adds quality requirements, including practical examples and verification warnings.
• It reduces ambiguity and makes the output easier to evaluate.

The generated presentation should still be reviewed for accuracy, suitability, copyright, and alignment with the course notes.

The ethical use of AI requires attention to both the output and the way the system is used.

Major ethical issues

• Bias and discrimination: AI may reproduce unfair patterns present in data or system design.
• Privacy: Prompts may expose personal, confidential, medical, educational, or organizational information.
• Misinformation: Generative AI can produce false claims, fabricated citations, and convincing synthetic media.
• Copyright and ownership: Training material, user inputs, and generated outputs may raise legal and licensing questions.
• Academic integrity: Submitting AI-generated work without permission or acknowledgment can amount to misconduct.
• Lack of transparency: People may not know that they are interacting with AI or viewing synthetic content.
• Accountability: It may be unclear who is responsible when an AI-supported decision causes harm.
• Security and misuse: AI can be used for phishing, impersonation, deepfakes, malware assistance, or manipulation.
• Overdependence: Excessive reliance may weaken critical thinking and professional judgment.
• Unequal access: Differences in access to advanced AI tools can increase social or educational inequality.

Guidelines for responsible use

1. Verify important information through credible and independent sources.
2. Keep humans responsible for high-impact decisions.
3. Do not enter sensitive data without authorization and suitable protection.
4. Test outputs for bias and include diverse perspectives.
5. Respect copyright, licenses, consent, and attribution requirements.
6. Label AI-generated or altered content when it could mislead an audience.
7. Follow institutional, workplace, and legal policies.
8. Disclose AI assistance in academic or professional work when required.
9. Provide ways to question, correct, or appeal important AI-supported decisions.
10. Use AI to support human creativity and judgment rather than to deceive or cause harm.

Responsible AI use combines accuracy, fairness, privacy, transparency, accountability, safety, and human oversight.