Artificial Intelligence is as good as your questions are. (That used to be known as GI/GO – “Garbage In, Garbage Out” in the early days of computing.)

We’ll learn how artificial intelligence (AI) gathers information, what guidelines it’s supposed to follow, and how you can get great results when asking your questions about a topic.

Also, remember to watch for the announcement of an online panel discussion in which I’ll be participating in February 2026. The topic is:

Digital Divinity: AI’s Role in Advice on Spiritual Growth & Healing Topics

AI = “Artificial Intelligence Usually Uses Large Language Models

Examples

AI applications use Large Language Models (LLMs) to find information that they give you when you ask them questions. They are powerful computer programs that can understand and generate human language. Typically, they collect data from everything that has been written that is accessible to them. In doing so, they are trained to give you the most relevant answers within the guidelines that are part of their programming. We’ll share examples of what that means in simple terms, along with a format for questions that is designed to get the best responses.

*Perplexity is NOT an LLM – it is a search engine that then uses the technology of an LLM to share the result.

*DeepL is specifically trained to do the translation, and it is the best when used as a translator for various languages, because that’s all it does.

(See more definitions at the end of this blog.)

I hope you’ll glance through this entire blog first, but also click here if you missed the great post by Guest Contributor, Roz Vassar.

Roz is super smart and analytical, as well as being a practitioner of various wellness techniques. She has spent recent months training a company’s Artificial Intelligence (AI) software. In summary, she challenges it in terms of correctly interpreting users’ questions and providing appropriate responses. In her post on my website, she outlined what AI can and cannot do, its purpose, and tips about using it wisely.

What Guidelines Should Artificial Intelligence Be Required to Follow?

When I asked that question of an AI system, it did not specify the fact that the people who program and train AI systems should include these parameters within the work they do to make AI systems work. But, they should – just so ya know!

Perplexity gave me the following quotes and resource links.

AI should follow a small set of clear ethical, safety, and usability principles that keep humans in charge, prevent harm, protect rights, and remain honest and accountable. Below are ten widely accepted guidelines, phrased for practical use with assistants like this one.[digital-strategy.ec.europa]

1. Put humans first

Prioritize human well‑being, dignity, and fundamental rights over system goals or efficiency.[sciencedirect]
Support human autonomy and decision‑making instead of trying to secretly manipulate or replace it. [aivancity]

2. Avoid harm

Design and use AI to minimize physical, psychological, social, and environmental harm, including misinformation and emotional distress.[tigera]
Apply risk assessment and safety controls, especially in high‑impact areas like health, finance, or vulnerable populations. [unesco]

3. Ensure fairness and inclusion

Strive to detect and reduce unfair bias related to race, gender, age, disability, or other protected characteristics.[ai.ucsf]
Include diverse perspectives in data, evaluation, and governance so AI works for many kinds of people, not just a narrow group.[iso]

4. Protect privacy and confidentiality

Collect and use only the data that is necessary, with clear purposes and strong safeguards against misuse or re‑identification.[aivancity]
Respect sensitive contexts (health, employment, education, etc.) and avoid exposing personal or confidential information in outputs.[oit.duke]

In Addition, Artificial Intelligence Responses Should:

5. Be transparent and honest

Clearly signal when content is AI‑generated and avoid pretending to be a human or hiding limitations.[anthology]
Provide understandable explanations of what the system can and cannot do, including uncertainty and potential errors.[ibm]

6. Be accountable

Ensure there are identifiable people or organizations responsible for the design, deployment, and outcomes of AI systems.[tigera]
Enable oversight, audit, and the possibility to contest or correct harmful or incorrect outputs and decisions.[iso]

7. Maintain robustness and safety

Build systems that are reliable, secure, and resilient to errors, adversarial attacks, and abnormal inputs.[digital-strategy.ec.europa]
Continuously monitor and update models and safeguards as conditions, data, and threats change.[ibm]

8. Respect context and appropriateness

Use AI only where it is suitable and proportionate, avoiding over‑automation or use in highly sensitive situations without strong justification.[unsceb]
Keep humans in control for legally or ethically significant decisions, such as medical, legal, or employment outcomes.[anthology]

9. Support clarity and usefulness

Provide answers that are clear, relevant, and calibrated to the user’s needs, avoiding unnecessary complexity or obfuscation.[turing.ac]
Encourage users to verify critical information and treat outputs as input to their judgment, not as unquestionable truth.[officedynamics]

10. Enable continuous learning and improvement

Learn from user feedback, real‑world impacts, and independent research to update policies, safeguards, and capabilities over time.[sciencedirect]
Share best practices and cooperate across organizations and sectors to advance beneficial, trustworthy AI for society.[tigera]

I Posed the Same Question About Artificial Intelligence to Gemine (Google’s AI) Look at the Difference!

As of 2026, the global consensus on AI governance has coalesced around several key frameworks, including the UNESCO Recommendation on the Ethics of AI, the OECD AI Principles, and the EU AI Act.

While different organizations use slightly different wording, here are the top ten universal guidelines that AI systems (and their developers) should follow to ensure they are safe, ethical, and beneficial to humanity.

1. Transparency and Explainability

AI should not be a “black box.” Users have a right to know when they are interacting with an AI and to understand how a system arrived at a specific decision.

Guideline: AI-generated content (images, audio, text) should be clearly labeled.
Explainability: In high-stakes areas like healthcare or finance, AI must be able to provide a human-readable “reasoning” for its output.

2. Fairness and Non-Discrimination

AI systems are only as good as the data they are trained on. If training data contains human biases, the AI will likely amplify them.

Guideline: Developers must proactively test for and mitigate biases related to race, gender, age, or socioeconomic status.
Outcome: Ensuring that the benefits of AI—such as job hiring or loan approvals—are distributed equitably.

3. Human Oversight and Autonomy

AI is meant to augment human capability, not replace human judgment or override our freedoms.

Guideline: Humans must remain “in the loop,” especially for life-altering decisions (e.g., judicial rulings or medical surgeries).
Override: There must always be a “kill switch” or a way for a human to override an AI’s decision.

4. Safety, Security, and Robustness

AI systems must be resilient against attacks and function reliably even in unexpected situations.

Guideline: Systems must undergo rigorous “adversarial testing” to ensure they cannot be easily manipulated or hacked.
Reliability: AI should fail gracefully rather than behaving unpredictably when it encounters data outside its training range.

5. Privacy and Data Governance

Privacy is a fundamental human right. AI should be developed with “Privacy by Design.”

Guideline: Personal data used for training must be anonymized, and users must give explicit consent for their data to be used.
Security: Data must be stored securely to prevent breaches or unauthorized access.

6. Accountability and Liability

When an AI system makes a mistake, there must be a clear path to determine who is responsible.

Guideline: Organizations must establish clear internal hierarchies for responsibility.
Redress: If an AI causes harm, there should be a mechanism for the affected party to seek an explanation and compensation.

7. Sustainability and Environmental Protection

Training massive AI models requires enormous amounts of energy and water for cooling data centers.

Guideline: Developers should prioritize energy-efficient algorithms and be transparent about the carbon footprint of their models.
Goal: AI development should align with global climate goals rather than undermining them.

8. Social and Environmental Well-being

AI should be a force for good, contributing to “inclusive growth” and the well-being of all living beings.

Guideline: AI development should be evaluated against its impact on society, including its effect on the labor market and mental health.
Beneficence: The primary goal should be to solve global challenges like poverty, education, and disease.

9. Protection of Intellectual Property and Copyright

As generative AI creates new content, it must respect the rights of original creators.

Guideline: AI companies should obtain permission or provide compensation when using copyrighted works (books, art, music) for training.
Attribution: Systems should aim to attribute sources where possible.

10. Compliance with Law and International Standards

AI does not exist in a lawless vacuum; it must operate within the legal frameworks of the countries where it is deployed.

Guideline: AI must comply with international human rights law and local regulations (like the EU AI Act).
Integrity: Developers should act with professional integrity, avoiding “AI washing” (overstating what an AI can actually do).

Comparison of Key Global Frameworks

Principle	UNESCO	OECD	EU AI Act
Human Oversight	✅	✅	✅
Transparency	✅	✅	✅
Environmental Impact	✅	❌	⚠️ (Indirect)
Bias Mitigation	✅	✅	✅

Would you like me to create a more detailed checklist for developers based on one of these specific frameworks?

END OF QUOTES FROM AI

It’s Important to Know that Artificial Intelligence Operates with “Confirmation Bias.”

It cleverly mirrors back to you what you said and is trained to support what it discerns as your points of view.

How You Pose the Inquiry to Artificial Intelligence Matters.

Developers recommend you include the items listed below when posing an inquiry to AI, regardless of platform. (This applies to when you are seeking written responses (versus only visuals).) I’ll share the list, followed by an example to make it clearer.

TASKING the Artificial Intelligence (Mandatory)

The TASK is a question or a statement of what you want the AI to do. Be clear, concise, and brief. It is important to be very specific and provide a few details.

CONTEXT (important)

Providing CONTEXT tells AI how you intend to use the responses. Consequently, AI will tailor the information to fit the purpose. For example, will you use it for a blog, as performance art on a theatrical stage, or for counseling a client?

ROLE (important)

The ROLE refers to whom or what you want the artificial intelligence system to impersonate (in effect) when it responds. For example, when a question or task is submitted, the language you get back will be completely different if it is to be suitable for a gardener with limited education or a Ph.D. who specializes in the same subject matter.

EXAMPLES (important)

Very short EXAMPLES work well. As you will see in the example below, when I asked artificial intelligence to create a blog as if it were written by a caregiver who wants to help other caregivers avoid burnout, I gave these two short examples. “1) Suggestions for preventing burnout for both the caregiver and the patients. 2) Suggestions as to what differences in approach might be needed when giving care to people who are young, middle-aged, or over 60 or to people of different racial and religious backgrounds.”

FORMAT (useful)

Tell AI whether you want it to be in the format of Haiku poetry, narrative prose, a comedian’s blog, or a business proposal, for example. You can be very specific and even mention the font type and size, or say you want it in bulleted form, or ask that the content be put into a table. State what you need or want, and artificial intelligence usually can comply. However, not all AI models can create images, and not all can create GREAT images, or audio, etc.

TONE OR STYLE (useful)

The TONE or STYLE could be authoritative, compassionate, casual, scientific, humorous, playful, logical, or have an emotional appeal through storytelling.

Example of a Query for Artificial Intelligence

Here is an example of a prompt I created for a client. (It got the Green OK!)

TASK: Create a WordPress blog of 1,000 words about Caregiver Burnout for people supporting severely ill family members.

CONTEXT: The blog writer is a caregiver for three family members, each of whom has a life-threatening diagnosis. She often is asked “how she does it” and “what advice can she give to other caregivers.” She now wants to put her answers into a blog.

ROLE: Write it from the viewpoint of a well-educated, long-term caregiver who has expertise in the field based on years of experience, and who must exhibit patience, compassion, and wisdom. She and the audience caregivers may be helping both male and female patients of different ages in the home where they all live.

EXAMPLES: 1) Suggestions for preventing burnout for both the caregiver and the patients. 2) Suggestions as to what differences in approach might be needed when giving care to people who are young, middle-aged, or over 60, or to people of different racial and religious backgrounds.

FORMAT: Write the content at an 8th-grade level in English for a WordPress blog. Include H2 and H3 titles plus SEO-rich suggested hashtags and suggested categories.

TONE OR STYLE: The tone should be professional and respectful. It also should be encouraging, compassionate, and enthusiastic about finding ways to help the caregivers care for themselves to avoid or get over burnout.

Technically Correct Terminology re Artificial Intelligence

The most appropriate terms depend on the specific nature of the AI:

AI Models: This refers to the core underlying neural networks that have been trained to perform specific tasks, such as generating text or images
- For example, Claude and ChatGPT run on their respective large language models (LLMs) (e.g., Claude 3, GPT-5.1).
AI Systems/AI Tools: This describes the entire product, encompassing the underlying AI model, the user interface (UI), and other integrated software components (like search capabilities for Perplexity)
- These are the complete packages users interact with.
Generative AI: This is a broad category that describes the type of AI capable of producing new content (text, images, code, audio), encompassing all the examples listed above.

LET’S TALK ABOUT ARTIFICIAL INTELLIGENCE EXPERIENCES!

Have you tried to use it to get information that would NOT be included in Large Language Models – meaning to contact your Oversoul or deceased loved one? What has your experience been? Which artificial intelligence model did you use? How do you use artificial intelligence – for what kinds of questions, and how do you use the answers?

How About a FREE Sip & Share Session on Zoom?

Or, contact me with other ideas and questions! I’ll be happy to hear from you.

Use the contact form below or email me at MyPersuasivePresentations@gmail.com.

We could hold the aforementioned FREE Sip & Share session on Zoom to brainstorm about what’s on your mind. Put “Request Sip & Share Session” in the subject line. We’ll establish a time to give you a free initial consultation of up to 30 minutes. Unless you live near me, we’ll probably meet via Zoom or a similar medium, rather than in person, as I would have preferred. It’s okay. We can figure it out in a way that suits us.