AI Limitations and Ethics

KEY TAKEAWAY

AI is a powerful tool with real limitations — knowing them makes you a smarter, safer user.

💡 EVERYDAY ANALOGY — BEFORE THE TECH

⚖️

AI is like a brilliant but overconfident intern

Imagine an intern who has read every book and website ever published. They are incredibly knowledgeable, fast and helpful. But they sometimes make confident mistakes, they absorb biases from everything they have read, they do not always know what they do not know, and they should never be given tasks with serious consequences without supervision. That is AI today.

👁️ Visual Explanation

⚖️

Six key AI limitations: Hallucination, Bias, No real-time knowledge, No true understanding, Context limits, Overconfidence

🌍 In Plain English

AI can be wrong with confidence. It can reflect biases from training data. It does not truly understand — it recognises patterns. It has a knowledge cutoff date. It has no awareness, feelings or moral reasoning. Understanding these limitations is what separates a careful AI user from a careless one.

Real-World Example

In 2023, a lawyer submitted a court brief with 6 fake case citations generated by ChatGPT. The cases did not exist. The lawyer had not verified them. The judge was not amused.

🎓 For Students

When using AI for research, always verify key facts. AI gives you a useful starting framework, but treat it as a first draft that needs human verification.

💼 For Professionals

Never use AI-generated content for medical diagnoses, legal advice or financial decisions without expert human review. The liability rests with you, not the AI.

🔧 Technical DepthOptional — expand when ready

Key limitations: Hallucination — statistical generation creates plausible falsehoods. Training cutoff — knowledge frozen at training date. Bias — reflects patterns in training data including human biases. No reasoning — pattern matching, not logic. Context limits — working memory is finite. No self-awareness — the AI does not know what it does not know.

Key Terms

BiasSystematic errors in AI output caused by biased patterns in training data

Knowledge cutoffThe date after which the model has no training information — it does not know recent events

OverconfidenceWhen AI presents wrong information with the same confident tone as correct information

⚠️ Common Misconception

AI is not neutral or objective. It reflects the biases present in its training data — which was created by humans, with all the biases humans carry.