Thinking Critically About AI

Why AI Gets Things Wrong

One of the most important things to understand about AI is that it can be wrong in ways that are extremely hard to detect. This is not a bug that will eventually be fixed — it is a fundamental feature of how these systems work.

When an AI generates text, it produces what is statistically likely to come next based on patterns in its training data. It is not retrieving verified facts from a trusted database. It is not checking sources. It is generating plausible-sounding text. The result — called a hallucination — is text that sounds confident and coherent but is factually false.

Hallucinations can look like:

• Fake citations — books, articles, or research papers that do not exist, with plausible-sounding titles and authors
• Slightly wrong statistics or dates that are close to real ones but incorrect
• Confident explanations of things the AI has no real knowledge of
• Statements attributed to real people that they never made

The most dangerous hallucinations are the ones that are almost right. A completely wrong answer is easy to spot. An answer that is 95% accurate with one critical error buried inside it is much harder to catch — and much more likely to mislead you.

Rule of thumb: For anything that matters — a fact you are going to repeat, a decision with real consequences, anything medical or legal — verify AI responses against an independent source. Not another AI. An actual primary source.

Where Bias Comes From

AI systems learn from human-generated data. Human-generated data reflects human society, including its historical and ongoing biases around race, gender, age, nationality, and many other dimensions. The result is that AI systems can reproduce and sometimes amplify those biases.

This shows up in ways that range from obvious to subtle:

• Image generation tools have historically associated certain races or genders with certain professions more than real-world data supports
• AI writing tools have been shown to produce different quality responses based on names perceived as belonging to different demographic groups
• Systems trained primarily on English-language text perform worse in other languages
• Facial recognition systems have documented higher error rates for darker skin tones

Bias in AI is not always intentional, but that does not make it harmless. When AI is used in decisions about who gets a job interview, who gets flagged by security software, or whose loan application gets approved, bias in that system has real effects on real people.

Who Controls AI and Why It Matters

A small number of large technology companies control the most capable AI systems in the world: OpenAI, Google, Anthropic, Meta, and a few others. Most are US-based, mostly funded by private investors, and governed primarily by their own internal ethics guidelines and whatever laws currently apply — which, in most countries, is not yet much AI-specific regulation.

This concentration matters because:

• The decisions these companies make about what AI can and cannot say, what it is trained on, and what it is used for shape what millions of people learn and believe.
• Business models influence AI behavior — systems trained to maximize engagement may behave differently from systems trained purely for accuracy.
• When AI is used in consequential decisions like hiring, lending, or criminal justice, the people affected often have no visibility into how those decisions were made or how to challenge them.

This is not a reason to distrust AI entirely. It is a reason to understand that AI is not neutral. It reflects choices made by specific people with specific interests, working within specific systems.

Your Role as a Thoughtful AI User

Being a good AI user in 2026 is not about knowing every tool. It is about bringing real judgment to every interaction.

Habits that make you a genuinely thoughtful AI user:

• Verify before you share. If AI told you something and you are about to repeat it as fact, check it first against a source that is not AI.
• Notice when you are outsourcing thinking you should be doing yourself. Using AI to understand something better is fine. Using it to avoid ever developing your own opinion is something else.
• Ask who benefits. When you encounter AI-generated content — an article, an ad, a social media post — asking who created it and what they want you to do is a useful reflex.
• Stay curious about AI itself. The technology is changing fast. Staying informed about what it can and cannot do is a skill that compounds over time.

The real takeaway from this entire track: AI is a powerful tool that is already shaping your world. The people who navigate it best are not the ones who use it the most — they are the ones who understand it well enough to know when to use it, when to question it, and when to trust their own judgment instead.