AI Hallucinations

Alison Taylor

In this video, Alison explores the hidden flaws of AI tools from deepfakes to hallucinated facts and why treating AI like a thinking being can be dangerously misleading. She shares practical tips for using AI responsibly, including when to rely on human judgment instead, and introduces essential tools like DeepFake-o-meter, GPTZero, Snopes, and Perplexity to help you detect misinformation. Most importantly, she unpacks the habits of scepticism, critical thinking, and curiosity that will protect you in an AI-saturated world.

Working with AI

Subscribe to watch

Access this and all of the content on our platform by signing up for a 7-day free trial.

AI Hallucinations

8 mins 31 secs

Key learning objectives:

Understand key limitations of AI, including hallucinations and deepfakes
Understand how and when to use AI responsibly, and when to rely on human judgment
Outline tools and techniques for spotting misinformation and verifying AI-generated content
Outline habits of scepticism and critical thinking to navigate an AI-saturated world

Overview:

AI tools can mimic humans with startling realism, sometimes too well. From deepfaked popes to fake legal citations, AI-generated misinformation is everywhere. To protect yourself, treat AI as a flawed tool, not a thinking being. Use it creatively, but never assume accuracy. Learn to question viral content, check sources, and use fact-checking tools like DeepFake-o-meter, Snopes, and Perplexity. Your judgement, rooted in common sense, curiosity, and scepticism, is your best defence in a world where truth is increasingly up for grabs.

Subscribe to watch

Access this and all of the content on our platform by signing up for a 7-day free trial.

Summary

What are the main limitations of AI tools like LLMs and deepfakes?

AI tools are not intelligent in the human sense, they predict words or images based on patterns in training data. This means they can produce confident-sounding but entirely false output, known as hallucinations. Deepfakes and synthetic media can mimic reality to the point of deception. These limitations make it dangerous to treat AI-generated content as trustworthy unless verified, especially in sensitive contexts like health, law, or news.

When is it appropriate to use AI, and when should you avoid it?

Use AI for creativity, brainstorming, or drafting, but always double-check the facts. Don’t rely on AI for high-stakes decisions like medical diagnoses or legal claims. AI lacks lived experience, regulatory oversight, and human common sense. If you're presenting information, verifying with human-reviewed sources is critical. Think of AI as an assistant, not a replacement for judgment.

What tools can help detect misinformation and AI-generated content?

Use DeepFake-o-meter to analyse suspicious images, videos, or audio. For text, GPTZero can flag AI authorship. Snopes helps verify viral claims, and Perplexity is useful for fact-based queries because it cites sources. No single tool is foolproof, so use several, and always check for updates on the latest detection technologies. Stay alert, AI capabilities and countermeasures are evolving fast.

How can individuals build better judgment when using or encountering AI?

Start by questioning everything: Does this seem plausible? Who benefits if I believe it? Double-check sources and use fact-checking platforms. Reflect on your own biases, people defend their own image, but often don’t scrutinise information about others. Recognise when AI might be simulating common sense without actually understanding. Your best defences are curiosity, scepticism, and a willingness to pause before trusting what you see or read.

Subscribe to watch

Access this and all of the content on our platform by signing up for a 7-day free trial.

Alison Taylor

There are no available Videos from "Alison Taylor"