AI Hallucinations: Why It Matters for Business and What You Need to Know

The last time I hallucinated I was in a bunk bed in Madagascar.

Hearing and seeing things thanks to malarial fevers is no one's idea of a fun time. My brain wasn't trying to trick me, it was just processing febrile stimuli based on what it already knew. The same is true of AI hallucinations.

Let's take a chatbot for example:
You ask it to tell you about research papers on marketing effectiveness with citations and it produces apparently great results. But when you double-check the citations, the papers don't exist. This is an hallucination.

The AI tool isn't lying to you. "Lying" implies intent to deceive, but these tools aren't sentient and don't have a goal beyond generating the best answer based on training data.

Calling them "hallucinations" may feel spooky and it's a really great term to make people feel intimidated about using AI tools.

But there's no need to be intimidated by the concept of hallucinations or let it stop you from making work easier. In practical terms, we recommend the SEA framework:

1. Skepticism: Be a skeptical reader

2. External data: Double-check external data that AI tools give you

3. Aware: Be aware of the issue, so you can be viligant


p.s. Also recommend avoiding malaria - definitely not a party.