Knowledgebase

When AI Is Confidently Wrong Print

  • aiethicsresponsibleuse, nigeria, guide, howto, solution, zillionkinghost, hosting, support
  • 0

Recognising it.

THE DIFFICULTY

There is no signal. Wrong output reads exactly like right output.

WHAT INCREASES THE RISK

Questions about recent events Obscure or specialist subjects Specific details rather than general principles Anything the model has little training data about Local and regional specifics

THAT LAST ONE

For Nigerian law, regulations, local conditions and institutions, coverage is thinner than for other contexts.

Errors are correspondingly more likely.

WHAT TO WATCH FOR

Unusual specificity about something obscure Confident answers about very recent matters Details that would be hard for anyone to know Anything that conveniently supports what you hoped

WHAT TO DO

Ask for the source, then check it Ask what the claim rests on Verify anything surprising

WHAT NOT TO DO

Assume a detailed answer is a researched one.

Detail is generated as readily as generality.

THE HABIT

Ask yourself what happens if this particular statement is wrong.


Was this answer helpful?
Back

Are you happy with your experience? Leave us a review on Trustpilot.


Trustpilot