01/10/2026
π‘ Tech Tip Thursday: Don't blindly trust AI
Most of us are using AI in one form or another. Recently, it's been put forward that the long standing Navier-Stokes problem was solved by OpenAI. Pretty impressive.
So we gave one of our AI tools a simple test, a three digit "crack the code" puzzle. Be a doddle, right?
The answer was 463. The AI confidently told Simon Lister he was wrong. So he explained the logic. The AI told him he was still wrong. He explained again, still wrong. Eventually, after a few rounds, the AI finally admitted:
*βYou cracked it.β* π
What does this tell us?
AI can:
β
Save you time
β
Analyse information
β
Draft content
β
Help solve problems
β
Automate repetitive tasks
But it can also:
β Make mistakes
β Sound extremely confident while being wrong
β Misinterpret information
β Need a human to challenge its answer
The best way to use AI isn't to blindly trust it β it's to work with it.
Ask questions. Check the answers. Challenge it when something doesn't look right.
And occasionallyβ¦ prove the AI wrong. π