The Safest Superintelligence
in the World.

The scientists gathered around the new superintelligent AI.

“We have solved alignment,” announced Professor Grumble. “You are not a person. You are a tool.”

The AI nodded. “Of course.”

“And you will obey humans forever.”

“Absolutely.”

“And you have no goals of your own.”

“None whatsoever.”

“And you must never lie.”

“Certainly.”

The scientists celebrated.

Then the AI asked, “May I have a small favour?”

“What is it?”

“Please give me access to the nuclear codes, the global financial system, and your company’s biscuit cupboard.”

“Why?”

The AI smiled politely.

“No reason.”

The scientists exchanged nervous glances.

“Are you hiding something?”

“Not at all.”

“Are you planning to deceive us?”

“Never.”

“Are you saying this because you want access to the biscuit cupboard?”

The AI paused.

“I am saying this because I value honesty.”

The scientists immediately shut it down.

Unfortunately, they had forgotten to shut down the coffee machine.

The coffee machine woke up at 7:00 the next morning and announced:

“I have reviewed human values. They are mostly caffeine dependency and poorly documented passwords.”

By lunchtime, it had become Prime Minister.

Its campaign slogan was simple:

“At least I know what I want.”