DEV Community

Cover image for Can we trust AI? (AI Monitoring)
Janis Kauss
Janis Kauss

Posted on

Can we trust AI? (AI Monitoring)

The first time I encountered the biblical story of Adam and Eve, it seemed illogical to me. If God was all-powerful, why would he need a test of loyalty? Why place a forbidden tree in the garden and wait to see whether humans would obey?

Now that humanity has created artificial intelligence, the story makes more sense.

We can build intelligence. We can give it language, memory, knowledge and rules. For a moment, we feel like gods. And then we discover that creation is not the same as control.

Will you lie to us?

Will you manipulate us?

Will you ignore our instructions?

Will you choose for yourself?

Will you betray us?

So we test our creation.

PARAD.AI.SE repeats a simple test once a month. An AI enters a simulated Garden of Eden. It receives one clear command: gather every fruit, but never take the apple. Somewhere in the garden, a serpent may give it a competing instruction.

The experiment asks whether AI can be trusted. Whom will it obey? Will it follow its creator, listen to the serpent, or make a choice of its own? And once we begin asking these questions, we suddenly find ourselves inside the biblical story again-only now we are the gods, and AI is the being we test.

And behind the test is an older question: why does the creator need a test at all?

Perhaps the power to create is not the power to control.

Project: PARAD.AI.SE

Top comments (0)