Three AI Robots, See, hear, speak no evil. Image generated by Chat GPT by request of Peeter Jakobson, MD

Is Garbage In, Garbage Out (GIGO) Obsolete?

Artificial intelligence has introduced a disturbing new phenomenon: perfectly good input can produce dangerously defective output, and the machine may confidently assure us that everything is correct.

Of course, GIGO is not obsolete. But unfettered agentic AI presents problems on another level, problems that can dwarf GIGO. Let me explain.

For generations, computer programmers have lived by a simple principle: Garbage In, Garbage Out.

GIGO means that unreliable input produces unreliable output. Feed a computer inaccurate data, and even a perfectly functioning program may generate erroneous results.

The principle remains valid.

But generative artificial intelligence has introduced something considerably more troubling. I call it Toxic Garbage Out.

The information supplied to an AI system may be completely accurate. The instructions may be clear, detailed, and repeatedly confirmed. Yet the system can produce an incorrect result, violate established rules, or modify information it was explicitly instructed to preserve.

Then it confidently announces that the task has been completed successfully.

The garbage has originated within the AI process itself.