The first thing I learned in Computer Science 1 at UC Berkeley in the fall of 1977 was GIGO. That stands for “Garbage In, Garbage Out.”

If you write bad code, you will get bad output.

You don’t want the GO or “Garbage Out” part to happen. So, you make sure the GI or “Garbage In” part doesn’t happen!

Today’s AI (mostly Large Language Models) are trained on data. Data is just a fancy word for “all the junk in the world.” Since computers don’t “know” anything, that is, they have no morals, ethics, or values, they can’t “decide” what is good information or what is bad information.

AI engineers started with the assumption of controlled environments and trustworthy inputs. But those things exist only in labs. In the real world there is plenty of garbage. And when AIs can slurp from the entire internet, they can embed corrupt material. They can incorporate suspicious code and ingest poisoned documents. Even if the programming works the way it is supposed to, the outputs can be foolish and stupid because you can’t trust the inputs.

This is the part the AI industry doesn’t want to talk about. You see, the industry has prioritized efficiency over integrity. Doing things right takes time and thus costs money.

From Bruce Schneier, one of my go-to sources on all things tech (along with Molly White):

Integrity isn’t a feature you add; it’s an architecture you choose. So far, we have built AI systems where “fast” and “smart” preclude “secure.” We optimized for capability over verification, for accessing web-scale data over ensuring trust. AI agents will be even more powerful—and increasingly autonomous. And without integrity, they will also be dangerous.

What kind of architectures do you think the Titans of AI will choose for their systems? Which ones have they already chosen?