Cage AI, Or Hand It The Keys?

Men & Women have spent a working life in the mining business.

Nobody ever argued about whether mining was dangerous. Everybody knew it was. Nobody argued about whether it paid, either. It did, and that’s why we were there.

What kept them alive wasn’t a ban on mining, and it wasn’t blind faith in the mine. It was how the work got done, every shift, by the people doing it.

I’ve been thinking about that, because the AI world has split into two camps, and neither one is talking about how the work gets done.

The first camp wants a cage.

On Wednesday, Senator Bernie Sanders and Representative Greg Casar unveiled a bill that would permanently ban what they call artificial superintelligence. The Associated Press had the story first.

The bill would pause work on the most advanced AI systems. It would create a new federal Department of Artificial Intelligence to write safety rules. Advanced systems would need government approval before they could be put to use, and some violations could bring up to twenty years in prison.

Senator Sanders put his reasoning in a few words. He told the AP, “It doesn’t take a genius to say, ‘slow it down.'”

That bill has something going for it. People who work inside the big AI companies signed on to support it, including a researcher at OpenAI and an engineer at Google DeepMind speaking for himself. When people inside the building ask for brakes, that ought to get your attention.

The AP also says the bill faces long odds in a Republican Congress. I’d say that’s putting it kindly.

The second camp wants to hand over the keys.

This week I wrote about Meta’s new AI agent, Muse. It reads your email, moves your calendar, and shops with your money. Two and a half million people downloaded it in two weeks.

So on one side, AI is going to end the world. On the other side, AI is going to find you a good deal on a Halloween costume. Both stories ran in the same week.

A columnist at Business Insider, Dan DeFrancesco, noticed the same thing, and he said something worth hearing. Both camps are selling you something.

His point goes like this. Some of the loudest voices warning about AI are the same companies building it. If you’re out front in a race, calling for everybody to slow down to an agreed pace helps you more than the fellow behind you. It also lets you help write the rules.

The other camp is selling too. He points out that Mark Zuckerberg spent years trying to put Meta between you and every business you deal with. An agent that shops for you is exactly that spot. Whoever gets there first sets the terms.

I think he’s right on both counts. Follow the money, and both stories make sense.

But he stopped short, and I want to finish the job.

He ended his column by saying we have to weigh the risks and understand the motives. That’s good advice, but it isn’t an answer. So let me ask the question he left on the table.

What is anybody selling about how the machine behaves?

The cage camp is selling control from the top. Ban this, pause that, license the rest. Even if that bill passed tomorrow, every approved system would go right back to work with no rules for how it conducts itself on an ordinary Tuesday, working for you.

The keys camp is selling access. Give it your passwords and let it run. How it behaves is left up to the company, and you find out when something goes wrong.

Neither camp has a dime riding on conduct. Nobody is selling that, and that’s the empty chair in this whole argument.

Let me go back underground for a minute.

The government did write safety rules for mines, and those rules saved lives. But no inspector could stand next to every man on every shift. What happened between inspections came down to how the crew worked. Did they check the roof before they went under it? Did they speak up when something looked wrong? Did they own up to a mistake before it got somebody hurt?

A rule from Washington can’t do that for you. Neither can a company’s promise. That part has to live in the work itself.

AI is headed to the same place. The machines are already running. They’re writing, answering, deciding, and now shopping and handling our mail. A ban on some future super machine does nothing about the ones working today, and handing over the keys doesn’t make them behave.

So here’s what I’d ask of any AI you let into your life, whichever camp you’re in.

Does it tell you when it doesn’t know something, or does it fill the hole with something that sounds good? Does it stop and ask before it acts, or does it guess? Does it show its work where you can check it? When it’s wrong, does it say so?

That’s conduct. It isn’t flashy, and it won’t send a stock up eleven percent. It’s what keeps men alive in a mine, and I believe it’s what will keep people whole in the age that’s coming.

Now, I’d be a hypocrite if I called everybody else a salesman and left myself out.

I built a set of rules for how an AI conducts itself, called the Faust Baseline, and it has a price tag. So point the same lens at me.

Here’s my answer. The rules are out in the open, in a dated public archive going back more than a year. You can read every one of them without buying anything. If they’re wrong, you can see where. That’s the only receipt I know how to show.

Senator Sanders says slow it down. Meta says speed it up. The columnist says watch your wallet.

I say watch how it behaves, because that part is left to you, and it always has been.

So I’ll put the question to you. If you had to choose today, would you cage it or hand it the keys? Or would you want to see how it works first?

Try it now for free, here is the file below, just load it before each AI session, thats it.

the-faust-baseline-public-release-1Download

” Attic Thoughts”-library – Intelligent People Assume Nothing

This post was drafted with AI governed assistance and reviewed and directed by Michael S. Faust Sr. before publication.