How do we control AI? Here’s what to say when it comes up at dinner

Image: ChatGPT/Kim Komando

Bill Gates just said on “Meet the Press” that AI could cause a billion deaths. Every news media is running that story today. The scarier part was quieter.

AI can’t easily tell a scientist curing a disease from someone building a bioweapon. Same questions. Opposite goals.

It’s already in your house. The AI that writes your reports writes a scammer’s fake bank emails. The AI that helps with homework can clone your voice and call Grandma begging for bail money.

It’s the oldest story in tech, really. Alfred Nobel invented dynamite to make blasting rock safer. It went on to blow up a lot more than rocks. Fritz Haber’s fertilizer breakthrough still feeds billions. He also helped launch poison gas in World War I. Eight years after Kitty Hawk, planes were dropping bombs.

So how do we control AI? Here’s the catch. A bad car gets recalled. AI can’t be. In 2023, Meta’s AI leaked online a week after release. Once it’s out, it’s out forever.

🍽️ Your dinner table answer

Your one-breath answer: “Lock it up, guard the ingredients and build the vaccine first.”

1. Lock it up before it’s out. The most powerful AI stays on the company’s own computers. You use it through an app, so the company can watch for misuse and cut people off. A federal lab has run more than 40 tests on top AI models, some before release, with the safety features turned off. Test first. There’s no after.
2. Guard the ingredients. A leaked AI can tell someone how to make something dangerous. It can’t hand over the ingredients.
3. Build the antivirus first. You can’t recall a computer virus. So you install antivirus. AI that finds software holes can patch them first. AI that could help design a bioweapon can help doctors spot outbreaks faster.

As I told the Daily Mail, I’m not afraid of AI. I’m afraid human competitiveness will outrun human wisdom.

🔒 Control the AI in your house

1. Pick a family safe word. If “your son” calls in a panic asking for money, ask for it. A cloned voice won’t know it.
2. Keep AI agents away. I’m talking about AI that clicks and types for you. Don’t give them access to your bank, your main email and anything that gets one-time passcodes. And as I said on the show last weekend, stick with the big AI companies and don’t install any other AI apps.
3. Be smart. Turn on two-factor authentication and automatic updates everywhere. That’s your personal kill switch.

📩 Send this to someone who always brings up AI at dinner.