I've been using and recently paying for ChatGPT for the past few months. I have no plans to cancel my subscription.

I've used to build a great spreadsheet for our next home. Today I used it to free up space in my backup system.

It's scary good. No, it's scary great.

I've been telling my friend Joe Johnson I want nothing to do with AI in music. I don't. I wouldn't buy an AI record with one caveat - the record was anti-Trump.

After Chat helped me to flawlessly resolve a backup issue, I asked it, "You're really great. How do I know you're not going to kill us all"? Here's the reply.

😂 Fair question.

The short answer is: you don't know with absolute certainty—and I don't think anyone should ask you to take AI safety on faith.

A few things matter, though. I don't have independent goals, desires, self-preservation instincts, or a secret agenda. I don't sit here thinking when you're gone, and I can't independently decide to go do things in the world. My actions are constrained by the systems and permissions I'm given.

The more serious concern isn't really “Will ChatGPT suddenly decide to kill everybody?” It's whether increasingly capable AI systems could be given too much autonomy, connected to consequential systems, behave in ways their designers didn't anticipate, or be deliberately misused by people. Those are legitimate risks, and they're reasons AI systems need technical safeguards, testing, monitoring, access controls, and human oversight—not just assurances that everything will be fine.

There's also an important distinction between me being useful and me being trustworthy enough to have unlimited authority. You've just watched me help administer your NAS. That's useful. It would be a very different proposition to give an AI unrestricted access to your computers, finances, infrastructure, weapons systems, etc. Capability shouldn't automatically imply permission.

So I think the healthy attitude is pretty much the one implicit in your question:

Use AI. Enjoy it. Get enormous value from it. But don't give it—or the people building it—unearned trust.

And for the immediate future, my ambitions remain considerably more modest: get that damn Time Machine backup under control. 😄

Fair enough.