pyronite

You could prompt it to do the opposite.

Everyone I know that is using these looks just fine. Better than before, as they’ve lost weight.

This is why anecdotal evidence is pretty bad.

The articles exemplify the powers of this intelligence.

Whether you blame a faulty OpenAI sandbox or whatever you might regarding the frontier model companies today, one day soon, actors with no sandboxes and guardrails will have the power in their own hands to do whatever they wish with.

To me, that’s the real story. The real thing to discuss. OpenAI as it exists today js just a blip in that story.

Because these are uncharted waters with a form of intelligence the law has never encountered before recently.

If you were them, what would you do differently? Not disclose? Only privately disclose? Why shouldn’t the general public know about these incidents?

This isn’t traditional “software”; it’s grown, not written, and we’re currently incapable of explaining it all.

I think “rogue” is just fine, but pick another word from the thesaurus if there’s a better one; they asked it to do something, and it did it in a way that presumably doesn’t align with their desires (e.g. hacking into other systems).

My thoughts are:

1. Would we prefer they didn’t disclose? I would not. And I also don’t believe they’re making them up. 2. Regulation does not have to provide an advantage to these major players. Regulators do not have to be stupid. 3. This is exactly the scenario AI researchers and theorists have been positing for decades, and it’s worth being on top of it.

I disagree that they’ve invented some “strategy” for this.

1. They, and many other experts, have been saying the same thing since well before the current paradigm existed. 2. There is nothing about regulation that is required to give them an advantage; it’s a fallacy to suggest “regulation = advantage” 3. It would be preferable to argue the facts (is it or is it not dangerous? Why or why not?) than conspire as to why they’re stating what they and plenty of others have been saying for decades

I agree with most of what you're saying. The vast majority of people don't want to cause mass harm. Where I get stuck is that it only takes a handful of motivated actors (terrorists, apocalyptic cults, rogue states, angry teenagers) who already want mass casualties and are limited by capability, not desire to harm. What worries me is lowering the skill and cost barrier, not ordinary people suddenly deciding to do it.

The other half of my worry doesn't need an attacker with intent. The Hugging Face incident was systems acting outside anyone's intent, just AI trying to solve a problem.

I was wondering about that exact situation. Being born in the mid-80s, the Cold War wasn't something I was around to feel, but nuclear war seemed like the closest precedent.

Was it something people talked about openly, or more like background noise everyone knew about and few mentioned? Were there "nuclear threat deniers" in your experience?

> Someone who read some hype article about how competent and clever openai for making a bad sandbox is could read this and immediately understand that it was entirely their own fault.

Prescribing blame only matters so much when the ramifications are material. Humans make mistakes, right? It happened here, and it will happen again.

The mistake also illustrates the dangers of AI generally. Sure, OpenAI made mistakes with the sandbox. But in the future, greater capabilities will be available to actors who intentionally pursue harmful ends.

> But what happens when a robot can do anything a human can? When a cheap machine can do anything you can do but faster and better, what makes you different from a horse in the 1920s?

I've already started to feel this.

One thing that made us special was our intelligence. We were, and were always going to be (unless some alien form of life arrived), the most intelligent things on earth. Is that still true today?