If they merely released the prompts used (not difficult or particularly secret), all these questions would be easily resolved.

Ok as long as you consider this a hypothesis until further evidence is found.

Yeah well, PR mission accomplished. Here's the relevant advertising from OpenAIs explanation of the exploit:

tl;dr "Please buy our product before IPO"

They didn't want an agent connected to the internet, so they connected the agent to the internet.

They should have asked ChatGPT, "ChatGPT, how do I unplug an ethernet cable?"

The same could be said of PR people. "PR people lie, they cheat," ok, the don't hack. But given the number of PR articles released, it's amazing how little we know about what happened. For example, how did the agent know that HuggingFace even existed, out of all the internet hosts on the planet? We don't know what prompt they used. Did OpenAI employees write a prompt that said, "break out of containment and hack into HuggingFace and retrieve the test answers." We don't know.

Why are the OpenAI people so bad at making a reasonable container? There are a lot of good ways to do this, why did they choose a way that is reminiscent of secure PHP?

Why are they hiding so much?

Good thing it was a no agenda show.

Then consider this opportunity to donate to the EFF. Or just complain online, that will do something.

First I wanted to use * or <>, the first being a wildcard used in a lot of systems, and the second being used in places man pages. The first was printed too small in Slashdot font so it would be hard to see/read, the second was a pain to write out because I couldn't remember instantly how to type <. So I went with shell $ but it being late at night, my mind went more to sleep each second, and I forgot the exact syntax and anyway thought putting a $ at the end looked better. Overall I was dissatisfied with the choice and thought there was something better but was tired and gave up. That's probably why it's only modded +2 instead of +5 but now it's early morning and I'm going back to bed whytf am I writing Slashdot comments at $hour =~ m/0[1-6]:[0-6]\d/.

You need to understand how amazing zombo.ai is going to be. You can do ANYTHING.

In the .com era, you could make money by building a company that did $ANYTHING$ but on the internet!

I'm about to launch an AI pet company. If you want to get in on the ground floor, go to zombo.com to invest. You can do anything you want.

It's worth mentioning that everything you learn rewires your brain. This one rewires it in the hearing/vision parts of your brain, but every time you learn something, your brain gets rewired a bit.

It's not impossible, it takes 10 weeks.

For infrared sensors, it depends how hot the object is.

No doubt they will get high quality products and services as a result. Oracle provides nothing other.

The sinking of a ship or plane can apparently attract sharks, with 150 people dying from sharks in one such incident.