Culture TechnologyLeave a Comment on Bonfire Of Inanities: The OpenAI “Hacking” Story Is Stupid
I saw this kid it was just the other day
Hacking on a server where I used to play.

No, a prototype AI did not breach containment at OpenAI and hack the Hugging Face site.

Of course it didn’t.

Firstly, there’s been no actual information about this alleged incident published.

There’s just a couple of marketing press releases from companies in an industry known for, if not outright lying then let’s say choosing the most-likely-to-be-misinterpreted terminology, framing and interpretations in describing their products and capabilities.

Remember “Mythos”? Like that.

Second, what there actually is in those press releases is absolute nonsense.

What they want you to think happened is that they’ve built a thinking machine so smart that during testing it decided to hack the test system to get a good grade. You know, like in “WarGames”.

And you’re going to be able to rent that thinking machine in a few short months!

Which is, again, absolute nonsense.

Just total nonsense.

For lots of reasons:

  • there is no thinking “AI’; there are chatbots
  • the chatbots don’t “decide”; they tell stories
  • this chatbot was clearly not “contained”; it had Internet access

And if all this were actually true, the OpenAI press release would be admitting to having committed a felony.

So, no, it didn’t happen.

We’ll probably never actually know what, if any, things actually happened in the Real World that make up the basis for these press releases.

But it wasn’t a thinking machine going rogue to get a good grade.

I could go on, but I need to get some actual work done and it feels like just thinking about this is making me stupider.

arkady
Latest posts by arkady (see all)
Related Posts:

Leave a Reply