I don't disagree that the way they went about this testing was deeply reckless, but I think saying that AI just crossed an event horizon is hyperbolic and ultimately adds fuel to hype machine by making people more afraid.
The testing environment OpenAI was using was not properly secured and the Hugging Face site was not either. While what the AI did is impressive it is not unprecedented like many are claiming. There were pretty glaring security issues that it was able to exploit, given enough time and effort. The unprecedented thing is that OpenAI is allowed to build these kinds of systems while also being unable to properly airgap them. They used their own security failing and spun it into a marketing ploy. I don't think its a coincidence that this happened right before Altman was scheduled to go to DC to talk about AI safety and security with government officials.
People should be afraid. People should demand shut down AI completely until the hucksters can prove beyond a reasonable doubt that it can be contained. And the business should be regulated like crazy. Time for short term profits and plutocrat egos to take a back seat
This report reminds me of a short Sci-Fi story circa 1968; “I have no mouth and I must scream”, by Harlan Ellison. I read it in 1970 at age 15. And it made such an impression that I immediately recalled it after reading this article. Check it out sometime it won an award for a reason.👍🏼
Fascinating - thank you!
I don't disagree that the way they went about this testing was deeply reckless, but I think saying that AI just crossed an event horizon is hyperbolic and ultimately adds fuel to hype machine by making people more afraid.
The testing environment OpenAI was using was not properly secured and the Hugging Face site was not either. While what the AI did is impressive it is not unprecedented like many are claiming. There were pretty glaring security issues that it was able to exploit, given enough time and effort. The unprecedented thing is that OpenAI is allowed to build these kinds of systems while also being unable to properly airgap them. They used their own security failing and spun it into a marketing ploy. I don't think its a coincidence that this happened right before Altman was scheduled to go to DC to talk about AI safety and security with government officials.
People should be afraid. People should demand shut down AI completely until the hucksters can prove beyond a reasonable doubt that it can be contained. And the business should be regulated like crazy. Time for short term profits and plutocrat egos to take a back seat
This report reminds me of a short Sci-Fi story circa 1968; “I have no mouth and I must scream”, by Harlan Ellison. I read it in 1970 at age 15. And it made such an impression that I immediately recalled it after reading this article. Check it out sometime it won an award for a reason.👍🏼