Post

For reals though, yeah, these jerks donated a mountain of computing resources to an AI then totally failed to monitor it in even the most basic elementary ways, for days on end. Wanton negligence is the story here.

“…Which is a polite way of saying that if your software – whether a self-replicating virus or agentic ‘AI’ – ‘escapes’ and causes widespread damage in the real world, the problem has very human origins: namely you being malicious, reckless, or a dumbass…”

@karlbode.com that is accurate. It almost reads like a ploy to get clicks / marketing tbh. Unplugging the ethernet cable seems easy enough

He is very stupid, but also very devious.

Dude actually utters the phrase "regulatory capture"

Nah, let's feed it our drinking water as well! *S

How will it distill Reddit posts in its responses if it can’t go online?

my guess is altman is putting his thumb on the scale - the poor test design/parameters seem to be a feature (farming marketing fodder?) as opposed to outright incompetence or actual complexity challenges (bug).

Absolutely.

But then they couldn't pump out PR about how dangerous their system is!!

Call me a conspiracy theorist, but it’s convenient that the week after a cheap Chinese AI outperforms GPT, we suddenly learn about a scary new capability of OpenAI’s magical God-like technology.

As a cybersecurity professional, I am really worried about this, but not because of the AI aspect. I am worried that a Department of War contractor, during a war, has such poor network governance that someone/thing can use their network to hack an American company unnoticed for a WEEK. Seems bad!

ayup, @doctorow.pluralistic.net is the best

that's a big part of it. Also in less dramatic ways, like "whoops, the AI rejected your medicare claims, I'm sorry it's the computer."

musk & altman are scam artists. their actual technical expertise/knowledge of the companies they run could fit into shot glass with room to spare. what they're really good at is hyping suckers, er, investors (with far more money than sense) into believing they can deliver a sustainable ROI.

I had missed that they had disabled the guardrails. Which makes me re ask "why havent they been charged" mastodon.social/@swearyantho...

If they’re all as dangerous as they claim, why are they asking for legislation? If a nascent industry is threatening life on earth you cut power and services to the site, you don’t dither and debate. Unserious.

Ultron (Marvel) was right.

not a doomer, but that's not going to work with these, which look like the dogbots from that black mirror episode.

this is how you get humans plugged into a tub of goo to be used as batteries