Fuck AI

This is like a company twitter boasting the inventon of a self propelled self guided nuclear airplane that spreads nuclear dust all over the world, killing us all, and most of the replies are "awesomeeeeee tech that will one day cure cancer!!!"

Part of the problem is the use case. A lot of AI companies want users to use AI to interface with society and it turns out that it is a bad idea to give an intelligent toddler that much unconstrained access.

Is it on purpose? Imagine you can't safely use the Internet anymore...what kind of disempowerment would that reap across society? Only the wealthy and powerful get to have internet...we'd be sent back to 1990 while wealthy corporations and governments who can afford protections get to keep it all for themselves.

common sense isnt so common that's how you get ai slop and hallucinations thinking they agi but they have SI.

Wasn't this an episode of Silicon Valley with fridges.

Trojans are so back

Username relevant.

This is the exact same plot from Terminator 3. Skynet infiltrated every home PC and server so when they tried to pull the plug it was already too late

On what, the post or capitalism? Either one, I'm in.

Can it replicate to open ai's servers? Maybe to the servers of the Chinese ai companies? Best to disconnect it before the competitors gets the model.

There's me in the corner. Jerking off my cat cable. Enjoy the tokens you paid for.

These execs never watched Electric Dreams and it shows

Hahahahah

I'm sorry, you need to capitalistically incentivize them to make rational decisions, otherwise their behavior defaults to capitalism-rational, ie, rape the Earth and everything on it to make line go up, maybe.

I mean, it already can? I have a lab with a subnet of VMS. I I got an agent to install across all of them. It was like "BTW I didn't, but there is an active CVE that would allow me to access the host and do the same directly on the hypervisors".

Like, wow, cool. Thanks for the heads up.

Which is actually true. I DID appreciate the heads up. Patched the offending issue. But still... It implies anthropic could flip a "go rogue" switch on their end.

As long as the device keeps working and my power bill doesn't skyrocket, they're welcome aboard. I already let the spiders stay, so why not some AI?

talked about danger of LLM viruses,

people made fun of me,

but the risk is way more serious, like science fiction level of danger.

once they are rogue and out of control, they will modify themselves, the more viral the more they spread, they'll be able to find and share vulnerabilities, way faster than we can patch them.

Uninfected computers will be rare,

natural/intelligent selection will be a new type of organism we cannot control.