Goldstein suggested more radical approaches, including using the courts to hold AI companies accountable through lawsuits when their systems cause harm.
The suggestion that this is a "radical approach" might actually be the most insane part of what is already a fairly insane article.
This is really what blows my mind the most. With all this talk about how much power LLMs and diffusion models use companies are still constantly cramming it into places where it's just running all the time passively doing things no one asked for.
Overall power use by these things would probably be cut down by an order of magnitude by just limiting it to directed, intentional use only.
It's a basic AC rectifier, the resistor represents an arbitrary DC load. You use similar circuits all the time, though generally with additional failsafes and some mechanism of smoothing out the rectified current.
"Conduit" is the word for those tubes for wires. Probably a shared etymology with "conductor" though.
Having the pipes in the mortar/bricks sounds like a maintenance nightmare.
What blows me away is how much of it is so obviously AI but people seem to just not notice. You do see it called out from time to time, especially on Lemmy, but in the wider Internet it's much more common to see people just engaging with it.
Ah, if you need to build a .NET project that makes sense
Nuget is a the .NET package manager. Like npm or pip, but for .NET projects.
If you needed it for a published application that strikes me as fairly strange.
...no? Why would it be?
Let's table that discussion.
Tap for spoiler
The meanings of "table" as a verb in US vs UK parliamentary usage are literally opposites. With the US meaning being to stop discussing or put aside for later, while the UK version means to begin discussing.
This actually caused confusion during allied meetings in WWII.
I'm okay with a DM ruling that it's possible to cast it in such a way that someone is taken off guard, sure. Maybe a performance or deception vs hostile creature(s) insight rather than the typical stealth vs perception when determining surprise from sneaking, which is not RAW, but I think sounds reasonable. I'd definitely not consider it to be an automatic aspect of the spell at any table I ran.
And you absolutely could not avoid a fight and just walk away from the situation with plausible deniability because you "only insulted them".
This is probably a diffusion model. LLMs don't create images.
On top of that they say that these sorts of behaviors only arise when the models are "stressed", and the article also mentions "threats" like being unplugged. What kind of response do they actually expect from a fill-in-the-conversation machine when the prompt it's been asked to continue from is a threat?