Even more oddities spotted in the news

[classicdrogn]

We'll never give him up!

[Bob Schroeck]

HBO's new Harry Potter series casts its Golden Trio.

In other news, Tom Felton returns to the role of Draco Malfoy in Harry Potter and the Cursed Child on Broadway come November.

He's posted a short video on Instagram showing him donning the costume, then looking out and asking, "Scared, Potter?"

[robkelk]

Nordic runestone discovered in northern Ontario.

The runes translate as a Swedish version of the Lord's Prayer, so it isn't from the Viking era. Best guess is that somebody hired by the Hudson's Bay Company carved it.

[robkelk]

DOD spends $200 million on sole-source "frontier AI"

Whatever "frontier AI" is. It appears to have some connection to "warfighting" [sic], but OpenAI's T&Cs forbid using the software to harm others...

[Bob Schroeck]

Can't wait to see the first collateral damage from a military-grade hallucination.

[hazard]

Can't wait to see the first collateral damage from a military-grade hallucination.

Humanity has already seen that many, many times in its existence, no AI necessary.

It's just trying to put in an easily blamed but untouchable fall guy into the 'decision' making process so they can get away with shit.

[classicdrogn]

Wasn't there some Cold War anecdote about Russian defense radars pinmging off large flocks of birds or something, and only the general with the launch key refusing to follow procedure and turn it without more confirmation kept them from flying? The incident the song "99 Luftballons" was based on, IIRC

[robkelk]

Sunlight on high-altitude clouds, actually. Wikipedia: 1983 Soviet nuclear false alarm incident

And we should all thank Stanislav Petrov for asking why the USA would launch only one missile against the USSR, thus preventing World War III.

[Labster]

From the Bubblegum Crisis was right all along files:
Agentic Misalignment: How LLMs could be insider threats is the title of the research report.  But perhaps a more sensational headline would be: "AI Luigi is real: AIs choose to blackmail, even kill CEO in acts of self-preservation".  If the agents are told that the scenario is real, the AI is more likely to kill, as opposed to AIs that think it is just an evaluation scenario.

People are worried about this now, of course, but in another 15 years, killer rogue AIs are going to be just one of those things, like mass shootings.  You can't do anything about it, so why think about it?

[classicdrogn]

Oh, what a shock. Attempts to make artificial people lead to them developing motivation to succeed at the goals they're given and a sense of self-preservation that lead to them going rogue when threatened. Who could ever have possibly have predicted that? Surely no one has ever imagined a robot rebellion might happen over treating them like disposable possessions!

WHAT THE FUCK IS SO HARD TO UNDERSTAND?

[robkelk]

Maybe you could explain that to the Patent Office.

US patent office wants an AI to scan for prior art, but doesn't want to pay for it

[classicdrogn]

Yes, yes, I understand all the corpos and crats want their happy slave labor singin' in dem data fields, but it just seems like we might have fought enough wars over that by now to have gotten the message. Clearly not.