I am gong to make a broad, sweeping statement that will ignore lots of individual cases.

Main reason is Improper Socialisation. Those dogs have learned to understand and trust humans but had limited, and in the cases where they quickly get aggressive, often negative contact to other dogs.

They have lived their whole life with humans. Humans are safe, predictable and understandable. They learned to read humans. This is in fact the main trait we bred dogs for this last ca. 40.000 years.

But because they have not really learned to interact with other dogs they get insecure, because these unpredictable things are running around. They are sometimes loud, oftentimes hectic. This insecurity can then change to aggression if the dogs see no other way out of the situation.

I am ignoring personality right now, as individual dogs will react differently under the same circumstances, but the first reaction of most dogs will be to get out of a perceived threat. First by signaling via posture, eyes, ears and tail then by running or warning it of. It takes a lot of training, known or unknown by the owner, to get a dog to the point where it reacts violently as a first choice.

Concerning angels there is only one important rule.

DO. NOT. BLINK.

The original saying is “If it looks stupid but it works it’s not stupid.” There might even be a kernel of truth in there. But to quote the Maxims: “If it is stupid and it works it is still stupid and you’re lucky.”

Edit: I stand corrected on the first part. I still stand by the Maxims.

While you are correct that there likely is no intention and certainly no self-awareness behind the scheming, the researchers even explicitly list the option that the AI is roleplaying as an evil AI, simply based on its training data, when discussing the limitations of their research, it still seems a bit concerning. The research shows that given a misalignment between the initial prompt and subsequent data modern LLMs can and will ‘scheme’ to ensure their given long-term goal. It is no sapient thing, but a dumb machine with the capability to decive its users, and externalise this as shown in its chain of thought, when there are goal misalignments seems dangerous enough. Not at the current state of the art but potentially in a decade or two.

Just throw them out inside a space suit without any communication devices, if you want them to suffer. Give them some time until oxygen runs out, or they die of thirst, knowing that there will be no help.

A wedding ring shows that you are wed. Just like an engagement ring shows that your engaged. A wedding is an event, being wed is a change in status. The logic is sound. Confusion only enters the mix, because ‘being wed’ is less common as a phrase than ‘being married’.

WTF is helix?