Spoilered as a pseudo content warning but uncertain of how to tag: advances in robots that would be detrimental to human survival should things go bad (but it also has great uses and is still counterable)
So this shit is kind of scary at first glance. It's a Boston Dynamics Spot robot dog (IDK if they're still BD anymore) doing some very much improved pathfinding over particularly difficult terrain; """extreme""" parkour (for a Spot robot, sure, but it's all pretty... Basic parkour, honestly.)
Someone referred to it as also being known as "pursuing human prey at high speed across varied terrain".
And while that's a valid concern to have in general, this is the worst form to sweep humanity with as an AI/robot uprising. Ambulatory hunting, and all of the decisions around it, took Life a looooong fucking time to get even close to right, and at the same time, the other side of Life kept advancing as well. The best form of hunting came when we rocked up - extremely versatile, fast-learning, highly adaptive and generally enduring creatures that also happened to be bipeds. ALL of the evolutionary gains surrounding and supporting that achievement are, generally, lacking here, especially in the factor of number of units.
These puppies ain't T1000's. They're not indestructable. And they are, and likely always will be, far too expensive to make in ways that are usable in this capacity, effectively. Especially if the goal is melee, or capture. Unless they also put ranged weaponry (which I know they're trying), these things are going to run up against the same thing we used to take down fucking Mammoths:
The trusty spear. A nice, long, bracing pole weapon, commonly found in construction sites and Australian backyards under the guise of 'star droppers', these little fuckers aren't going to get shit all done if they can't get to you at the far end of a steel spear.
Sadly, and quite earnestly, if an AI/robot uprising is going to sweep the human populace from the world in a clean and fast manner that preserves the rest of it, the short film Slaughterbots captured it all pretty well (additional content warning: distressing even for a fictional short film; school shooter vibes, terrorist attack found footage vibes).
Tiny, low cost, swarmable drones that can simply flood a space and, even without the specificity of face-targeting for a precise headshot, tiny shaped charges would be more than enough to kill if faced with multiple per target.
And that's if they even care about leaving a majority of infrastructure. If they don't, all the tools needed already exist. Pretty sure there's enough extant munitions, even w/o resorting to nuclear weapons, to old-school drone-strike us into oblivion.
The key thing to remember is we're A: not dead yet, and B: people seem too focused on using AI for... Capitalistic shit than anything else, so I think we're safe in that regard, for now.
ALL THIS BEING SAID.
This parkour stuff is actually a really useful step forward. Why? These things may not be mass-producible to a swarm-killing extent, but they certainly can be made in quantities usable enough for Search And Rescue and other Disaster Relief Efforts!!
Being able to navigate rubble and other extremely dangerous environs, scan using RADAR, LIDAR, audio, various heat-sensing technologies, etc.
Imagine if we could send a handful of people-piloted Atlas robots, followed by tool-carrying Spot robots, in to do maintenance on a reactor, or in place of the Fukushima 50. Big Dog robots could carry the injured in protective casings, or deployable shelters.
It's a very promising field. It saddens me that they want to put fucking GUNS on these things, but the problem isn't the robot - it's the gun, and the person giving it the command to go and do things with that gun. Take that away, and we might have a better future with these.
Sorry for the rambling. This is not-high Bear talkin', too; normally this level of ramble would be reserved for post-fat-tokes, but here we are.