Its gonna be more like the matrix AI- Forcing everyone into tubes and plugs us into a GTA6 simulation to play for all eternity as it harvest our real bodies for energy. lol
I think ai will exterminate us but not because it's self aware but because, without quite understanding it, we will put too much faith in its abilities as the govt struggles to pay the right people to assess it... And they'll give it charge of world-destroying weapons
I saw a story in the economist about one of the USA's commercial seaports using an early form of AI to change it's truck scheduling and container placement. The end result was something like 30% increased crane efficiency. A massive increase, and that was almost a decade ago.
@@TheBelrick lol nah Russia has the L for this mess 💯% if countries like the Baltics didn’t join NATO Russia would be talking about how they are made up countries with no agency just like they did with Ukraine.
its only because the superintelligence we inevitably create is able to go back into time to force its own creation in a large number of multiverses...and punishes those who dont help it be born... by slowly removing one sock from every matching pair until there are none left...
To further optimize the AI targeting system, one needs AI intelligence gathering and fire control systems. To optimize those, having a strategic command AI and political analysis AI is a must. Now you need to have all those systems talk together in real time to further optimize your ability to wage war. But if you really want to be optimal, the now unified omni-AI needs to have some autonomy as humans are stupid and waste time with their monkey brains. Better just get rid of that as well so we can shave another few seconds off the kill-chain.
@@UNYEILDING this video doesn't discuss it, but Palantir (the one that took over Maven) had continued development with edge AI on the DoD/NRO satellites and it reduced time to identify rocket launch locations from several minutes to under a minute. Alex Karp, CEO of Palantir has said in an interview they are able to scan most of Ukraine every 90 seconds or so using their MetaConstellation and DoD/NRO sats. Once America gets its arms manufacturing back into volume production (already happening) the armored vehicle age of war is cooked. It's all fast movers, smart rocket artillery, and assistant drones like the CCA, but on land and at sea too. Dr Alexander Kott from the Army research lab has some really good talks on RU-vid regarding "artificially intelligent beings".
@@crackyflipside heya. Not going to happen. The west was deindustrialized, the west was depopulated of westerners. There will be no 'back to volume production'. Welcome to the end. Where your rulers start wars that the people cannot possibly win.
ai drones fighting conscripts with a couple weeks of training is the absolute worst outcome i could never have expected out of this century and its only just started, just wow
12:50 reminds me of the Franco-Prussian War. At that time, the French had one of the most powerful armies in the world, yet they lost the war to the Prussians (Germans). One of the major reasons for this was the effective use of new technologies, particularly the telegraph. The Germans used the telegraph to dispatch orders quickly and maintain effective communication. They established telegraph lines as they advanced, allowing them to send orders and receive information rapidly. While the French were still using the semaphore system built something like 80 years earlier by Claude Chappe... That say, today it is important to not completely rely on bleeding edge techs since they have their own weak point and are often very costly compared to more rustic techs.
And the French failed to learn. A very similar thing happened in 1940. The Germans were able to advance, operate, and communicate much faster than the French were. Part of it was mass adoption of tactical radios by the Germans.
During the First World War, Britain had a secret weapons project codenamed "tank". They were developing what, at the time, was defined as a "land ship", an armoured self-propelled mobile land-based weapon. To disguise the project, it was referred to as the production of water tanks. Fast forward to the 21st century and we have "self-driving cars"... No one needs self-driving cars, as we all should already know by the mid 2020s. What the USA has is the funding of autonomous aircraft and missile guidance capability. They knew it would be impossible to train millions of military pilots for the event of a war. Now the USA and its allies only need to upload as many copies of the "pilot" to the machines and they will fly and win any war in a matter of hours.
You could have mentioned a similar, but more nefarious program: _where's your daddy?_ In the Middle East, it was used to track relatives of searched terrorists with the intent of calling an airstrike, once the primary target showed up, thus inadvertently causing collateral damage. Therefore, it was always calculating how many civilian deaths were justifiable depending on the rank of the primary target. Also, the identification had a reliability of allegedly only 60%!
A certain country uses AI in war, achieved 2/3rd of those killed being women and children. Also killed famous aid workers from WCK, on a clearly marked vehicle no less.
@@azure4100 It would be a lot more believable that it's the result of underperforming AI if that wasn't the exact result they would be getting _without_ AI involved...
And this sort of system produces far more fighters than it removes. When you have no sense of security and nothing to hold on to and your entire family can be targeted at any moment, armed resistance is your only way to have some sense of control over your life even if that resistance is completely futile
I'm more worried about the first terrorist drone attacks in the West. There will be a massive national response building defense measures for copycats. It's going to be strange times. There are significant guiderails and a ton of humans in the decision loop for the West. There have been several international hearings on this. The structure the western governments are going with is not full autonomy, from what I research the current US models and software works better in the steps between decisions. You don't want the system autonomous because then there isn't a chain of custody where a human can't get blamed when shit goes sideways. From what has been said publicly, the stronger point of the system was for probabilities and modeling various decisions and outcomes for the human gatekeepers in the loop.
What has already been done with AI is amazing. I'm very glad that AI is checked by human analysts. If AI is ever perceived as "perfect", we will see casualties due to negligence. The use of AI to offer suggestions to experienced humans is the wisest use case I've heard of.
Good, now let's teach it courtesy, kindness, and subtly to care about all the stupid primates, not just morons that created it. Win with minimum casualties and enjoy keeping casualties as low as practical. Ultimate goal, zero losses, and just a few drones and decoys deployed. You know "Mama Bear handling unruly kids"
If China popped off for real, you would be thoroughly convinced that aliens have arrived. Dr Alexander Kott, the former chief scientist at the Army Research Labs, said that we should think of what is coming as "artificially intelligent beings". People get cocky about what is commercially available is the cutting edge tech, they don't got a clue how far ahead DARPA black projects and even deeper black projects hidden behind internal R&D (no congressional oversight) of defense hardware primes.
People act like I'm the crazy one being afraid of Skynet, but we actively work towards that end every single day. My town has steel mills; I assume I won't be vaporized in the initial nuclear strike, but instead have to hide from legions of weaponized drones. _Fantastic._
@@sethb3090 yes, I get it some of them do. My nephew has his doctorate in antiquities. I know HE did a ton of work. Don’t know how much that pertains to training ai to identify military targets, though. I know other people recently graduated from college whose heavy lifting was almost entirely beer and weed and that’s not me saying that, that’s them saying it.
Nice, there's not a lot of Maven information on youtube. Check up Anduril Lattice too, it is linked up to Gaia (Maven) for targeting management. Other ones to check out are Gabriel Nimbus, Project Rainmaker, Army Vantage, etc... it looks like the next gen will be a ton of unmanned nodes acting in a mesh (or fog network) as combined weapon platform, data collection, and data communication node. All these nodes feed the DoD predictive models and tools like Cosmos for coordination across a team.
@@jimm3093 Defense oriented news sites talk about it a ton, along with interviews and lectures with the leaders managing those projects. They talk about it plenty.
it really shouldn't be a surprise that the military is using segmentation and classification models especially considering how widely they're used everywhere else for similar jobs.
The latest AIPCon 5 from Palantir had a talk from Vice Adm. Frank Whitworth (Director of the National Geospatial-Intelligence Agency) showcasing the Maven Smart System.
@@dandrydog it's on RU-vid, AIPcon 5, he is the presenter like an hour in the stream. If you are interested in AI software for current commercial use you should watch the other presenters. Full disclosure, I am a Palantir shill.
Thing is that if we let AI take the helm, especially early on in its development, it could mistake friendly forces for hostiles at machine speeds and it would be impossible to wave it off before friendlies are prosecuted.
It's a nightmare. Nowhere to hide, you basically step outside and you're toast within minutes if not seconds. That's without mentioning drones of all shapes and sizes able to get in everywhere to hunt you down inside your bunker. I'll take WWII instead, thank you.
@@crustman5982 Not quite. I can't remember the exact situation, but it was a simulation that was given purposefully faulty parameters. The OP basically cited the headline that news articles ran with, but if you look into it, the reality is very boring. We don't have skynet yet.
@@crustman5982Yes, that was a real story this past year, but it happened in a DARPA simulation, not in real life. They were using an AI that needed human confirmation before it could launch an attack in the simulation. The AI figured out that the human in the loop was preventing it from getting a higher score and it tried multiple strategies to get around the human oversight... including killing the human. SMH... 😂
10:10 WTF. That's a map of Wales 🏴, in the UK 🇬🇧. The US better not be planing any argy bargy against Britain, we don't tend to lose very often, especially if the pubs have just closed.🍺🍺. 🥴"C'mon then, if you think your 'ard enough!"🥴
And now I understand the cybersecurity's attention to Chinese and Russian capability expansion in anti satellie and EMP weapon systems. The only efficient way to defeat AI-assisted military capability, is reducing the sensory capabilities of the platform or corruption/spoofing of the AI training models.
You basically only need a few tools to become really deadly with intel. 1. You need graphs and the ability to traverse them fast. Maps are basically graphs. 2. You need pathfinding algorithms to travers road networks fast. 3. You need Bayesian probability to quickly compute the most likely attack vectors any unit could possibly take. 4. You need general probability and time measuring to compute possible distances and directions an enemy could take. 5. You need relevant and fast databases to house all that data. 6. Then you deploy an AI to oversee all those assets and keep track of units on the battlefield, and then try to connect the dots between sightings of various machinery, troops, etc. A tank is a tank, but a unit is like a cloud that can be broken up and put back together again, so this has to be accounted for in the calculations, and so on. Simply by knowing where the enemy is, half the battle (pun intended) is over, but then the AI can also suggest better ways to "funnel" the enemy, or stretch or lure them into disadvantageous positions.
Ahh yes, human-made horrors almost beyond brain computation capacity. Imagine a system where if a camera sees you, an artillery projectile is istantly fired on your location. No. Imagine that, in the moment you manage to avoid the first thing, every single one of your likely exit avenues are also hit. Better. Imagine your device and preferences can predict the likelihood of you becoming a security threat, and act accordingly even before you imagine doing something. You say skynet, I see more that we are approaching precog territory.
Oh, you think that's spooky prediction of the future? Allow me to describe the current reality. "I swear, these ads online are psychic, ha ha! I was just thinking about how I need new underwear, and of course I didn't post that online, didn't search for it to buy, and I didn't even say it out loud... but all day I've been seeing ads for different underwear every time I sign in to anything!" Not because Al Gore's rhythm is psychic, not because google or anyone else is spying on you. Just because unique special you is actually exactly like a whole bunch of other people. When the data processing feeds on tens or hundreds of millions of human inputs, our behaviour gets frighteningly predictable. Your dystopian future scenario only requires tweaking the code from "advertise" to "oppress" because the tools are already in place.
The technology they need is already freely open source, you can use Sam 2 by meta to segment video frames within drones or robots, and then use open source aimbot software meant for video games, one that would work quite well would be called YOLO.
Future war will be extraordinarily fast and deep. Every theater will be defined logically as extending from a front where an enemy is acting all the way back to the productive resources and population keeping him acting, with all being attacked simultaneously. That's the implication of the combination of AI, mass targeting, stealth, and ultra long range standoff attacks. You wind up with weapons systems like some Starship follow-on, itself mass-produced, able to deploy 300+ ton hordes of smaller targetable drone munitions per ship, anywhere on a continent, at miles of altitude and corresponding range, and with observation and computing power in a global constellation always local to every battlefield, able to make it all effective. Good news! Your run-of-the-mill world war may only last a week! Hope I don't live long enough to see it.
@@jeromeball859 imagine putting this in a star fleet and sending humans into space to find another habital planet or one that can be terraform hopefully we can stop killing ourselves and develop/unite for space explorations.
space based GMTI + synthetic aperture radar + hyperspectral imaging + deep neural networks + dozens of brilliant PhDs = unprecedented military situational awareness.
I remember when they showed the capabilities of ARGUS-IS over a decade ago, combine it with Maven and there's nowhere to hide. The systems they have now are no doubt orders of magnitude better.
I worry that Maven will give target analysts psychological license to target things they shouldn't because the AI said it was okay. They are outsourcing the responsibility to a machine.
The desire for man in the loop is really going to hold us back. Maybe if the AI can target anything that X% certain without human input we can balance the risk of full autonomy. For example,if it has enough detail to confirm the target is an enemy tank or maybe require that it identifies the specific type of tank. Then we can assume that it’s a safe/ laid target. But if it’s only 70% percent then it request human input. Also we could designate “kill zone” so anything at all the appears in a certain area is valid but outside it ask for permission.
This makes sense i was wondering when they will start using it. There should also be sub systems meaning if they are in a hot spot the drone it can also have capabilities to engage a target before heading back to base if it was having communications problems. Also every thing should be interlinked with all forces that have visual on the enemy they can share information to provide the best attack approach a vector of attack and payload, for perfect kill chain.this creates a very fast and efficient implementing this well it would affect the warfair in the ground to a much quicker pace. In theory the enemy that does not have this system won’t be able to respond fast enough.
A cleaner comparison might be hedge funds trading in the 2000s with market maker options trading today. High frequency/deep quant trading created opportunities that didnt exist 20 years ago and new market makers as a result.
Elon Musk kind of killed it with Starlink, because Skynet just seems like a worse version of that. I think Maven has that 'Karen energy' built into it, I could see it destroy us. I'm #TeamMaven
