DARPA Air Combat Evolution (ACE) Program

Seeing as he didn't even touch the brakes, I'm going to say yes.

vtti.vt.edu/featured/?p=422

Caveats:
>ALL driverless crashes must be reported
>not all human driver crashes reported (ie., fender benders), with estimates being up to 50% of minor bumps being unreported, or police determining they arent enough to warrant paperwork

>level 1 crash: car rolls over, a fatality, etc, a high delta V impact
>level 2: $1500+ damage. Hitting a sign, most medium animal (deer) impacts
>level 3: small animal impacts, going over curb, hitting a guard rail at low speed.
>level 4: tire-strike only (clipping a curb)

As seen in pic, theres a significant deference, though this is 2016 data taken from google vehicles.

Attached: figure1.png (800x474, 40K)

Attached: Screenshot_20190625-083917_Amazon Kindle.jpg (1080x1920, 154K)

>US planes using Belkan technology

what the fuck, is this a crossover episode

Attached: B E L K A.png (1601x1088, 9K)

based and acecombatpilled.

thread theme: youtube.com/watch?v=mjY6awCr2_c

LMAO. AI isn’t even 1/100th as good as a human. Those numbers are fake and don’t account for the times humans have had to intervene to prevent crashes in Autopilot aka disengagement. And the majority of those miles are highway miles, the easiest to solve.

Are you that fucking guillable?

almost all planes today are running on autopilot, right now, user. the air is a much easier place to have AI than the ground.

Attached: 1561351292937.jpg (2724x2744, 818K)

Commercial airline autopilot has nothing to do with AI.

>the-the AI numbers are fake!
>meanwhile, the study states that probably up to 50% of human crashes and curb incidents go unreported

How often do humans have to dodge other humans? More so? AI dont text and drive. AI dont get DUIs. AI dont get distracted by fiddling with the radio, or talking to someone, or eating in the car. AI dont unnecessarily speed. AI back in this 2016 study analyzing largely 2014-2015 data suggest a significant improvement on driving. By 2020, 2025 it will be the clearly superior driver outside going off road entirely. And it wont be long until insurance charges you extra for a manual car, rather than one with a default automated car.

>AI don’t
You sound like a pajeet.

AI is way overhyped for autonomous operations. It’s good as an adjunct to a human but fails terribly with the simplest human tasks.

Not him, but have you been living under a rock? AI has seen enormous advances even since last year, let alone 2016. They’re more than capable of handling air operations and it won’t be long before air-to-air roles are mostly occupied by drones like in OP

AI dont is viable depending depending if you are referring to plural objects, a singular entity, or a group referred to as a singular entity. Driving is a monotonous task with set rules, thats the sorta thing AI is good at.

user, AI is not magic. Airliner autopilot absolutely counts as "AI" because "AI" is a massive misnomer meme term to begin with. AI is nothing more than the use of algorithms or collections of weighted algorithms to make control decisions.

Literal brain let. AI in contemporary computer science refers to DNNs or deep neural networks. It has nothing to do with airline autopilot which is hard programmed human parameters.

If you’re 90 IQ don’t respond to this thread. Stick to your guns bullshit.

There’s no evidence that AI has improved that much. This is a low IQ belief. The problem with AI is that it works 99.9% of the time but there’s a .1% or less where the generalization of the algorithm utterly fails. Those edges cases of failure have not been improved upon and still happen frequently.

Dnns aren’t the only form of AI, user....

>Literal brain let. AI in contemporary computer science refers to DNNs or deep neural networks.
Physician, heal thyself.

99.9 is still better than any human, and you’re forgetting the multiple advantages AI has inherently over humans. Faster response times, better reflexes, etc. and you’re simply ignorant or failing to read up on this stuff if you haven’t seen stuff like ALPHA from Psibernetix crush human and AI pilots alike in simulated air combat

>better than humans

So buy a Tesla let it drive you around end never take over a disengagement if you believe that.

Obviously you won’t because it’s an utter lie.

>ALPHA from Psibernetix
That was really, really over-hyped by the media, probably intentionally from the company. The pilot that was beaten was also an employee.

Except people do this all the time...
Doesn’t discount the fact that he is an experienced pilot which is more than enough to prove my point. Also citation needed on whether he’s an employee of theirs

>DNNs or deep neural networks
Why did you say it twice?

>So buy a Tesla let it drive you around end never take over a disengagement if you believe that.
Theres a video on pornhub of two people fucking in a Tesla while doing this. Even Elon left a comment.

Elon Musk made a pornhub account in order to comment on a video of two people fucking in one of his cars? Hard to believe.

The last fighter pilot has already been born

>Deep learning exists
>Suddenly all other forms of AI are no longer considered AI

You might like to know that DNNs also have human hardcoded parameters in many cases

I’ve read the ALPHA paper for my work. While I’m not an expert in air combat they specifically mention trying to even the playing field on both sides. The human pilots had superior aircraft, short range missiles, and long range range missiles. The AI only had medium range missiles but generally (not always) had a 2 to 1 numerical advantage. From what I’ve read on air combat losing sight of your opponent means death, and it’s likely much harder keeping track of two opponents. Though maybe the human advantages do actually make up for it, I dunno

Attached: 96C9378E-6C69-4909-9752-B21F21F2F503.gif (750x545, 126K)

I mean the video did go crazy viral and pornhub has a huge twitter presence.
Also its Elon Musk. A known shitposter. Hard to believe my ass.

There won't be fighter pilots soon. Air-to-Air combat will be completely automated, including fighter jets, bombers, and even transport planes.

What will remain of flying is simply recreational use.

Do you have a link to that document? I find stuff like this to be incredibly interesting..

Here ya go user

google.com/url?sa=t&source=web&rct=j&url=https://scholar.uc.edu/downloads/bc386t26n&ved=2ahUKEwiHptyngoXjAhVKG80KHQ9VA8wQFjABegQIAhAB&usg=AOvVaw1lAbblA2fW16u5aUHl0H8o&cshid=1561479044283

Thank you.

There's nothing intelligent about DFCS, it's nothing more than if this then that or else. All it's doing is taking voltage readings to determine flight control positions, running it through ADIRU and FMC data then keeping course as selected. It's a dumb system. T. Avionics fag

The future is now, old man

>implying he didn't already have one

Even something as simple as a search algorithm is considered AI. I’m not an avionicsfag but but the voltage readings and data sounds like it could be represented as a position in high-dimensional space and keeping course as selected sounds like finding the position in that space to maintain your current state. At least with that description, it becomes an AI problem. But I’m not an avionicsfag and don’t actually know what’s going on under the hood

>Said I’m not an avionicsfag twice

That’s what i get for not proofreading

Seriously. People that keep thinking self-driving cars are worse than human drivers haven’t been following the technology. Also, any time there’s a crash involving a self-driving vehicle, of course it’s gonna be heavily reported on. Meanwhile, there’s probably thousands of car crashes every day that never make the local news.

How many miles is the average Tesla autopilot disengagement? It's less than 100 miles probably. Tesla won't even California test it. That means without a human a Tesla would crash every 100 miles. And that's mostly cruising miles, since autopilot doesn't work at red lights or stop signs.

Jesus Christ you people are fucking retarded.

Here's the disengagement rate in the California test for self driving cars, the majority of which are one every less than 50 miles.

Attached: 4F011014-EC37-4AF7-9AB4-1EA11CB1331E.jpg (1623x1166, 396K)

I honestly don't get Elon Musk.

On the one hand he's a thief and plagiator (Hyperloop), borderline if not outright conman (Tesla) and weirdo (Boring Company) . On the other hand he's also the guy behind SpaceX which is a huge success (I think??).

He's clearly a driven genius that knows how to achieve the things he wants to do, but he's not the perfect human being his fans seem to paint him as.

Yeah that's a good way of putting it. He's obviously an outstanding businessman on the surface (to be fair though, weren't his parents multi millionaires already?) but he's clearly not a saint.

That being said, I don't think we're getting the full story. A lot of the hype or meme business people (latest example: Elizabeth Holmes) turn out to be something completely different than what they were presented to be...Just a hunch...

Attached: FB_IMG_1558568354511.jpg (480x372, 10K)

>friendly fire never happened for humans

>all the children in this thread

Attached: 1561209416638.jpg (620x1108, 63K)

Its nowhere near that complex, its literally just as i said, pilot sets a course. ADIRU provides gyroscopic and air data knowledge for the FCC. FCC receives voltage readings from sensors in flight surfaces and summnates changes based on prior measurements. FMC provides GPS data for following said course. Its not going to fly itself in an emergency situation of any kind its quite simply go from point A to point B here's your numbers. There's also a TCAS computer for avoiding other aircraft via summnation values then providing an climb/descend callout to the pilot. There is no Artificial intelligence to it. Its just math. It doesn't think, there's no learning. Not yet at least.

Air is hard for humans because 3d but easy for computers because no clutter. Vice versa for ground.

This lil' nigga got his ass handed to him.

>>AI will decide which planes to shoot
>>AI can't even avoid driving into 18 wheel semi
You got the difficulty of those two things completely backwards
Which is easier;
a) a literal multiple choice question
b) spatial recognition, land nav and collision avoidance
One takes three million lines of code.
The other takes six.
AI in the air is fucking easy.

People drive like assholes and tailgate semi's all the time

Robots are getting scary

>muh automation
>muh AI
whatever

I doubt they'd employ a complete slouch to fly against ALPHA when the point is to put it's problem solving capabilities to the test. Besides it demonstrated other impressive abilities like pulling off high G maneuvers to throw off enemy missiles and extreme accuracy in missile guidance, enough to put it's missile through the cockpit of simulated enemy planes and into the head of an enemy pilot. At the very lead it might be beneficial to get a stripped down version of ALPHA as part of your targeting system to allow for superhuman guidance precision.

What happens when we lose the keys to the machines?

Lyft is doing massive testing with autonomous vehicles.

oh I forgot that a fucking PLANE does not need to worry about navigation or any of the other things you listed

retard

It's hard to have discussions about AI because everyone seems to have a different idea of what AI is. Some of the simplest forms of AI have no learning capabilities. Our most sophisticated form of AI today is still essentially just math. If thinking is a requirement in your definition of AI, then it's likely that AI has not even been created yet. It's may not be possible for the two us to reach a consensus at this rate

Attached: 1523114478199.jpg (1200x859, 967K)

>I doubt they'd employ a complete slouch to fly against ALPHA
The guy they had doing tests was an air battle manager/AWACS. Not to say that position isn't incredibly important, but that isn't really a fighter pilot.
>when the point is to put it's problem solving capabilities to the test
The point also could've been to get name recognition to the public for their company.
>Besides it demonstrated other impressive abilities like pulling off high G maneuvers to throw off enemy missiles and extreme accuracy in missile guidance, enough to put it's missile through the cockpit of simulated enemy planes and into the head of an enemy pilot
Source? Because nothing I've seen stated that ALPHA was doing that, it was more oriented towards ordering other planes around.

So long and thanks for all the fish!

Attached: armor of the softness.jpg (480x360, 18K)

Compared to cars, it doesn't. AI is not like humans. What's easy for humans is not easy for AI. What's easy for AI is not easy for humans.

Might have been a while since I read through the PDF, I might be referring to the iteration actually "flying" the simulated planes which ALPHA was giving instructions to when discussing specific combat capabilities.

I think the difference here is that Tesla and Space-X are delivering actual products and services instead of empty hype backed by bald faced lies like Theranos.