Page 2 of 6

Re: No more F-22s for you!

Posted: Wed Jul 22, 2009 3:57 pm
by Isgrimnur
While these aircraft are indeed nifty, they are all tasked with a ground role. I have yet to hear of a UAV that would be capable of supersonic flight, which would be key to an effective unmanned air superiority platform.

If anything, I see these craft as eventual replacements for the F-35 rather than the F-22.

Re: No more F-22s for you!

Posted: Wed Jul 22, 2009 4:18 pm
by Quaro
I think Gates makes the case quite well:
The grim reality is that with regard to the budget we have entered a zero-sum game. Every defense dollar diverted to fund excess or unneeded capacity… is a dollar that will be unavailable to take care of our people, to win the wars we are in, to deter potential adversaries, and to improve capabilities in areas where America is underinvested and potentially vulnerable. That is a risk that I will not take and one that I cannot accept.
Simply put, the F-22 it hurts the troops because it's money that could be going to areas where we are weaker.

Re: No more F-22s for you!

Posted: Wed Jul 22, 2009 4:31 pm
by Isgrimnur
My question about that is: how the budget is parsed to account for that money? Does that money saved go back into the Air Force's coffers, or does it go back into the general fund? Since this was a Congressional line item, I don't believe that this is money that was in a general Pentagon fund. Maybe they will be more willing to approve that amount for another project, but I don't see how this results in more body/Humvee armor as a matter of this being rolled into a product that was under funded.

Re: No more F-22s for you!

Posted: Wed Jul 22, 2009 5:16 pm
by Combustible Lemur
Isgrimnur wrote:While these aircraft are indeed nifty, they are all tasked with a ground role. I have yet to hear of a UAV that would be capable of supersonic flight, which would be key to an effective unmanned air superiority platform.

If anything, I see these craft as eventual replacements for the F-35 rather than the F-22.

These particular aircraft perhaps. But it's disingenuous to say that there is no UCAV technology in development even close for air superiority. UCAV's in the past 9 years has gone from from RC prop driven unarmed predator drones, to jet driven team tactic behaving independantly acting stealth fighters.Technology is fast, we have nearly 200 F-22's, we need to replace F16's A10's and all the other 40 year old technology, and Every major player is gearing up to go UCAV for most air operations over the next 20 - 50 years.

I don't think anyone is debating that the F-22 isn't the baddest monster air superiority fighter on the block. Or that in a war against a foe with an equal level fighter that has huge numbers of aircraft that we had previously no knowledge of them having, we would prefer to have lots more than 187 F22's assuming the f35 isn't improved to F22 quality by that time. The military is just saying that in the current and forseeable future the Bismark is superfluous to the needs of the services compared to the immediate gain of F-35's that are still decades ahead of fourth generation fighters..

Re: No more F-22s for you!

Posted: Wed Jul 22, 2009 5:32 pm
by The Preacher
I'm not sure what to make of the F-22 demise. What I will say is that I'll believe the UAV's will own the air when it happens. Most military strategists said the same about missiles 40 years ago and yet...

Personally, I think UAV's will be a great supplemental strategy but nothing beats "eyes in the sky."

Re: No more F-22s for you!

Posted: Wed Jul 22, 2009 5:40 pm
by Isgrimnur
Combustible Lemur wrote:These particular aircraft perhaps. But it's disingenuous to say that there is no UCAV technology in development even close for air superiority. UCAV's in the past 9 years has gone from from RC prop driven unarmed predator drones, to jet driven team tactic behaving independantly acting stealth fighters. Technology is fast, we have nearly 200 F-22's, we need to replace F16's A10's and all the other 40 year old technology, and Every major player is gearing up to go UCAV for most air operations over the next 20 - 50 years.
Independently acting stealth fighters? No. Independently acting stealth bombers, sure.

I'm not trying to be disingenuous. I have yet to see anything that will fight for air cover. The bomber will not always get through. Most of these projects won't see operational status before 2015. But in terms of aircraft performance, UAVs stand at a pre-X-1 level.

Technology may be fast, but if we cut our margins too close in terms of the numbers of aircraft available until the tech catches up, we're going to have difficulties fielding enough operational and training programs to keep the F-22 a viable weapon until such time as they are replaced. We will lose aircraft to accidents. The Air Force is planning on keeping about as many F-15s in service until 2025. It's not much of a replacement if you have to keep the old versions around.

The problem becomes that if you de-fund the program now, it's never going to get re-funded at a later time. The needs of the services will change. Honestly, I hope you're right and that we'll have plenty and the most action they see is at airshows.

Re: No more F-22s for you!

Posted: Mon Feb 11, 2013 1:29 pm
by Isgrimnur
F-35 downgraded
For the second time in a year, the Pentagon has eased the performance requirements of the F-35 Joint Strike Fighter (JSF). The reduced specs — including a slower acceleration and turning rate — lower the bar for the troubled, trillion-dollar JSF program, allowing it to proceed towards full-rate production despite ongoing problems with the plane’s complex design. Under the old specs, the stealth fighter, due to enter service in 2018 or 2019, probably wouldn’t pass its Pentagon-mandated final exams.
...
For the pilots who will eventually take the F-35 into combat, the JSF’s reduced performance mean they might not be able to outfly and outfight the latest Russian- and Chinese-made fighters. Even before the downgrades, some analysts questioned the F-35′s ability to defeat newer Sukhoi and Shenyang jets. Despite the JSF’s lower specs, Lockheed bizarrely claims its new plane is now more maneuverable than every other fighters in the world except the company’s own F-22.
...
“The program announced an intention to change performance specifications for the F-35A, reducing turn performance from 5.3 to 4.6 sustained g’s and extending the time for acceleration from 0.8 Mach to 1.2 Mach by eight seconds,” DOT&E stated. The F-35B and C also had their turn rates and acceleration time eased. The B-model jet’s max turn went from 5.0 to 4.5 g’s and its acceleration time to Mach 1.2 was extended by 16 seconds. The F-35C’s lost .1 g off its turn spec and added a whopping 43 seconds to its acceleration.

The changes likely reflect higher-than-expected drag on the JSF’s single-engine airframe, according to Bill Sweetman of Aviation Week. The implications for frontline pilots are pretty serious. Less maneuverability makes the F-35 more vulnerable in a dogfight. And the slower acceleration means the plane can spend less time at top speed. “A long, full-power transonic acceleration burns a lot of fuel,” Sweetman explained.

This is not the first time the Pentagon has altered its standards to give the JSF a pass. In early 2012, the military granted the F-35 a longer takeoff run than originally required and tweaked the plane’s standard flight profile in order to claw back some of the flying range lost to increasing weight and drag.
...
In any event, the F-35 is likely to get even less maneuverable as development continues. Gilmore’s report warned that the F-35A’s tightly-packed airframe has essentially zero room for weight growth without losing nimbleness. “The program will need to continue rigorous weight management through the end of [development] to avoid performance degradation and operational impacts.”

But in the same report, the Pentagon admitted to a chain of safety problems that could force Lockheed to add weight to the radar-evading plane. Extra mass doesn’t necessarily affect the JSF’s ability to avoid detection, but it does impact maneuverability. Several years ago, to save around 50 pounds, the F-35′s designers removed some fuel safety valves. As a result, the JSF is now 25-percent more likely to burn if struck by enemy weapons, making it “overall more vulnerable [to fire] than most” older warplanes, Jennifer Elzea, a DOT&E spokesperson, told Bloomberg.
:doh:

Not to mention that the VTOL version loses to John McClane in a close quarters battle situation.

Re: No more F-22s for you!

Posted: Tue Feb 12, 2013 3:51 pm
by NickAragua
RE: Widespread use of AI controlled craft

Completely unacceptable, due to the following factors:
1. Software is inherently incredibly buggy.
2. Software written by government or on government contract is doubly likely to be so - after all, it's written by the lowest bidder.
3. Getting software to acceptable reliability costs a lot of time and money, which is excluded by the fact that we're talking about "lowest bidder" software.

I do software development for a living and spent my grad school time doing AI research, and I wouldn't trust a computer to drive my car. The error rate in common situations is way too high (1%? .5%?), and I don't even want to think about emergency situations. Given that, I certainly wouldn't want a fully autonomous AI in charge of an aircraft firing missiles and such. There needs to be a human in the loop - partially so you can have someone to blame if you shoot a missile at a wedding party (for example), but partially because automated systems just aren't good enough - you still need a human to babysit the machine so it doesn't mistake a school for a military training camp.

It may be that, in the future, people will be comfortable handing out death dealing capabilities to fully autonomous machines. I sincerely hope I'm not around to see it.

Re: No more F-22s for you!

Posted: Tue Feb 12, 2013 4:57 pm
by GreenGoo
NickAragua wrote: 2. Software written by government or on government contract is doubly likely to be so - after all, it's written by the lowest bidder.
So are the vehicles. So is nearly every piece of military hardware in the US arsenal.

Re: No more F-22s for you!

Posted: Tue Feb 12, 2013 5:12 pm
by LawBeefaroni
GreenGoo wrote:
NickAragua wrote: 2. Software written by government or on government contract is doubly likely to be so - after all, it's written by the lowest bidder.
So are the vehicles. So is nearly every piece of military hardware in the US arsenal.
Lowest qualified bidder. Only pre-approved companies can bid and often (as with a fighter jet program) a full design competition is also part of the process. Even "minor" stuff like small arms ammo is limited to qualified bidders.

Re: No more F-22s for you!

Posted: Tue Feb 12, 2013 11:20 pm
by GreenGoo
LawBeefaroni wrote:
GreenGoo wrote:
NickAragua wrote: 2. Software written by government or on government contract is doubly likely to be so - after all, it's written by the lowest bidder.
So are the vehicles. So is nearly every piece of military hardware in the US arsenal.
Lowest qualified bidder. Only pre-approved companies can bid and often (as with a fighter jet program) a full design competition is also part of the process. Even "minor" stuff like small arms ammo is limited to qualified bidders.
Sure. Do you think ai for an autonomous vehicle would be less stringent or more stringent than for the vehicle itself?

Of course it's moot since it would be part of the whole package anyway. A vehicle intended to be autonomous would be designed from the ground up to that purpose. It's not like they'd throw ai onto an F-14 and send it off into combat.

There is no reason to believe that (any) military hardware would go through a satisfactory bidding process but the software would somehow be inherently (more) faulty for going through the same process.

Re: No more F-22s for you!

Posted: Wed Feb 13, 2013 12:36 am
by NickAragua
GreenGoo wrote:Sure. Do you think ai for an autonomous vehicle would be less stringent or more stringent than for the vehicle itself?

Of course it's moot since it would be part of the whole package anyway. A vehicle intended to be autonomous would be designed from the ground up to that purpose. It's not like they'd throw ai onto an F-14 and send it off into combat.

There is no reason to believe that (any) military hardware would go through a satisfactory bidding process but the software would somehow be inherently (more) faulty for going through the same process.
We'll leave my opinion of the "satisfactory bidding process" aside for this discussion (as it's based only on anecdotal evidence and personal experience). Suffice it to say that I don't believe a large, bureaucratic organization capable of producing a high-quality complex software product in the general case.

The main point I was trying to make (and I could have emphasized it more) is that, even assuming well-written software, we're talking about an incredibly complex chunk of code that has to solve a bunch of pretty hard (generally NP-complete) problems to a very high degree of accuracy in real-time(!). A discussion of the specifics is probably out of scope for this thread, but I'd be glad to get into it somewhere else. As a result of this complexity, it doesn't matter who's writing our hunter-killer AI, you're going to get a steaming pile of poop unless you sink a lot of resources in there. Now, when Watson doesn't answer three out of sixty questions (or however many get asked in two jeopardy rounds) correctly, or when Deep Blue whiffs in four out of six matches vs Gary Kasparov, it's no big deal. When, out of sixty autonomous HKs, three shoot a bunch of civilians or friendlies, well, I hope you guys are happy that there was a "satisfactory bidding process".

Re: No more F-22s for you!

Posted: Wed Feb 13, 2013 1:18 am
by Rip
Guess what, those nuclear ICBMs are pretty complex beasts and lobbing several independent warheads from each one each hitting a different target with accuracy is pretty complex as well. Not to mention the systems that secure the communications or the systems that detect and opposition launch.

Don't even get me started on the complexity of the NSAs intel intercept systems. Suffice it to say they can handle complex design and more importantly testing to determine if it works or not. I think you could even fit the entire software system that performed calculations for the moon launch into something that would easily fit into a car.

Re: No more F-22s for you!

Posted: Wed Feb 13, 2013 1:36 am
by GreenGoo
My point is that either the bidding process can produce high quality products, and I think evidence has shown at least a few winners in this area, which Lawbeef also seems to support, or the bidding process can only produce low quality products.

I grant you that ai programming for an autonomous weapon would be so incredibly complex as to be guaranteed to be flawed in some way or ways (which you stated). My interjection was that the bidding process is not inherently going to produce lower quality software, just as it did not inherently produce lower quality hardware or software in other military products.

Both Lawbeef and Rip seem to support that quality products can be had via this process.

Not trying to make a huge brawl, it's just that I hear the lowest bidder line used to automatically dismiss everything a federal government does as shit by default.

I think the complexity of making a viable ai is going to produce enough problems in and of itself, no matter how it is acquired.

Re: No more F-22s for you!

Posted: Wed Feb 13, 2013 2:32 am
by NickAragua
GreenGoo wrote:My point is that either the bidding process can produce high quality products, and I think evidence has shown at least a few winners in this area, which Lawbeef also seems to support, or the bidding process can only produce low quality products.

I grant you that ai programming for an autonomous weapon would be so incredibly complex as to be guaranteed to be flawed in some way or ways (which you stated). My interjection was that the bidding process is not inherently going to produce lower quality software, just as it did not inherently produce lower quality hardware or software in other military products.

Both Lawbeef and Rip seem to support that quality products can be had via this process.

Not trying to make a huge brawl, it's just that I hear the lowest bidder line used to automatically dismiss everything a federal government does as shit by default.

I think the complexity of making a viable ai is going to produce enough problems in and of itself, no matter how it is acquired.
I'm perfectly willing to leave product quality out of this discussion - that's a practical concern and, like you appear to be saying, doesn't change the difficulty of the problem at hand. I guess it's ultimately irrelevant.

I'm really just here to argue that, when it comes to killing people (possibly a lot of people), there always needs to be a (non-psychotic) human in the decision loop. AI systems can be good at flagging things that may be relevant in a mountain of data for human review, and they're great at plotting trajectories and flight angles or whatever, but, until we start seeing the next generation of data analysis and decision making algorithms... would you really trust the decision to kill a bunch of people to an AI that has a 52% confidence and makes that decision in 5 milliseconds? Sure, people have based such decisions on pure lunacy, historically speaking, but then we don't really need the AIs for that.

Re: No more F-22s for you!

Posted: Wed Feb 13, 2013 2:47 am
by Rip
NickAragua wrote:
GreenGoo wrote:My point is that either the bidding process can produce high quality products, and I think evidence has shown at least a few winners in this area, which Lawbeef also seems to support, or the bidding process can only produce low quality products.

I grant you that ai programming for an autonomous weapon would be so incredibly complex as to be guaranteed to be flawed in some way or ways (which you stated). My interjection was that the bidding process is not inherently going to produce lower quality software, just as it did not inherently produce lower quality hardware or software in other military products.

Both Lawbeef and Rip seem to support that quality products can be had via this process.

Not trying to make a huge brawl, it's just that I hear the lowest bidder line used to automatically dismiss everything a federal government does as shit by default.

I think the complexity of making a viable ai is going to produce enough problems in and of itself, no matter how it is acquired.
I'm perfectly willing to leave product quality out of this discussion - that's a practical concern and, like you appear to be saying, doesn't change the difficulty of the problem at hand. I guess it's ultimately irrelevant.

I'm really just here to argue that, when it comes to killing people (possibly a lot of people), there always needs to be a (non-psychotic) human in the decision loop. AI systems can be good at flagging things that may be relevant in a mountain of data for human review, and they're great at plotting trajectories and flight angles or whatever, but, until we start seeing the next generation of data analysis and decision making algorithms... would you really trust the decision to kill a bunch of people to an AI that has a 52% confidence and makes that decision in 5 milliseconds? Sure, people have based such decisions on pure lunacy, historically speaking, but then we don't really need the AIs for that.
Which is why you have that human making the decision on when to deploy them and what situation they will be appropriate for. Not like they just ship a bunch of planes to a combat zone and turn them on. Even AI systems require some strategic and operational oversight.

Re: No more F-22s for you!

Posted: Wed Feb 13, 2013 7:50 am
by RLMullen
NickAragua wrote: We'll leave my opinion of the "satisfactory bidding process" aside for this discussion (as it's based only on anecdotal evidence and personal experience). Suffice it to say that I don't believe a large, bureaucratic organization capable of producing a high-quality complex software product in the general case.
Space Shuttle control software. 135 missions, 2 failures, neither of them software related. You lose. :horse:

Re: No more F-22s for you!

Posted: Wed Feb 13, 2013 10:16 am
by stessier
RLMullen wrote:
NickAragua wrote: We'll leave my opinion of the "satisfactory bidding process" aside for this discussion (as it's based only on anecdotal evidence and personal experience). Suffice it to say that I don't believe a large, bureaucratic organization capable of producing a high-quality complex software product in the general case.
Space Shuttle control software. 135 missions, 2 failures, neither of them software related. You lose. :horse:
Also - current Mars rover was outside of human control for 7 minutes during which it did some really amazing stuff without embedding itself in a crater.

Re: No more F-22s for you!

Posted: Wed Feb 13, 2013 10:20 am
by Gavin
You'd think they'd do more in the improving existing technology department instead of regularly scrapping entire projects at every turn. One piece of software I work with is 25 years old. You wouldn't know it to look at it because it looks really new. The 25 years have been spent making it better, less buggy, and more aesthetically pleasing. The benefit? It's an easy application that simply works. Calls I get for it are almost always user error.

Imagine if we had one unified fighter OS that had been firmed up over all these years instead of a bunch of infant codes that are buggier than hell. NASA is an example of an agency that knows how to improve on what they have and recycle it. The military drops the ball there. I should ask my buddy that's a mechanic for these fighter jets what he thinks about it.

Re: No more F-22s for you!

Posted: Wed Feb 13, 2013 11:22 am
by Isgrimnur
There are F-15s in service that were made when I was a toddler. There are B-52s still in service that were made when my father was a toddler.

Military aircraft are continually being upgraded from a physical and software perspective.

Re: No more F-22s for you!

Posted: Wed Feb 13, 2013 11:28 am
by Gavin
Isgrimnur wrote:There are F-15s in service that were made when I was a toddler. There are B-52s still in service that were made when my father was a toddler.

Military aircraft are continually being upgraded from a physical and software perspective.
Maintained is not upgraded. The software also isn't translating into new craft. Every time we get a completely new OS with all the bugs to work out. There has to be a way to standardize software.

Re: No more F-22s for you!

Posted: Wed Feb 13, 2013 11:31 am
by RLMullen
Gavin wrote:
Isgrimnur wrote:There are F-15s in service that were made when I was a toddler. There are B-52s still in service that were made when my father was a toddler.

Military aircraft are continually being upgraded from a physical and software perspective.
Maintained is not upgraded. The software also isn't translating into new craft. Every time we get a completely new OS with all the bugs to work out. There has to be a way to standardize software.
The B-52 was bid in 1946 and its maiden flight was in 1952. I can assure you that aircraft has been significantly UPGRADED in the intervening years.

Re: No more F-22s for you!

Posted: Wed Feb 13, 2013 11:34 am
by Isgrimnur
B-52 Avionics
Ongoing problems with avionics systems were addressed in the Jolly Well program, completed in 1964, which improved components of the AN/ASQ-38 bombing navigational computer and the terrain computer. The MADREC (Malfunction Detection and Recording) upgrade fitted to most aircraft by 1965 could detect failures in avionics and weapons computer systems, and was essential in monitoring the Hound Dog missiles. The electronic countermeasures capability of the B-52 was expanded with Rivet Rambler (1971) and Rivet Ace (1973).

To improve safe day and night operations at low altitude, the AN/ASQ-151 Electro-Optical Viewing System (EVS), which consisted of a Low Light Level Television (LLLTV) and a Forward looking infrared (FLIR) system mounted in blisters under the noses of B-52Gs and Hs between 1972 and 1976. The navigational capabilities of the B-52 were later augmented with the addition of GPS in the 1980s. The IBM AP-101, also used on the Rockwell B-1 Lancer bomber and the Space Shuttle, was the B-52's main computer.

In 2007 the LITENING targeting pod was fitted, which increases the combat effectiveness of the aircraft during day, night and poor weather conditions in the attack of ground targets with a variety of standoff weapons, using laser guidance, a high resolution forward-looking infrared sensor (FLIR), and a CCD camera used to obtain target imagery.

Re: No more F-22s for you!

Posted: Wed Feb 13, 2013 11:37 am
by Gavin
To repeat: The software also isn't translating into new craft. Every time we get a completely new OS with all the bugs to work out. There has to be a way to standardize software.

My comments aren't that we should be using old planes and just upgrading them. My point is that with each new plane we largely scrap what made previous models successful and rewrite the whole thing manually.

Re: No more F-22s for you!

Posted: Wed Feb 13, 2013 11:41 am
by Isgrimnur
Isgrimnur wrote:The IBM AP-101, also used on the Rockwell B-1 Lancer bomber and the Space Shuttle, was the B-52's main computer.
I'm certain that they wouldn't have completely scrapped the software. And I highly doubt that Lockheed and Northrop are completely scrapping the OS for new versions. Also, I know for a fact that there are export restrictions in place so that our F-16 export customers aren't getting the avionics suites that we use, so even though they get the airframe, our software is better than what we give them.

Re: No more F-22s for you!

Posted: Wed Feb 13, 2013 11:47 am
by Gavin
Isgrimnur wrote:
Isgrimnur wrote:The IBM AP-101, also used on the Rockwell B-1 Lancer bomber and the Space Shuttle, was the B-52's main computer.
I'm certain that they wouldn't have completely scrapped the software. And I highly doubt that Lockheed and Northrop are completely scrapping the OS for new versions. Also, I know for a fact that there are export restrictions in place so that our F-16 export customers aren't getting the avionics suites that we use, so even though they get the airframe, our software is better than what we give them.
Ok? We've had a lot of planes over the years. Do you think you could find 10 examples of significant rebuilding of previously successful tech if you looked? I'm not asking you to wade through and find them. I'm just asking if you feel that the example you gave here is indicative of the norm?

Re: No more F-22s for you!

Posted: Wed Feb 13, 2013 11:55 am
by Isgrimnur
I absolutely feel that it is indicative of the norm. The A-10 interior lighting was reworked to allow the pilots to use NVGs, something that had been pioneered in other aircraft. Weapon systems are developed and tweaked to allow existing platforms to deliver newer and more effective ordinance.

The lifecycle of military equipment is one of continuous upgrades, not least of which because the contractors want to earn more money by improving their existing equipment and capabilities.

Re: No more F-22s for you!

Posted: Wed Feb 13, 2013 12:00 pm
by NickAragua
Successful NASA software.
Well, you guys got me. Have fun getting blown up by AI controlled aircraft that mistakenly classify your house as a terrorist training camp.

Re: No more F-22s for you!

Posted: Wed Feb 13, 2013 12:18 pm
by Rip
NickAragua wrote:
Successful NASA software.
Well, you guys got me. Have fun getting blown up by AI controlled aircraft that mistakenly classify your house as a terrorist training camp.
Does it matter if it is an IAI controlled craft versus a human controlled drone? Chances are about the same of mistakenly classifying you as a target. In fact I would prefer the AI, at least it should be consistent.

Re: No more F-22s for you!

Posted: Wed Feb 13, 2013 12:51 pm
by RLMullen
NickAragua wrote:
Successful NASA software.
Well, you guys got me. Have fun getting blown up by AI controlled aircraft that mistakenly classify your house as a terrorist training camp.
You set yourself up as a self-styled expert in AI software because you once did a grad school project in AI. You then make the broad-based claim that no good software can come from a large bureaucratic bidding process. Now you are surprised that you got snarky replies that completely discount your broad-based claim regarding "government software"?

I'll take this one step further...
NickAragua wrote: would you really trust the decision to kill a bunch of people to an AI that has a 52% confidence and makes that decision in 5 milliseconds?
First of all, there is no decision in your example. You've simply given an analytic confidence rating and processing time. Am I to expect that your fictitious software takes a kill shot with a 52% confidence rating? Why would your software do this? Did you hardcode 52 as a magic number? Is 52% confidence simply "good enough" based on the current engagement parameters? I think that you were maybe pulling numbers out of your ass in order to trigger an emotional response as your talking point. I've certainly been guilty of this tactic, and it didn't work for me anymore than it is now working for you.

Also, why would trusting a kill decision to a emotionless software routine be any worse than trusting a kill decision to an emotional, albeit trained, human? I think the LAPD just proved that a trained human is still capable of pumping several magazines of bullets into a civilian vehicle with a confidence rating somewhere damned near 0%. The citizens of LA may indeed prefer Robocop!

As far as AI driving cars... that is coming, and sooner than any of us expect. My prediction is that AI driven cars will have a much better safety record than their human piloted counterparts. I'll also note that the chance that the last commercial flight that you took was landed by the aircraft's AI is somewhere above zero. AI has been an active participant in aviation, both military and commercial, for many years now.

For someone who has expressed some authority in the field of AI, you might want to get your nose out of your monitor and take a look at the world around you. The technology is moving much quicker than you think.

Re: No more F-22s for you!

Posted: Wed Feb 13, 2013 1:56 pm
by Gavin
Isgrimnur wrote:I absolutely feel that it is indicative of the norm. The A-10 interior lighting was reworked to allow the pilots to use NVGs, something that had been pioneered in other aircraft. Weapon systems are developed and tweaked to allow existing platforms to deliver newer and more effective ordinance.

The lifecycle of military equipment is one of continuous upgrades, not least of which because the contractors want to earn more money by improving their existing equipment and capabilities.
I mean, are things from previous planes utilized in newer planes? Software, hull design, etc.? Not are the individual planes upgraded. I'm talking long term improvements being standardized across planes instead of not crossing over where they may be beneficial.

Re: No more F-22s for you!

Posted: Wed Feb 13, 2013 2:16 pm
by Isgrimnur
The contractors are all about lessons learned and applying those techniques to upgrades to existing aircraft and applying them to their new aircraft. Obviously, a lot of the details are going to be quite esoteric, if not outright classified.

Thrust vectoring is a good example. It stared in development of rockets and missiles and is now part and parcel of the F-22 and F-35 designs to increase maneuverability.

Re: No more F-22s for you!

Posted: Wed Feb 13, 2013 5:19 pm
by NickAragua
RLMullen wrote:
NickAragua wrote:
Successful NASA software.
Well, you guys got me. Have fun getting blown up by AI controlled aircraft that mistakenly classify your house as a terrorist training camp.
You set yourself up as a self-styled expert in AI software because you once did a grad school project in AI. You then make the broad-based claim that no good software can come from a large bureaucratic bidding process. Now you are surprised that you got snarky replies that completely discount your broad-based claim regarding "government software"?
Who said anything about being surprised? Snark is kind of the way things work here. Nice ad-hominem attack by the way. Can we skip this "debating 101" bullshit?

I made a blanket statement that's not true in many cases - as you mentioned earlier, NASA was able to produce software that worked pretty well. Luckily, I had a slight qualifier, important part bolded for you reading pleasure.
Suffice it to say that I don't believe a large, bureaucratic organization capable of producing a high-quality complex software product in the general case.
Obviously, there are going to be exceptions, and, if you sink enough resources into a problem, you'll be able to solve it. With that in mind, I'll retract my claim of "poor quality software produced by government bidding process". I would urge you to have a healthy dose of skepticism, however, because, for every Curiosity, there's a Beagle 2.
RLMullen wrote:
NickAragua wrote: would you really trust the decision to kill a bunch of people to an AI that has a 52% confidence and makes that decision in 5 milliseconds?
First of all, there is no decision in your example. You've simply given an analytic confidence rating and processing time. Am I to expect that your fictitious software takes a kill shot with a 52% confidence rating? Why would your software do this? Did you hardcode 52 as a magic number? Is 52% confidence simply "good enough" based on the current engagement parameters? I think that you were maybe pulling numbers out of your ass in order to trigger an emotional response as your talking point. I've certainly been guilty of this tactic, and it didn't work for me anymore than it is now working for you.

Also, why would trusting a kill decision to a emotionless software routine be any worse than trusting a kill decision to an emotional, albeit trained, human? I think the LAPD just proved that a trained human is still capable of pumping several magazines of bullets into a civilian vehicle with a confidence rating somewhere damned near 0%. The citizens of LA may indeed prefer Robocop!

As far as AI driving cars... that is coming, and sooner than any of us expect. My prediction is that AI driven cars will have a much better safety record than their human piloted counterparts. I'll also note that the chance that the last commercial flight that you took was landed by the aircraft's AI is somewhere above zero. AI has been an active participant in aviation, both military and commercial, for many years now.

For someone who has expressed some authority in the field of AI, you might want to get your nose out of your monitor and take a look at the world around you. The technology is moving much quicker than you think.
You're right, 52% was pulled directly out of my rectum. I was fresh from reading an article about Watson, the Jeopardy computer, which gave Jeopardy answers with confidence scores such as "32%", to which "52%" seemed vaguely similar.

I'm not sure where you got the idea that I'm opposed to any automation whatsoever. ILS landing systems are a superb piece of technology, but there always needs to be a pilot there to make sure our glorious AI overlord doesn't fly our commercial aircraft through another plane. I'm (once again) arguing that, given the current state of the technology, there still needs to be a human in the decision loop.

Anyway, there is an implicit decision in my example. The decision is: do I fire the missile or do I not.

But let's discuss for a second what kind of problems we may expect our HK drones to solve. Here's a sampling:
- Route to target
- Threat detection (e.g. storm clouds, enemy fighters, incoming missiles, friendlies in the flight path)
- Threat mitigation (evasion/offensive action/whatever)
- Target/Non-target identification (one of the problematic parts)

The first two are pretty much solved problems. The third one is an adversarial search problem (you have a set of incoming bad things, and a set of actions you can take. Find a 'good enough' solution.) I hear the Iron Dome folks have that problem partially solved, with a 90% interception rate on incoming missiles, rockets and potentially aircraft. So, problem solved, we just strap an Iron Dome unit onto our drone and we're good to go. Either that, or we just ignore the problem because we have a thousand of these HK drones and don't care if one or two get blown up.

The fourth one, well, let's say we need to blow up a car belonging to a guy who makes bombs. We have pictures of the car, so we can pick it out from 20000 feet or whatever. Simple so far. The decision: do we shoot at the car or do we not. Now, it's a common car, and it's driving around in a populated city, and there's an occasional cloud getting in the way and we maybe don't want to kill civilians. So now we're tracking more than one potential target car and have to hold our fire until it's in a clear area - so now, not only do we have to identify what to fire at, but what *not* to fire at, and we can attempt to predict the path of our target car under cover, but only with a certain confidence level. Also, we will want to probably make sure that the target is actually in the car, and it's not just his cousin going out for a coffee, using his car. All of these are non-trivial problems and must be solved to some acceptable degree of confidence and in real time. I haven't really read about the technology to accomplish that yet.

So, what's an acceptable degree of confidence on a kill shot for you in that kind of situation? Are you happy with a Watson 32% on a jeopardy question? Maybe you want a Deep Blue 50% on a chess match? I accept the argument that humans make mistakes all the time, but we can only make so many of them in a minute. If you give up complete control to a machine, you're losing intuition (the ability to skip many, many logical steps), the ability to adapt to a completely changed situation (what if our drone suddenly needs to get re-tasked to do something completely different that it can still accomplish with its onboard equipment?) and the ability to say, "hey wait a minute, that's not a brown sedan, it's a blue pickup truck, don't fire the missile!".

Of course, all of these problems can be solved given enough time and resources. The question is, should we be spending all this time and money building autonomous HK drones? Or maybe something a little more useful. I mean, Terminator was a great movie but do you really want to live in that kind of setting?

As for cars, well, you can have your auto-drive at 55 miles an hour in the center lane. I'll keep my stick-shift, thanks.

Re: No more F-22s for you!

Posted: Wed Feb 13, 2013 6:10 pm
by RLMullen
NickAragua wrote: Who said anything about being surprised? Snark is kind of the way things work here. Nice ad-hominem attack by the way. Can we skip this "debating 101" bullshit?
Nope... not going to happen. When you make broad unsubstantiated claims, expect to be called on it. It happens to me all the time.
NickAragua wrote: The fourth one, well, let's say we need to blow up a car belonging to a guy who makes bombs. We have pictures of the car, so we can pick it out from 20000 feet or whatever. Simple so far. The decision: do we shoot at the car or do we not. Now, it's a common car, and it's driving around in a populated city, and there's an occasional cloud getting in the way and we maybe don't want to kill civilians. So now we're tracking more than one potential target car and have to hold our fire until it's in a clear area - so now, not only do we have to identify what to fire at, but what *not* to fire at, and we can attempt to predict the path of our target car under cover, but only with a certain confidence level. Also, we will want to probably make sure that the target is actually in the car, and it's not just his cousin going out for a coffee, using his car. All of these are non-trivial problems and must be solved to some acceptable degree of confidence and in real time. I haven't really read about the technology to accomplish that yet.
http://www.youtube.com/watch?v=QGxNyaXfJsA
Take a look at that video. Who do you want tracking your target car through traffic, a human operator or "the code"? Who do you think is going to be able to compute a better firing solution within the engagement parameters, a human operator or "the code"?

Your fourth problem is currently solvable, and I expect that if DARPA is letting NOVA produce that video, then they have already solved it. It's just classified.
NickAragua wrote:So, what's an acceptable degree of confidence on a kill shot for you in that kind of situation?
Degree of confidence depends on the strategic, operational, and tactical situation; the same as it is with humans. I'm not sure why you are creating a fictional situation where the HK drone is working free from a layered command and control structure? We don't put pilots in aircraft with instructions to just go shoot shit, I'm not sure why we would be doing that with HK drones.
NickAragua wrote:Of course, all of these problems can be solved given enough time and resources. The question is, should we be spending all this time and money building autonomous HK drones? Or maybe something a little more useful. I mean, Terminator was a great movie but do you really want to live in that kind of setting?
All of these problems will be solved regardless of where the resources originate. Either the US military solves them or some other military will. The scary part is that commercial institutions have a good chance of solving some of these problems before military institutions. Note that the ARGUS-IS uses commercial camera chips!
NickAragua wrote:As for cars, well, you can have your auto-drive at 55 miles an hour in the center lane. I'll keep my stick-shift, thanks.
Is your car a Red Barchetta?

Re: No more F-22s for you!

Posted: Thu Feb 14, 2013 12:29 am
by NickAragua
RLMullen wrote:http://www.youtube.com/watch?v=QGxNyaXfJsA
Take a look at that video. Who do you want tracking your target car through traffic, a human operator or "the code"? Who do you think is going to be able to compute a better firing solution within the engagement parameters, a human operator or "the code"?

Your fourth problem is currently solvable, and I expect that if DARPA is letting NOVA produce that video, then they have already solved it. It's just classified.
That's an impressive surveillance system. I noticed that, as that machine is tracking some moving cars, I can see one move behind a tree (the red jeep with the green box ~1:44), and then the tracking box disappears (and doesn't reappear after it comes out from behind the tree!). So, probably, the motion tracking software involved doesn't have a concept of a "car" or a "person", just "this is a moving thing I should flag it". It's actually a pretty common approach to motion tracking software, you take a comparison of the current frame and the previous X frames, and flag any differences. Actually identifying the moving object as a "car" or a "person" requires considerably more computational effort. So I don't think it quite solves the problem I was positing. It is, however, an incredibly useful tool to help a *human* with vehicle tracking. Instead of having to eyeball fifteen square miles, you can now have the computer flag potentially important stuff for human review, thus making the information gathering process much more efficient, but still not fully automated.
RLMullen wrote:
NickAragua wrote:So, what's an acceptable degree of confidence on a kill shot for you in that kind of situation?
Degree of confidence depends on the strategic, operational, and tactical situation; the same as it is with humans. I'm not sure why you are creating a fictional situation where the HK drone is working free from a layered command and control structure? We don't put pilots in aircraft with instructions to just go shoot shit, I'm not sure why we would be doing that with HK drones.
So, in that case, we seem to be in agreement. There still need to be people in the decision making process.
RLMullen wrote:
NickAragua wrote:As for cars, well, you can have your auto-drive at 55 miles an hour in the center lane. I'll keep my stick-shift, thanks.
Is your car a Red Barchetta?
Nope, just a shitbox Toyota. It's not quite a race car, but it's my car, and I'd like to drive it myself.

Re: No more F-22s for you!

Posted: Thu Feb 14, 2013 1:37 pm
by Boudreaux
Just as an aside, the whole "government contract to the lowest bidder" criticism isn't really accurate. Price is a factor in any DoD procurement process, but it's not the only criteria. Technical performance, requirement satisfaction, management, and company history all play into procurement/contract award decisions. Price is only a part of it, and going in with the lowest cost isn't a guarantee of success.

Re: No more F-22s for you!

Posted: Thu Feb 14, 2013 3:17 pm
by Gavin
Boudreaux wrote:Just as an aside, the whole "government contract to the lowest bidder" criticism isn't really accurate. Price is a factor in any DoD procurement process, but it's not the only criteria. Technical performance, requirement satisfaction, management, and company history all play into procurement/contract award decisions. Price is only a part of it, and going in with the lowest cost isn't a guarantee of success.
Let's also not forget that contractors aren't above greasing palms.

Re: No more F-22s for you!

Posted: Thu Feb 14, 2013 5:19 pm
by Enough
F-35, the most expensive weapons program in human history, an interesting article timely to this thread that just came out.
If sequestration happens March 1, F-35 officials have made it clear they will be forced to slow production and delay flight tests. Both steps will make each plane that is ultimately bought more expensive.

But thanks to $4.8 billion in Pentagon contracts for 31 planes pushed out the door barely 100 hours before the original Jan. 2 sequestration deadline, much of the program will continue on autopilot.

"The F-35 program has built up a good buffer by getting the most recent lot of aircraft awarded in time," says Todd Harrison, a defense-budget expert at the independent Center for Strategic and Budgetary Assessments. "That means Lockheed and all the subcontractors have a backlog of work that won't be affected by sequestration, so they can continue working as planned for the time being."

Apparently the F-35 may end up being pretty stealthy after all.

Read more: http://www.time.com/time/magazine/artic ... z2KuWVIQIQ" target="_blank

Re: No more F-22s for you!

Posted: Sat Feb 16, 2013 5:39 pm
by Boudreaux
I read that in my print issue yesterday, interesting article. There are some really boneheaded pieces to the F-35 program that are mentioned there. Starting LRIP production before even 1% of flight testing was complete means billions are being spent to "fix" jets with newly uncovered problems. This was an issue on F-22 as well, and one of the big lessons learned from that program was not to do this - and yet it happened anyway. Weight growth has caused all sorts of compromises, performance requirement reductions, and capability deferments. Commonality, which was supposed to be a huge cost savings, is now down to something like 30%.

Unfortunately DoD doubled down on it so many times, it's almost impossible to cut now without it being an enormous embarrassment to the government.

Re: No more F-22s for you!

Posted: Sat Feb 16, 2013 5:44 pm
by GreenGoo
Boudreaux wrote:Unfortunately DoD doubled down on it so many times, it's almost impossible to cut now without it being an enormous embarrassment to the government.
Our government is making the exact same mistake. I assume it has to do with some sort of political commitment at this point. There seems to be little reason to pick a few of these up for our air force but DND and our conservative government are plowing ahead doggedly regardless.