For Questions That Don't Merit Their Own Thread

[magnemoe]

50 minutes ago, JoeSchmuckatelli said:

A purpose built guided launcher might be built smooth bore to better service the rounds - but most artillery is operated as an area effect weapon, rather than a point target, and thus cheap spin stabilized rounds are fine.  So a multi purpose gun is likely to be rifled to send the cheap stuff - meaning the expensive guided round needs to be a little more expensive to compensate. 

Offering the expensive stuff gives the GP arty a bit more flexibility - and while they're talking about increased range, the big benefits are reduced collateral damage when aiming at bad guys using civilian areas as cover 

Two other tactical benefits, first if he enemy use you as cover as in getting so close to you traditional fire support get very dangerous for you but with guided shells this get easier. 
You can also be closer to the enemy then you start shelling them so you hit much faster after you stop. 

The US navy looked into an cannon for only smart shells only at one time.  They did some calculations and found that an Excalibur style shell would get maximum range if fired almost strait up, the shell would turn toward target up in the atmosphere reducing air resistance over the ballistic trajectory. Other benefit was taking very little deck space, easy to use an very long barrel, easy recoil handling as in against the keel and very easy to reload as next projectile only has to be shifted below the gun and rammed up. 
But the idea was dropped as it could only fire smart shells, it they went for this an smoothbore would make more sense, as this would be an entire new gun I would just as well make it 8" then fire 5-6" as discarding sabot   
Had made more with one of these and an standard gun on the zumwalt destroyers. 

[magnemoe]

17 hours ago, MKI said:

There have been documented cases where the stock market "drops" nearly instantly due to some form of "trolling/fake-out", like hacking the US presidents twitter account, posting a purposely designed tweet to send the stock market diving for a spit second before recovering. But this requires access to some kind of leverage that isn't easily gained. 

If your a criminal, it is actually easier to fool fellow humans than try to fool AI. There's also a large influx of humans ready to exploit.

Fake road signs could fool real people too, so I wouldn't call it jamming, just attempts at faking/fooling.

Its worth mentioning that a well trained AI within its problem domain can perform better than a human in many cases, or to help mitigate issues with an AI system, you can give it more input than a human to help it make decisions. This is why airline pilots usually let the machine do its job at landing during low visibility weather and no one bats an eye. Its also one area where self-driving cars currently excel at. You have 2 eyes, but your car can have 25+ high def cameras, range finders, lidar and radar. Putting them all together and you get a safer system than the average driver. (yes this is right now)

Yes, but it has been some cases stock marked AI goes into panic mode just because an blip in the marked put them into the sell now mode, the problem is them many uses the same AI and their sales get the others to sell too. Stock exchanges tend to disallow automated trading then its lots of changes. 

Agree its easier to fool humans, but you can find an AI fail mode and exploit it at scale. This is mostly an issue in cases like fraud or warfare, not things like self driving cars there you don't have an intelligent enemy. 

And yes combining AI and humans is an force multiplier if they human does their work that is  

[Jacke]

On 8/4/2021 at 9:35 PM, MKI said:

Its worth mentioning that a well trained AI within its problem domain can perform better than a human in many cases, or to help mitigate issues with an AI system, you can give it more input than a human to help it make decisions. This is why airline pilots usually let the machine do its job at landing during low visibility weather and no one bats an eye. Its also one area where self-driving cars currently excel at. You have 2 eyes, but your car can have 25+ high def cameras, range finders, lidar and radar. Putting them all together and you get a safer system than the average driver. (yes this is right now)

I'm very skeptical of a truly workable self-driving car.  One thing I just thought of: where are the self-driving trains?  It's a simpler problem, which is already done on some LRT systems--or are they remote control?  But I don't think there's any sign of AI for freight or passenger trains.  Which is definitely a simpler set of use cases to handle.

It's not the numbers of cameras.  It's the visual processing and control system.  The brain and body are purpose-built to handle that with the added training for operating the vehicle.  Sure, it's not perfect.  But I think the person is still likely better than the AI.

I'll believe it when it can't be spoofed by a team of engineers and magicians and passes a broad set of worse-case conditions.

[SunlitZelkova]

1 hour ago, Jacke said:

I'm very skeptical of a truly workable self-driving car.  One thing I just thought of: where are the self-driving trains?  It's a simpler problem, which is already done on some LRT systems--or are they remote control?  But I don't think there's any sign of AI for freight or passenger trains.  Which is definitely a simpler set of use cases to handle.

It's not the numbers of cameras.  It's the visual processing and control system.  The brain and body are purpose-built to handle that with the added training for operating the vehicle.  Sure, it's not perfect.  But I think the person is still likely better than the AI.

I'll believe it when it can't be spoofed by a team of engineers and magicians and passes a broad set of worse-case conditions.

https://english.kyodonews.net/news/2020/11/dec36c0fae5c-jr-east-to-test-driverless-shinkansen-bullet-trains-in-2021.html

and

https://asia.nikkei.com/Business/Companies/Driverless-train-makes-the-rounds-on-Tokyo-commuter-route

But-

https://japan-forward.com/concerns-over-japans-driverless-trains-remain/

Note that despite some safety concerns, they are still going forward with the vehicles in the first two articles.

That said, even if there was a perfect self-driving train or car, it just screams "cyber-attack target" to me personally.

Edited August 6, 2021 by SunlitZelkova

[JoeSchmuckatelli]

A bit of fluff for anyone interested 

https://www.cnn.com/interactive/2021/08/world/human-body-in-space-quiz-scn/

 

[wumpus]

On 8/5/2021 at 8:45 PM, Jacke said:

I'm very skeptical of a truly workable self-driving car.  One thing I just thought of: where are the self-driving trains?  It's a simpler problem, which is already done on some LRT systems--or are they remote control?  But I don't think there's any sign of AI for freight or passenger trains.  Which is definitely a simpler set of use cases to handle.

It's not the numbers of cameras.  It's the visual processing and control system.  The brain and body are purpose-built to handle that with the added training for operating the vehicle.  Sure, it's not perfect.  But I think the person is still likely better than the AI.

I'll believe it when it can't be spoofed by a team of engineers and magicians and passes a broad set of worse-case conditions.

A bigger issue is with self-guided trains, I'm not sure the human can do much other than get PTSD from an approaching accident.  In highly litigious countries (i.e. the USA), this makes sense as a CYA move, and can help the jurors feel empathy for the hapless engineer unable to stop the train in time.  Elsewhere, I'm unsure why you would inflict PTSD on the engineer because somebody just *had* to sit on the tracks/beat the train across a road (the train isn't going to stop regardless of whether or not the AI begins braking).

Remember that considering a single AI driven death unacceptable sacrifices 30k (-1) Americans (I'm not aware of the international numbers) to the "perfect safety" gods.  I'd like better standards comparing the expected number of accidents (and fatalities) per mile in terms of BAC (blood alcohol content) or eventually the numbers of miles (or km) an AI can drive compared to a human with equal safety.  An AI with a safety rating of ".8" or better might be vastly safer for a driver with a DUI conviction than a BAC interlock test.

Note: In the US, the NHTSA (National Highway Traffic Safety Administration) has its headquarters in Washington DC, home of some of the worst drivers in the nation.  Not sure how many of the decision makers work there compared to "regional offices" (which appear to be more any state vehicle/highway department than anything else).

PS: the cyberattack issue is much deeper and having AI drivers barely increases the attack surface.  Once I learned that Tesla cars had their brakes altered (anti-lock braking fixed) over a pushed wireless update, I just gave up on the hope of any attempted defense against cyberattacks (it doesn't help to work for a defense company that is deliberately removing security under government orders).

[kerbiloid]

They should just officially declare a saboteur everyone getting closer than 10 m 30 ft to the railway outside of official pedestrian passageways, and let AI self-defend on its own.

Edited August 8, 2021 by kerbiloid

[magnemoe]

3 hours ago, wumpus said:

A bigger issue is with self-guided trains, I'm not sure the human can do much other than get PTSD from an approaching accident.  In highly litigious countries (i.e. the USA), this makes sense as a CYA move, and can help the jurors feel empathy for the hapless engineer unable to stop the train in time.  Elsewhere, I'm unsure why you would inflict PTSD on the engineer because somebody just *had* to sit on the tracks/beat the train across a road (the train isn't going to stop regardless of whether or not the AI begins braking).

Remember that considering a single AI driven death unacceptable sacrifices 30k (-1) Americans (I'm not aware of the international numbers) to the "perfect safety" gods.  I'd like better standards comparing the expected number of accidents (and fatalities) per mile in terms of BAC (blood alcohol content) or eventually the numbers of miles (or km) an AI can drive compared to a human with equal safety.  An AI with a safety rating of ".8" or better might be vastly safer for a driver with a DUI conviction than a BAC interlock test.

Note: In the US, the NHTSA (National Highway Traffic Safety Administration) has its headquarters in Washington DC, home of some of the worst drivers in the nation.  Not sure how many of the decision makers work there compared to "regional offices" (which appear to be more any state vehicle/highway department than anything else).

PS: the cyberattack issue is much deeper and having AI drivers barely increases the attack surface.  Once I learned that Tesla cars had their brakes altered (anti-lock braking fixed) over a pushed wireless update, I just gave up on the hope of any attempted defense against cyberattacks (it doesn't help to work for a defense company that is deliberately removing security under government orders).

The issue here is that the train driver is usually unable to do anything to avoid an accident as the train braking length is longer than the line of sight. Some exceptions mostly then already stopping the train at an station and someone fall onto the track the driver can do an emergency braking and stop early, also cases with good visibility and cars stuck on the crossing, how does this happen so often? 
We simply accepted that trains can not stop if its something on the track who is not another train, other trains are not an major problem as trains are already managed by computers. 
Now with an AI system and cheap cameras this should be pretty easy to solve. simply put cameras on all the poles. 
it also solves the problems then the train control get bugs or stuff like drunks throwing shopping carts on the line. 
Nowday you have to wait for inspection or drive un-managed, there the driver has to be able to stop withing line of sight. Again much easier to solve with cameras and AI. 
it would also prevent derailing and damage to trains because of damage to tracks are not picked up. 

[SunlitZelkova]

7 hours ago, wumpus said:

A bigger issue is with self-guided trains, I'm not sure the human can do much other than get PTSD from an approaching accident.  In highly litigious countries (i.e. the USA), this makes sense as a CYA move, and can help the jurors feel empathy for the hapless engineer unable to stop the train in time.  Elsewhere, I'm unsure why you would inflict PTSD on the engineer because somebody just *had* to sit on the tracks/beat the train across a road (the train isn't going to stop regardless of whether or not the AI begins braking).

Remember that considering a single AI driven death unacceptable sacrifices 30k (-1) Americans (I'm not aware of the international numbers) to the "perfect safety" gods.  I'd like better standards comparing the expected number of accidents (and fatalities) per mile in terms of BAC (blood alcohol content) or eventually the numbers of miles (or km) an AI can drive compared to a human with equal safety.  An AI with a safety rating of ".8" or better might be vastly safer for a driver with a DUI conviction than a BAC interlock test.

Note: In the US, the NHTSA (National Highway Traffic Safety Administration) has its headquarters in Washington DC, home of some of the worst drivers in the nation.  Not sure how many of the decision makers work there compared to "regional offices" (which appear to be more any state vehicle/highway department than anything else).

PS: the cyberattack issue is much deeper and having AI drivers barely increases the attack surface.  Once I learned that Tesla cars had their brakes altered (anti-lock braking fixed) over a pushed wireless update, I just gave up on the hope of any attempted defense against cyberattacks (it doesn't help to work for a defense company that is deliberately removing security under government orders).

I think the issue is preventability.

Humans will continue to have issues forever- there will always be those who become intoxicated and cause harm to others in some way, that is just something generally accepted by humans in general.

Whereas some AI caused accidents could be prevented by the vehicle being controlled by a human. The AI car's existence is at fault and there is a clear method by which future accidents can be prevented- not having AI controlled vehicles on the road. Except instead of happening after the accident, this is happening before.

This isn't to say one or the other right or wrong- this is just the mechanism of what is happening.

Thanks for the info on cyberattack potential. I was completely unaware some human controlled vehicles were vulnerable too.

[JoeSchmuckatelli]

Robot Dawgs on Mars? 

https://www.cbsnews.com/news/nasa-human-like-robots-boston-dynamics-spot-explore-caves-mars/

[kerbiloid]

10 minutes ago, JoeSchmuckatelli said:

Robot Dawgs on Mars? 

Do they read this forum or watch sci-fi?

Once again, for the NASAn slowpokes.

Spoiler

 

 

[JoeSchmuckatelli]

Also - for nerds with no life...   https://m.huffpost.com/us/entry/us_610e96dae4b05f81570ac962

Edited August 9, 2021 by JoeSchmuckatelli

[DDE]

10 minutes ago, JoeSchmuckatelli said:

Also - for nerds with no life...   https://m.huffpost.com/us/entry/us_610e96dae4b05f81570ac962

Another year inside? Ugh.

[DDE]

 

[ARS]

Does ballistic weapon the worst choice for space-based weapon compared to energy and missile based? Since it needs physical projectile (more mass to carry), generate heat (especially for railgun, since in space you really need highspeed projectile to hit anything) and the need for counter-thruster to mitigate recoil (more propellant dedicated for the gun). Compared to the guidance capability of missiles and sustained use of energy weapons

[JoeSchmuckatelli]

53 minutes ago, ARS said:

Does ballistic weapon the worst choice for space-based weapon compared to energy and missile based? Since it needs physical projectile (more mass to carry), generate heat (especially for railgun, since in space you really need highspeed projectile to hit anything) and the need for counter-thruster to mitigate recoil (more propellant dedicated for the gun). Compared to the guidance capability of missiles and sustained use of energy weapons

Everything involves some compromise or another.

IIRC Niven or Pournell had guided missiles with nuke detonators that actually powered an energy weapon (the bomb was only the fuel source for the killing beams).  

Typically, you need to first assess your target before designing a weapon to defeat that target - so... What do your targets look like?  

If they're just satellites orbiting this planet, I would argue that terrestrial or in atmosphere weapons are likely sufficient to defeat those. 

If you are talking interplanetary craft hauling stuff - perhaps a combination of guided and directed energy are needed. 

The other thing is to know what you want your weapon to do - if you want to disable the craft and kill crew so you can harvest resources... You have additional limitations.  If you want to obliterate it - you have to build more robustly.

 

Edit - but ballistics ain't bad... A bunch of cheap, tiny, fast moving rocks have been used successfully for thousands of years to defeat a variety of targets 

Edited August 10, 2021 by JoeSchmuckatelli

[kerbiloid]

You need 100 .. 250 m/s to deorbit the shell.

A gun can provide ~1 km/s of delta-V.

So, it may push it sideways-backwards to deorbit, and this gives ~1 km/s of perpendicular velocity, and thus up to 1 000 km of side distance.

If take  M109 Palpatine and put in orbit, it could spread 28 40 kg shells in such wide area.

The problem is accuracy and recoil.

But the recoil could be compensated by shoting in turn to the left and to the right, so the howitzer would oscillate about some average orbit (anyway it will be destroyed in minutes).

Te accuracy requires to use some guided shells.

The heat protection forces to use several hundred kilogram capsules, similar to the film capsules of good old times, and make the gun bigger.

Advantage: simple shells, single expendable cannon.

[DDE]

9 hours ago, ARS said:

Does ballistic weapon the worst choice for space-based weapon compared to energy and missile based? Since it needs physical projectile (more mass to carry), generate heat (especially for railgun, since in space you really need highspeed projectile to hit anything) and the need for counter-thruster to mitigate recoil (more propellant dedicated for the gun). Compared to the guidance capability of missiles and sustained use of energy weapons

It has the advantage of cheap projectiles with destructive power that doesn't decay over range, so it has a niche in saturation fire.

Dumbfire shells and guided missiles also form the two extremes of a whole continuum of physical projectiles.

[JoeSchmuckatelli]

Is there any reason to make a space missile cylindrical?  

Would a spherical missile be more maneuverable given that everything is closer to the COM? 

(Smart Guys: how would YOU design a spacecraft launched antispacecraft missile?) 

[SOXBLOX]

8 minutes ago, JoeSchmuckatelli said:

(Smart Guys: how would YOU design a spacecraft launched antispacecraft missile?) 

I know you weren't asking me, but I'll add my two cents anyways. A hexagonal cross-section would make a projectile that's really easy to stack. Not sure if stacking missiles is a good idea, but it maybe it could work.

I guess the major reason for cylindrical missiles in the first place is that cylinders have the second-highest volume-to-surface-area ratio, after spheres.

[JoeSchmuckatelli]

10 minutes ago, SOXBLOX said:

guess the major reason for cylindrical missiles in the first place...  cylinders have the second-highest volume-to-surface-area ratio, after spheres

One of those things that I don't know off the top of my head - grin!

I always assume that the shape is for aerodynamic performance - which you don't need in space.  Thus I wonder if tube shaped missiles on a spaceBattleKruizer are an anachronism 

[kerbiloid]

31 minutes ago, SOXBLOX said:

A hexagonal cross-section would make a projectile that's really easy to stack.

Iirc, BDArmory HE-KV is squared.

[DDE]

3 hours ago, JoeSchmuckatelli said:

Is there any reason to make a space missile cylindrical?  

Would a spherical missile be more maneuverable given that everything is closer to the COM? 

(Smart Guys: how would YOU design a spacecraft launched antispacecraft missile?) 

Well, in all likelihood you need to push it out through a hole in your spacecraft. A completely spherical design isn't terribly conducive to that.

It also gives too small a lever for RCS.

[kerbiloid]

https://www.globalsecurity.org/space/world/russia/17f111.htm

[Souptime]

What the heck is an electron orbital? why can scientists just spot the electron?