Killer Robots

Fathom · 01-06-2018 7:55pm

Krishnan, A. (2016 ebook), Killer Robots, 1st Edition, London: Routledge. Also... Daniel Wilson (2005), How to Survive a Robot Uprising, Bloomsbury Publishing.

Black Swan · 02-06-2018 4:35pm

It appears that the UK and USA are standing together regarding the use of future autonomous killer robots before the UN.

Fathom · 04-06-2018 7:00pm

USA & UK protocols? Or lack of?

Black Swan · 08-06-2018 10:42pm

Fathom wrote: »

USA & UK protocols? Or lack of?

Noted earlier, Hayley Evans of Lawfare (13 April 2018) was a good source for detailing the issues faced by both UK and USA regarding the prohibition of autonomous weapons systems. In summary, it is too early to tell what the UK and USA will do regarding Killer Robots. It would seem obvious they are developing such, but as to deploying them in sufficient numbers to make a difference in modern warfare, that remains problematic at present.

AlphabetCards · 09-06-2018 1:05pm

It has got to the point that off-road autonomous gun carriers (small tracked vehicles with chain gun) are being trialed with line infantry in 29 Palms with some US Marine Corps units. Suffice to say, there must be some operational autonomous units with the likes of their special forces.

Fathom · 09-06-2018 9:33pm

Aircraft carrier groups. 5 defensive weapons systems. Faster than humans. CIC now has final keyed button. For now.

AlphabetCards · 10-06-2018 8:20pm

To be fair, close-in weapon systems on naval assets work to a set of very well pre-defined boundaries and rules. For example, there will be little to no reason for an airborne object to approach a carrier group at mach 2 or on some sort of terminal descent, similar for the anti-FIAC situation - if a group is well clear of allied littoral regions and there's a number of small craft inbound at speed, it's not likely to be a friendly boarding party. Where AI will come into it's own, is when a crew have been on 6-on/6-off for months on end and they are getting sloppy (it happens, no offence to any matelots on the board) and they react slowly to dealing with inbound threats. At least AI will pick it up at earliest opportunity.

Black Swan · 10-06-2018 9:15pm

AlphabetCards wrote: »

To be fair, close-in weapon systems on naval assets work to a set of very well pre-defined boundaries and rules.

What about the "pre-defined boundaries and rules" pertaining to civilian passenger Iran Air Flight 655 flying from Tehran to Dubai (3 July 1988), shot down by an SM-2MR surface-to-air missile fired from USS Vincennes, a guided missile cruiser of the United States Navy? Also, commercial flight pilot error, flight software and hardware program failures, and flight course changes to avoid adverse weather may result in problematic behaviour of civilian flights, which may, or may not intrude on "pre-defined boundaries and rules" programmed into autonomous weapons systems. Of course, the weapon system was not entirely autonomous, and required a human decision before fining, but had it been autonomous, who knows?

AlphabetCards wrote: »

Where AI will come into it's own, is when a crew have been on 6-on/6-off for months on end and they are getting sloppy (it happens, no offence to any matelots on the board) and they react slowly to dealing with inbound threats. At least AI will pick it up at earliest opportunity.

The USC Immersive Naval Officer Training System (INOTS) attempts to address such problems through active training and simulations, etc., with results that suggest improved efficiency and effectiveness for this problem mentioned, as well as a host of other problems.

Fathom · 11-06-2018 1:43am

Encyclopaedia Britannica:

Iran Air flight 655, flight of an Iranian airliner that was shot down by the missile cruiser USS Vincennes on July 3, 1988, over the Strait of Hormuz, killing all 290 people on board. The passenger plane, which was in Iranian airspace, had been incorrectly identified as a fighter jet.

Capt'n Midnight · 11-06-2018 6:13pm

Fathom · 11-06-2018 6:20pm

Killer robots or robot killer?

AlphabetCards · 12-06-2018 6:05pm

Of course, the weapon system was not entirely autonomous, and required a human decision before fining, but had it been autonomous, who knows?

I was referring to CIWS systems such as Goalkeeper and Phalanx which are guns (and have a range of about 4km) rather than missiles (the one you refer to goes well past 70km). The criteria for both these zones and modes of threat engagement are hugely different - anything 4km from a battlegroup will be well assessed and characterised by the time it gets there.

AlphabetCards · 13-06-2018 12:01pm

Capt'n Midnight wrote: »

Jesus christ. This is surely a wind-up? The recoil from that will destroy the whole unit, and there's not a hope of accuracy unless they fire point-blank.

Capt'n Midnight · 13-06-2018 6:26pm

AlphabetCards wrote: »

Jesus christ. This is surely a wind-up? The recoil from that will destroy the whole unit, and there's not a hope of accuracy unless they fire point-blank.

what recoil ?

Rubecula · 14-06-2018 11:59am

some folks just want to kill, there will always be people like that I am sorry to say.

Black Swan · 15-06-2018 10:13pm

Rubecula wrote: »

some folks just want to kill, there will always be people like that I am sorry to say.

Makes me wonder if there was an evolutionary survival advantage for such a characteristic, if in fact it was somehow inherited or culturally learned? If so, what happens when such a characteristic is somehow programmed into autonomous weapons systems; or not?

Capt'n Midnight · 15-06-2018 10:39pm

Black Swan wrote: »

Makes me wonder if there was an evolutionary survival advantage for such a characteristic, if in fact it was somehow inherited or culturally learned? If so, what happens when such a characteristic is somehow programmed into autonomous weapons systems; or not?

I'd reckon it's genetic. Compare chimps to bonobos.

Our civilisation is a balance between a few people being able to accumulate vast riches and power at the expense of others, and a general lifting of the rest of us.

42 people now have the same resources as half of all humanity.

The rich are getting richer. It used to take a lot more billionaires to match the poor and there's been population growth. Things like the weekend and annual holidays weren't given freely. "I predict a riot"

Here in Ireland we've had a very long period of relatively quite industrial relations as some wealth was shared. In France the air traffic controllers are having their annual strike.

There's an old saying that a rising tide lifts all boats and social coefficients predict that societies are more stable if wealth growth is shared.

I'm less envious if I'm getting richer too. But if you are getting richer and I ain't then don't expect sympathy.

So for robots on auto it depends. If they are working together or if they are competing.

*googles DIY EMP devices*

Black Swan · 15-06-2018 11:04pm

Capt'n Midnight wrote: »

So for robots on auto it depends. If they are working together or if they are competing.

AI programmer colleague claims that there are AI systems that have built-in competition in their algorithms, and from this competition the AI programme learns and advances, while continuing to compete at an extraordinary rate. He had a name for this self-competitive algorithm, but I cannot remember it at this moment. I can just imagine such an AI system shaping and rapidly evolving the programming of killer robots.

Capt'n Midnight wrote: »

*googles DIY EMP devices*

Grand humour!

Fathom · 15-06-2018 11:42pm

Black Swan wrote: »

He had a name for this self-competitive algorithm, but I cannot remember it at this moment.

Evolutionary algorithms?

Rubecula · 16-06-2018 2:01am

we see competitive spirit in humans as a good thing but not to the point of destroying somebody else (in general terms anyway) we can only hope AI systems of the future will be as moral. It is always possible that the competitive streak is an aberration too.

Black Swan · 16-06-2018 3:32pm

Rubecula wrote: »

we can only hope AI systems of the future will be as moral.

"If one uses morality in its descriptive sense, and therefore uses it to refer to codes of conduct actually put forward by distinct groups or societies, one will almost certainly deny that there is a universal morality that applies to all human beings." Does this suggest that what may be "moral" for one group or society may not be for another? This reminds me of the cliche, "One group's asset is another group's liability." Or yet another paraphrase: one group's terrorist is another group's freedom fighter. Will AI killer robots evidence group-specific morality accordingly?

Source:
The Definition of Morality, Stanford Encyclopedia of Philosophy, Revised 8 February 2016.

Rubecula wrote: »

It is always possible that the competitive streak is an aberration too.

My housemate and I have been rereading The Lessons of History by Will and Ariel Durant, wherein they conclude that war is the ultimate form of human competition, and that war historically is normal, and peace is not. "War is a constant in history, there being only 268 years of no war in 3421 years of history."

To what extent do these conclusions suggest that "killer robots" may be a normal extension of human behaviour, that may evidence group-specific morality?

Dr. Bre · 16-06-2018 3:37pm

AI will be the death of humanity .. humans are the future sheep cows. That’s if we are lucky enough for the robots to even let us be slaves

Rubecula · 16-06-2018 6:15pm

Dr. Bre wrote: »

AI will be the death of humanity .. humans are the future sheep cows. That’s if we are lucky enough for the robots to even let us be slaves

we will not be alive to see that outcome, imho.

Black Swan · 16-06-2018 6:27pm

Dr. Bre wrote: »

AI will be the death of humanity .. humans are the future sheep cows. That’s if we are lucky enough for the robots to even let us be slaves

Rubecula wrote: »

we will not be alive to see that outcome, imho.

When Skynet becomes self-aware.

Capt'n Midnight · 18-06-2018 7:46pm

Rubecula · 18-06-2018 10:50pm

Capt'n Midnight wrote: »

omg .. :eek::eek:

Fathom · 18-06-2018 11:51pm

Turing Test. Would we fail too?

Rubecula · 19-06-2018 3:19am

Fathom wrote: »

Turing Test. Would we fail too?

most certainly

Black Swan · 20-06-2018 7:26am

Rubecula wrote: »

most certainly

Raises hand.

Fathom · 21-06-2018 11:14pm

Ex Machina exceeds expectations?

Killer Robots

Comments