ttbn,
Let us not forget that invention that has helped mankind and womankind to a happier and healthier lifestyle.
Maybe the French didn’t invent it but they have it named after them, at least in English.
The well known and appreciated “French Letter”; rarely does it fail to do the job for which it is intended and many of them helped in WWII by being streatched over the muzzles of artillery pieces to keep out sand and water.

Vive Le Franger Francaise!!

plantagenet: I don't think we've the same idea about autonomous vessels in mind. I not discussing something like a remote controlled machine, but one directed/operated by an onboard artificial intelligence.

So what I've in mind is a hypothetical AI autonomous vessel which is basically a submarine but with no human crew (and as a bonus consequence they can be a lot smaller since they don't need a great void in them for humans to work and live in and also need none of the human life support infrastructure like the oxygen and temperature regulation, etc.). So the tasks they perform are very similar to whatever it is that submarines do now- eg: Note that here I know next to nothing about submarines so I'm just speculating but I imagine the a human captain receives orders for higher up to do things like track a particular ship, or harass/attack any enemy vessel that enters a particular area. And for the sub AI, just as for a human captain, once given a mission it is up to it implement the details without the need for further control/communication from those that gave the orders.

This is currently technically possible with planes and other machines so I don't see how it is impossible to implement it on a submarine. Here is an aircraft that is available to today (in fact these have been around for years) that operates autonomously: http://www.iai.co.il/p/harpy . Basically you give this thing the command to go over to the enemy's side and loiter/wait in the air there scanning the battlefield below and if it detects a worthwhile target it will destroy it (the targets it's interested in are enemy radar- it is not specifically designed to kill people). So you just set it, then forget it and leave it to do its thing, ie: once launched it acts by itself.

The previous example is designed not specifically for killing people, but there do exist autonomous drones that are (in addition to doing other stuff as well). And they're available for purchase today, they're not just prototypes/concept aircraft.

--continued below --

--from above --
Indeed there is speculation that such a drone has already conducted a fully autonomous kill in battle: http://www.newscientist.com/article/2278852-drones-may-have-attacked-humans-fully-autonomously-for-the-first-time (the drone in this article is the Turkish made Kargu-2).

Now, in these examples given I don't think the AI onboard is very advanced. In the first case, from what I understand, the Harpy more-or-less just flies around and listens for enemy radar signals and attacks its source. For the case of the Kargu-2, because it is such a small drone I don't think it could have much in the way of processing power on board so it couldn't be a very advanced AI.

A few years ago AI technology in general was still quite primitive. However, over the last 5 or so years AI (especially neural network based stuff) has really started to rocket ahead. It is truly amazing what current AI can manage to do. One of the best known astounding examples is autonomous driving: cars are rapidly improving and are becoming proficient at self-driving. In some respects and circumstances they are already better than human drivers. There is no reason to not believe that they won't be able to surpass the average driver in terms of safe, smooth and efficient driving in just mere years. So it seems plain to me that given the current rate of general AI improvements that in 20 years it will be mind-blowing what machines can do by themselves. And one very obvious application of this will be in defence. And this is why I suspect a human-crewed sub that will take 15-20 years to become operational will not be in-use for very long before it is obsolete. It may even be obsolete before it is even deployed.

Hi thinkabit

Thanks for advising:

"Note that here I know next to nothing about submarines so I'm just speculating but I imagine the a human captain receives orders for higher up to do things like track a particular ship, or harass/attack any enemy vessel that enters a particular area."

So basically you don't know enough about the inter-action of human crewed submarines with (alaways only partial) IA drones in operational situations.

Cheers

Pete
Of "Submarine Matters" after more than 10 years talking to submariners.
http://gentleseas.blogspot.com/

plantagenet:

/"So basically you don't know enough about the inter-action of human crewed submarines ..."

Why would you say this? The whole point of an autonomous machine is to remove the human element! So it's not at all important what the interaction between the humans on existing crewed submarines is because on an autonomous vessel there are no humans. An example of this is the Norwegian cargo ship that I linked to in a previous post: in just a few months this will operate COMPLETELY autonomously, there will be NO CREW on board! So the fact that current human crewed ships have a Captains, Chief Mates, Second Mates, Boatswains, Able seamen, First Engineers, Second Engineers, Stewards, Cooks, etc. is irrelevant. An autonomous ship simply doesn't need to be designed and operated around the current fixed notions of what people in these positions are responsible for and what they do. All this cargo ship has to be able to do is complete its overall objective: which is to safely and in accordance with the law/customs move cargo from one dock to another. Ie: dock, be unloaded any cargo for delivery, reload with any new cargo, undock, sail to the next destination and then repeat this process (with incidentals like refuel which in this particular case is be plugged in since it's electric, monitor and report on problems, etc.).

/"... IA drones needs to be only partial IA"

Again, why do you say this? The whole point of an completely AUTONOMOUS AI war machine is to be just that, ie: autonomous! There is no fundamental law of physics or computation that says that AI enabled machines need to have a human involved with its operation. It's perfectly plausible to speculate that AI is will one day soonish (almost certainly with two decades) be smart enough to operate a killing machine completely autonomously at a comparable level of proficiency to that of humans. Once we're at this level, you just give the machine an appropriate overall objective and then it goes off on its mission and tries to accomplish this objective.

--continued--

-- from above --

How it goes about this business is up to it, according the standards that its been designed to. Now (hopefully), because Australia is a country that respects human rights and is a signatory to several conventions (esp: the Geneva conventions) any killing machine that we deploy will have very high standards, eg: not killing non-combants, not deliberately destroying civil property, etc. However, other countries (eg: most of Africa and the Middle East) where they just don't give a f!ck about human rights then their killer robots/machines will most likely be allowed to be more cavalier in their methods. But an the end of the day there is no fundamental technological reason why an autonomous killing machine needs a human to be involved in its operation.

In general, from reading what you written, I'm willing to bet that you're not up-to-speed about how advanced current AI technology is. But here's something for you to think about: today I drove my car about 80 km's on an outing. On that drive I passed, within just mere meters, at least 100's of other cars driving the opposite direction at deadly speeds. Also I navigated passed cyclists, over pedestrian crossings, through roundabouts and intersections (some with lights / some not) and dodged debris and potholes. This is an incredibly complicated task that requires me to be capable of operating the car, know the road rules and understand how other people behave so that I can anticipate their actions. And I do all this within a tolerance of just seconds or less of response time and within the tight spatial bounds of the driveable road/lane width. And the most amazing thing is that TODAY CARS CAN DO THIS almost as well as a humans!! After thinking about this, hopefully you can acknowledge just how vastly complicated safely driving a car is* and how astounding it is that we are currently of the cusp of full-self driving cars

-- continued below --