As opposed to tired, stressed and easily overwhelmed humans making tactical and strategic decisions? I share your concern but I don't think its impossible for an AI to get to the point where it makes less mistakes than a human.
The stereotypical danger isn't that they'll make fewer mistakes than a human, it's that the mistakes they do make will be much bigger--continually offloading decisionmaking and decision enforcement to AI without a coherent theory of friendliness is like using Martingale betting: You don't lose often, but it only takes one to wipe you out.
True. I would imagine that the way to use AI in the context of managing information overload on the battlefield is to have AIs prioritise information and propose options and have humans review the options and take decisions. By having the AI deal with incoming information and humans making the decisions you get the best of both worlds, at least with current technolody. Taking humans out of the loop entirely won't be a good idea for a long time yet.
Keeping humans in the loop doesn't do as much as we think, because humans have a natural inclination to trust computers. In a high-stress situation, such as a battlefield, a human will defer to a computer even if the computer reports information that, under normal circumstances, would be obviously false. In 1988, for example, the USS Vincennes shot down a civilian airliner because its AI targeting system had misidentified it as an enemy aircraft. Giving individual humans access to an off-switch is a pretty suboptimal way to prevent AIpocalypse.
Human brains have highly evolved empathy software that can cause a human soldier to pause before taking an action that will end with a lot of casualties. True, the process of making soldiers involves a lot of de-sensitizing these instincts. But my point is those instincts are present in normal human beings and have to be over-ridden with extensive training, vs. robots that are likely to be intentionally designed to be psychopaths.
Winning is the point of warfare. Inflicting enemy causalities just happens to be part of winning.
Ok, so rather than 'winning' it's achieving the desired change through application of force. If the point of warfare was inflicting enemy casualties, than we would have already won in Iraq and Afghanistan... and Vietnam. We've been moving away from the whole 'kill as much people as possible' method of warfare for quite a while.
Omelets require breaking eggs, but the point of omelets isn't to break eggs.
I've worked with small robotic military platforms, and AI must overcome major physical limitations before anyone even worries about whether the AI is making good decisions!
Battery life. If necessary, humans can stay active for a long time. Robots can't. If it dies, you either manually get the robot, or leave it for the enemy. The more processing and sensors on a platform, the sooner it dies. An almighty AI will likely require an almighty amount of power.
Cameras. Humans swiftly adapt to lighting, can deal with very dark and very bright conditions, have an amazing field of view, can accurately track tiny objects very far away, and very quickly refocus and swivel our eyes and head. You'll have a lot of trouble finding cameras+mounts that meet all of these criteria that aren't a giant bullseye. It's more likely that you need to tailor your sensors for particular situations. What we see as a person at 80 meters might be a few pixels if the camera is zoomed out. If you zoom in, you'll see the enemy, but not your surroundings.
Communications. Quickly transmitting binary data at the ground level is a very open-ended problem.
Positioning. GPS doesn't work within buildings. Some of your platform power will be allocated to dealing with this fact.
In an interview with Andy McNab (former member of the SAS and now a fairly successful author) in the Scotsman at the weekend I was rather surprised that he has been taking part in a UK research program to develop technology to suppress the parts of the brain that "implement" fear and empathy - arguably making a person more machine like. He was being studied as he clearly doesn't do much in the line of fear or empathy!
Or the neural "priming" mods from Greg Egan's Quarantine - or even, perhaps more alarmingly if the changes can be made semi-permanent, focus from Vinge's A Deepness in the Sky.
The future of warfare is in robots making war time strategic decisions, guys, can't we just take this to its logical conclusion and settle matters over a nice game of chess.... Best 2 out of 3?
Ultimately war will exist entirely online, because all of the most important things will be in cyberspace. Once reality as we know it has been reduced to simply hosting the processors that host our made up worlds, then maybe we can permanently keep war where it should be, inside video games.
No... war will now be over the processors and whatever keeps them running. If all of humanity lived their lives in virtuality, whoever controls the hardware is king.
"The future of warfare is in robots making war time strategic decisions, guys, can't we just take this to its logical conclusion and settle matters over a nice game of chess.... Best 2 out of 3?"
This was the plot of a Star Trek episode, where the losers had to docilely line up to be incinerated when the computer determined that they had been "killed" in the simulated battles. It was all much less messy that way.
Of course, when Kirk destroyed the incinerators (or was it the master computer?) they had to decide whether to fight the messy way, or maybe just not fight at all.
> The future of warfare is in robots making war time strategic decisions, guys, can't we just take this to its logical conclusion and settle matters over a nice game of chess.... Best 2 out of 3?
Why would the other side agree to use chess to determine the outcome? More to the point - if they would, why would you?
War is a decision-making process, but one of the constraints is that both sides aren't willing to give in absent significant costs.
There is a point to warfare. It is to make things so unpleasant for the other side that they have no choice but to submit.
The more we try to computerize warfare and make it antiseptic, the easier it becomes for us to ignore that fact. Admittedly this is effective - when we don't feel the pain of war we are able to inflict more of it - but fundamentally I think it is wrong.
There was a star trek original series episode about this. Two planets who had evolved warfare to the point that it was completely carried out in simulations, and casualties were executed by each respective side.
I believe kirk ended the show with destroying their computers and giving them a moral lecture on the horrors of war...
29 comments
[ 2.9 ms ] story [ 61.6 ms ] threadhttp://wiki.lesswrong.com/wiki/Paperclip_maximizer
A slight mis-prioritization on the AI's goals could easily lead to spectacularly unintended results.
Ok, so rather than 'winning' it's achieving the desired change through application of force. If the point of warfare was inflicting enemy casualties, than we would have already won in Iraq and Afghanistan... and Vietnam. We've been moving away from the whole 'kill as much people as possible' method of warfare for quite a while.
Omelets require breaking eggs, but the point of omelets isn't to break eggs.
Battery life. If necessary, humans can stay active for a long time. Robots can't. If it dies, you either manually get the robot, or leave it for the enemy. The more processing and sensors on a platform, the sooner it dies. An almighty AI will likely require an almighty amount of power.
Cameras. Humans swiftly adapt to lighting, can deal with very dark and very bright conditions, have an amazing field of view, can accurately track tiny objects very far away, and very quickly refocus and swivel our eyes and head. You'll have a lot of trouble finding cameras+mounts that meet all of these criteria that aren't a giant bullseye. It's more likely that you need to tailor your sensors for particular situations. What we see as a person at 80 meters might be a few pixels if the camera is zoomed out. If you zoom in, you'll see the enemy, but not your surroundings.
Communications. Quickly transmitting binary data at the ground level is a very open-ended problem.
Positioning. GPS doesn't work within buildings. Some of your platform power will be allocated to dealing with this fact.
[This is based upon my experience of trying to keep up with my son playing Halo rather than Ender's Game]
The interview is here:
http://www.scotsman.com/spectrum/Interview-Andy-McNab-soldie...
The bit about the brain scans is on page 5:
http://www.scotsman.com/spectrum/Interview-Andy-McNab-soldie...
Ultimately war will exist entirely online, because all of the most important things will be in cyberspace. Once reality as we know it has been reduced to simply hosting the processors that host our made up worlds, then maybe we can permanently keep war where it should be, inside video games.
This was the plot of a Star Trek episode, where the losers had to docilely line up to be incinerated when the computer determined that they had been "killed" in the simulated battles. It was all much less messy that way.
Of course, when Kirk destroyed the incinerators (or was it the master computer?) they had to decide whether to fight the messy way, or maybe just not fight at all.
Why would the other side agree to use chess to determine the outcome? More to the point - if they would, why would you?
War is a decision-making process, but one of the constraints is that both sides aren't willing to give in absent significant costs.
The more we try to computerize warfare and make it antiseptic, the easier it becomes for us to ignore that fact. Admittedly this is effective - when we don't feel the pain of war we are able to inflict more of it - but fundamentally I think it is wrong.
I believe kirk ended the show with destroying their computers and giving them a moral lecture on the horrors of war...
Anyway, your post reminded me of it.