AI and Robotics

I watched Horizon last night, as I saw it was about artificial intelligence, which I find quite interesting.

There were a few examples of present robots and computers, varying from the incredibly knowledgeable but non-robotic Watson computer (which triumphed in a special edition of the US quiz Jeopardy) to more anthropomorphic creations.

The programme did pose some interesting philosophical questions about intelligence and whether AI can truly exist. The presenter, mathematician Marcus Du Sautoy, did a Chinese room experiment. He was in an enclosed room with a letterbox through which a fluent Chinese speaker posted simple messages. He then replied, despite knowing no Chinese at all and having no idea what he was saying, by posting back messages in accordance with a rulebook he had in the room. The Chinese speaker said that if she’d been speaking with him online she would have thought he knew it fluently and would have no idea he was utterly unaware of what he’d been saying.

The point of that is that we cannot know whether a robot is truly thinking as we do, or whether it’s blindly following instructions. Du Sautoy countered this by suggesting that humans themselves follow ‘instructions’ in the way we think (a view to which I’m inclined).

He also encountered a robot deliberately made to have many human structures, with a plastic skeleton, joints and thread doing the work of tendons. The scientist behind that creation thought that real intelligence was contingent upon having a physical form that interacts with the world. He pointed out that robots are phenomenally good at things we thought they might find hard (playing draughts) but very bad at things we thought they’d find easy (moving the pieces).

The final chap, a German, that Du Sautoy met had perhaps the most interesting robots of all. There were three of them, roughly 60% the size of a human and with a similar composition (2 legs, 2 arms). They invented their own language, giving a word to a gesture and then teaching it to one another and then Du Sautoy (who got it wrong at first). The German scientist suggested that robots really need to evolve in the same way that humans develop from infancy to adulthood, and the fascinating interaction between his creations seemed to back this up. It is hard to know, however, how much behaviour is simply programmed and how much is due to the robots ‘thinking’ for themselves.

For those Britons wanting to watch the programme, here’s a link to it on the iPlayer, which should be up for the next few days:

http://www.bbc.co.uk/iplayer/episode/b01fmbvb/Horizon_The_Hunt_for_AI/

There’s also an interesting moral question. If we created a robot species with roughly equal intelligence to our own, would they be free, or slaves? Would the abolition of ‘off’ switches become a political rallying point for robot rights activists?

Whilst we’ve had tremendous progress in computing power I’m not sure how long it’ll be before those questions become live issues. It’s worth mentioning, though, that mobile telephones were fiction not so very long, as was the internet.

In related news, I read a tweet by the lovely Miss Plato featuring a new video from those clever boffins at Boston Dynamics. This time they’ve made a little robust robot that can jump rather impressive heights:

http://www.youtube.com/watch?v=6b4ZZQkcNEo&feature=youtu.be [For some reason embedding isn't working]

Thaddeus

The Robotic Cheetah

Military robots have long been desirable. They won’t ask for wages, won’t have qualms about doing unpleasant but necessary tasks and won’t even complain. It would also be nice if patriotic citizens didn’t get killed, and instead hardware got forcibly decommissioned.

Some time ago the dog (which is more of a mule, as it’s intended for weight-bearing primarily) was unveiled. It’s not, I think, finished, but it’s clearly pretty advanced. The robotic beast of burden has phenomenal balance, a steady pace and the ability to walk on pretty much any terrain.

Now the clever chaps at Boston Dynamics have created a robotic cheetah, which has recently broken the speed record for robots (now 18mph, having previously been 13mph).

Technological progress and non-human combatants have always been critical in warfare. Whether exceptional cases (Hannibal’s use of snakes in naval warfare or Wojtek, the Polish bear who fought in WWII) or more common ones (elephants, horses, camels) they can add a tactical or strategic advantage.

I think robotics could play a great role in logistics and intelligence (it’s not hard to imagine even more advanced surveillance robots as well as the dog or cheetah above), but I’m doubtful as to whether armed robots would be wise. They’d need to be able to differentiate between enemies and allies and I don’t know whether that will ever be possible, given the wide variety of context. If a robotic mule breaks then gear has to be carried by something else, but if a robotic velociraptor broke down it could pose a serious risk to civilians or its own side.

Thaddeus