Sequential Decision Making
By Todd Bierman. In one tiny phrase, the key to better machine driven cars – in fact, to any task oriented pursuit including scientific research – is within reach. Google has announced the creation of an algorithm that successfully negotiates common video games by learning to better execute sequential decisions that lead to success.
Some call it trial and error, but the growing field of machine learning is striving for something a bit further up the evolutionary tree, hence the name Artificial Intelligence. The present iteration of self driving cars depend on detailed maps instead of experience. In effect, every move is pre-planned. The goal of the new algorithm is to allow a machine to learn the maps on its own, along with other important real life aspects of driving like adjusting to weather, and of course the biggy, avoiding other cars.
The technology isn’t evolving in isolation however. Recent advances in the field ofmetamaterials, that is, the creation of man made materials with atomic level precision are quickly leading to powerful and relatively less expensive sensors that can better see the self driving auto’s surrounding world. New metamaterial flat antenae can sense the entire road, its environs, and other cars – and all with no moving parts. Goodbye to the expensive if cool looking rotating radar sweeps.
Similarly, DARPA has announced the development of self-guided bullets. In addition to allowing a relative amateur to hit a fleeing opponent at 1000 yards, this technology allows real time flight adjustment of fast moving metal, and cars are often that.
The sum of things
The final puzzle parts to self driving cars are already here. Cheaper, massively parallel cloud computing let’s the better algorithms reach out and touch Detroit steel wherever it may be – even if it’s really Japanese aluminum. And so, with visions of Stephen King’s haunted car Christine dancing in my head – no sugar plum fairies here – we may want to insist that our AI is consciousness free, or at very least devoid of malevolent spirits.
By Todd Bierman. Thank you, TiA.