In the last two years, major advances have been made in deep learning. Applications have been developed that can scan through images and automatically pick out detail, objects and people within seemingly random pixels, without using any subject specific algorithms, just by feeding massive amounts of input through a closed system. This is truly a breakthrough with enormous implications. Google, Facebook and several others have gone on a spending spree, snapping up companies for 100’s of millions to acquire the technology.
Yet this isn’t A.I., even though it may match the current dictionary definition.
Artificial Intelligence, as understood by most of us, is something else. Deep learning represents one of the initial challenges one the path to true artificial intelligence. And there is no doubt it is a major step forward. There are even larger leaps forward that are required to truly take us on that path. No one is even sure quite how or if those leaps can be made.
A number of industry figures have spoken of the dangers of A.I., the threat it may pose to humanity. And they are right. Would a truly sentient, artificial being consider humanities needs above those of other species, or the planet? Or would it consider us a plague, a disease to be dealt with and perhaps eradicated. No one really knows.
Advances in deep learning will continue. In perhaps five or ten years, it will become integrated into our lives in ways we cannot yet imagine.
As to A.I., there needs to be vigilance, in the scientific community, research labs, start-up, to what creatures they may inadvertently create.
Update 19th July 2015: it seems I am not the only one to classify deep learning as “not quite A.I”. The article here describes deep learning as Applied, Weak and Narrow Artificial Intelligence.