谷歌沉[尋]夢是什麼來的?
WHAT IS GOOGLE DEEPDREAM?
People who know a thing or two about artificial intelligence believe this term is a misnomer and that the field will never progress beyond the machine learning that defines most of it in use today – with the possible exception of the DeepMind program, developed by some wunderkinds in the UK and acquired by Google (now Alphabet) in 2014. It now has research centers in London, Canada, France and the United States.
If you ever wondered why YouTube bothers to host billions of hours of video (an insanely expensive proposition) and how Facebook and Google can break even in a weak advertising market, remind oneself of the new mantra that, “Data is the new currency.”
But in what kind of reality could data ever truly be the new currency? In a Virtual Reality modeled on and simulating meatspace that the powers that be are desperate mirror perfectly, all the better to control reality, once every last person and object becomes a node on the Internet of Things.
Which brings us to DeepDream, which gives a visualization of how AI works, in a program that uses a convolutional neural network to find and enhance patterns in images via algorithmic pareidolia and/or facial recognition. The program was originally trained on animals and thus heavily favors dogs and birds.
Still, the resemblance of the imagery to LSD- and psilocybin-induced hallucinations suggests the artificial neural networks may be mimicking certain layers of the visual cortex.
Happy Trails!
沒有留言:
發佈留言