Google Creates Artificial Brain That Loves Cats
As a way to create a simulation of the human brain, a group of researchers at Google constructed a neural network made up of 16,000 computer processors. Then, they turned their creation loose on the internet where it immediately developed a taste for cat videos. This thing is so human it's scary!
The brain's creators fed it millions of random thumbnails extracted from YouTube videos. After processing the data, the neural network actually taught itself to recognize cats with no further help, which is an amazing cognitive feat for a machine.
The brain was even able to assemble an image of a cat, mimicking what humans do in their visual cortex. “We never told it during the training, 'This is a cat,'" said Google fellow Jeff Dean. "It basically invented the concept of a cat."
Researchers point out that the artificial brain is still tiny when compared to the human brain, which is many times larger in terms of neurons and synapses. Still, the research is important since it essentially proves that machine-learning algorithms improve greatly when exposed to large amounts of data.
Now, if this thing could only learn to recognize corgis. That would be truly impressive.