>>831647
What makes you think they're actually from /pol/?
I'm actually going to make a semi-constructive post instead of being fucking triggered on the assumption that OP is just naive and acting in good faith.
Google's Tensor flow is a bunch of neural networks cascaded together. They've trained the most basic networks that do feature extraction of lines and whatnot themselves. A bit of background for OP who may not know, a neural network s just a series of nodes connected by weightings. It's literally just a bunch of summation and multiplication operations, the amount of multiplication is what you are training for.
So google has given you a bunch of math and some tools to produce some more math by calculating more multiplication factors based upon images you feed into it. These images are stored on your computer and you can see in the source that nothing is sent back to google. You can even monitor your network to see if it is making requests if you want.
How google makes a profit from this is 1) They wanted image recognition for themselves so they made it 2) They sell hardware that accelerates the software they have written and want you to buy it.