Veil of the Minorities
Veil of the Minorities is an image produced using machine learning and neural networks. The principle component of training a neural network is data which the network uses to learn patterns. Lyonel Feininger’s Church of the Minorities is the input data, and the neural network attempts to reproduce the image. The training process allows the neural network to move towards a perfect reproduction of the original input image by reducing the number of pixels that are wrong in the output. Yet the final image is blurred, as though the observer perceives Church of the Minorities through a veil. No matter how long the network trains or how many resources it has, it converges to the blurred image you see above. The reason the network cannot reproduce the original image is because of a glass ceiling inherent in the architecture of the network; a point of diminishing returns is reached where further training produces negligible changes in the output.
Video of the training proccess on the left. On the top right is the original image, and on the bottom right is a graph showing the success of the neural network.
The composition quickly takes shape early in the training process, and later phases illuminate finer details like windows and the individual wearing red. The graph in the bottom right of the video above shows the success of the network converging towards an upper bound at the end of training. The vibrations of the aging neural network display its struggle to achieve the original artwork, even though that outcome is impossible given limitations inherent in the architecture of the neural network. How often does the architecture of a system, neural or societal, prevent achieving success? Society holds us to pixel perfect images of success, yet the system within which we live and our position within this system determine the bounds of achievement. It is not a failure of effort, but rather a failure of a system too simple to render success. Effort should not be wasted on minimal gains in a rigged system but rather on discovering better systems that allow opportunities for success.
Kawara is a clock that merges digital ease with analog elegance. It uses subtle lines and a strong font to bring style to your watch. The watch face uses the bold Futura font to display the time in a digital format, but the numbers draw from analog timepieces by using lines that run parallel to the hands on an analog watch.
Google Fit Watch Face
Google Fit makes tracking your fitness a breeze, just simply wear the watch and it automatically detects your walking, running, and biking. Google Fit encourages people to live a healthier life by motivating them to keep moving. This includes a watch face to track your daily activity, and smart notifications to keep you updated.