12/7 Reading

Machine learning, a subfield of artificial intelligence, only recently became increasingly prevalent due to availability of things such as GPUs and publicly available large data sets. It really is quite remarkable how the field of machine learning has suddenly become the forefront of computer science research and development. What's especially interesting with regards to applying machine learning to things such as image recognition or translation in this case, is that training a neural network to do so doesn't actually require a whole lot of human intervention. You set up a neural network architecture, give it example translations, and let it learn by itself how to differentiate between good/bad, right/wrong translations. Machine learning is essentially programmers telling computers how to learn without needing to describe what to learn. You can give a computer a whole bunch of examples of translations between certain languages, and given enough training time and a relatively solid architecture, the computer will learn by itself to distinguish between certain features (i.e. grammar structures, punctuation, etc.) Though it will most definitely require a large amount of time and programming ingenuity to program and train a neural network capable of perfect (or more realistically, near-perfect) translations, I do believe that it possible to do so in the (near?) future.