In the field of science, unexpected turns often occur. So is AI. I am researching algorithms that use AI to measure vehicle speed with CCTV or recognize robot motions with cameras. This research topic may seem unconnected, but it has happened thanks to the rapid advances in deep learning. In the past, it would have taken a long time for research to solve a problem in one field to spread to another, but this is not the case these days. Current AI research seems to be full of many undiscovered breakthroughs. And one of the breakthroughs is ‘translation’. Not translating into a language we are familiar with. In other words, a surprising shift in thinking is taking place, ‘translating the world’. I want to tell that story.
Machine translation that resembles living things
First, let’s go back to the history of machine translation. Machine translation is one of the long-standing challenges of AI researchers. Its roots go back over 70 years. It was around the time the first generation of computers appeared. At that time, researchers were full of confidence. I thought the problem would be fixed soon. But waiting for them was not a pleasant ending. What became clear as time went on was the only sign that their research was not going to come to fruition.