Artificial Intelligence helps to break down language barriers
Jakob Uszkoreit works for Google’s Artificial Intelligence department defining new automated translation services.
Imagine that we can maintain a conversation in two completely different languages at the same time and we don’t even notice / Imagina que podemos mantener una conversación como esta en dos idiomas completamente diferentes al mismo tiempo y que ni siquiera lo notamos”. An instant translation like this, at the very moment it is produced, is the near future that Jakob Uszkoreit visualises when he talks about the possibilities of artificial intelligence applied to translation.
For Uszkoreit, a member of Google’s Brain Artificial Intelligence group, the future of automated translation systems is about improving the interfaces and the user’s ability to translate what they need. That is to say that the key for him is that more and more languages can be translated from a larger number of applications and technologies. For example, referring to a live conversation, Uszkoreit values being able to have a headset with a built-in microphone that translates the messages it receives in real time. It would be like simultaneous translation, except that on this occasion the translator is an artificial intelligence engine instead of a person.
His team’s work focuses on two pathways. On the one hand, they are training the algorithm to understand more languages; from those spoken by a large number of people to minority languages whose study requires research with the community itself to “understand their speech, writing, and the way they express themselves”. The other branch of his project is about finding new services and simpler interactions with users.
“The basis of people’s interaction with Google is no longer as visual, nor are traditional graphic interfaces used. Instead, the current key is language– you ask a question and you get quite a natural response”, explains Uszkoreit. In his head, the futurelooks like a reality in which interaction with machines through language will be done in a natural way: “The lines between analogue communication and communication via a device will become blurred, and machines will participate in conversations translating, offering information or responding to questions that arise as if they were simply another conversational partner.”
His interest in computing and language comes from when he was little, and Uszkoreit’s journey to end up working for Google has been long and winding. Born in Silicon Valley to a family of German immigrants, he moved to Germany at a young age, where his father – a specialist in computational language – introduced him to this field. He was then trained in Computer Science at the Berlin Institute of Technology, and would end up receiving a grant from Google in 2008 and returning to Silicon Valley.
There he began to collaborate with the machine learning team that Google Translate was developing. “At that time, mentioning this service made people laugh, but the team had a lot of energy, a desire to learn and very smart people,” explains Uszkoreit, adding: “They knew they could break down language barriers”. Now, Uszkoreit is expanding his vision toward machine learning in images and videos, and he has undertaken a second return journey, this time to Berlin to take charge of Google Berlin’s Brain Artificial Intelligence group.