Please follow and like us:

On June 27, 2023, the Circle hosted an online event about AI and a technology known as Augmented Translation. The featured speaker was Miguel A. Jiménez-Crespo, an ATA certified English to Spanish translator. He also holds a PhD in Translation and Interpreting Studies from the University of Granda and is a Professor in the Department of Spanish and Portuguese at Rutgers.   He is particularly interested in how humans interact with technology so that the technology can become more ergonomic.

According to Mr. Jiménez, almost all human translation nowadays involves an interaction with a machine or some external source of information other than our own brains. For example, if a translator uses a computer and a Word program to prepare a translation, the translator is interacting with a machine. In using translation memory, we are actually using someone else’s brain to do our work. Even the use of a dictionary involves dependence on an external source of information.  These external sources enable us to do our work better and more rapidly.

Human Interaction with machines on the part of translators can often generate negative emotional reactions to automation. For example, translators are concerned about the downward pressure on rates posed by machine translation which may negatively impact their opinion of machine translation. Some translators may also have negative reactions to certain features of machine tools such as CAT tools. These reactions can create anxiety on the part of the translator. This kind of reaction is known as cognitive friction.

Mr. Jiménez then introduced the concept of Augmented Translation. Its purpose is to help humans solve problems better and faster. It is a human-centered approach to translation in which the human is in control. It seeks to enhance the work done by humans by uniting the strengths of both humans and machines. This technology is different from AI since in AI there is no human involvement.

Mr. Jiménez then discussed the technology known as neural machine translation or neural architecture.  This is the technology used by DeepL, ChatGPT and Google translate. It seeks to replicate the workings of the human brain.

While machine translation is statistical in nature in that it generates word choices by means of probabilities, neural machine translation captures the long-term dependencies between words in a particular language. By accurately analyzing the positions of words within sentences, the technology chooses the word that is correct for a particular context. It is not that the machine is actually thinking: it is merely predicting the next word because it has been pre-trained with huge amounts of data culled from the human translations available on the internet.

Mr. Jiménez sought to dispel the fear that technology will replace human beings. He believes that humans will continue to surpass machines when they are translating high value content or engaging in tasks that involve creativity, critical thinking, judgment and storytelling. AI on the other hand is more appropriate for lower order tasks and lower value content.

The speaker believes that the future of AI in translation will involve striking the right balance between humans and technology and determining which tasks can be automated, which can be augmented, and which can be left exclusively to humans. The level of technological support will depend on the complexity of the task at hand.

Patricia Stumpp


Please follow and like us: