In fact, companies have been insisting that translation firms slash their fees and work alongside machine translation systems from DeepL, and Google. The only problem is that automatic translation systems are universally shite and it is only a matter of time before some big corporate gets taken to the cleaners in court for using them. While they can be useful for cut and pasting clear text, legal, or technical translations (which make up the bulk of translation work) are at best unreliable, and at worse outright dangerous.

For example here is a Deepl translation from Italian for a legal document to give you an idea.

“The company owning the property, herein sold, having to dispose of numerous real estate units that make up the building and for which the finishing works have not been carried out, reserves the right to carry out what will be agreed with the eventual purchasers, also with the displacement or construction of partitions and walls, with the creation of utilities, services, installations, accesses for the single building units without having to ask for authorisations to the purchasing party, respecting in any case the units sold or promised to the purchasing party and respecting also, as far as possible, the indications provided by the purchasing party for the finishing touches to be carried out in the common parts of the building.”

To be fair, the Italian was shite too, but this sentence caused smoke to pour out of Deepl’s ears.

A human translation was:
"The company owning and selling the property must dispose of several units that make up the unfinished building. It may carry out what will be agreed with future buyers, by moving or constructing partitions and walls, creating utilities, services, installations, accesses for individual building units without having to ask the buyer’s permission. The integrity of units promised or sold to the buyer will be maintained as will instructions for the finishing to be carried out in the building’s communal areas."

Microsoft has an additional problem in that its machine translation system is even less useful than Deepl or Google and has been churning out translations that are even more literal.

Now Vole has announced an update to its translation services that, thanks to new machine learning techniques, promises significantly improved translations between a large number of language pairs.

Based on its Project Z-Code, which uses a "spare Mixture of Experts" approach, these new models now often score between 3- 15 per cent better than the company's previous models during blind evaluations.

Z-Code is part of Microsoft's wider XYZ-Code initiative that looks at combining models for text, vision and audio across multiple languages to create more powerful and helpful AI systems.

"Mixture of Experts" isn't a completely new technique, but Vole claims it is especially useful in the context of translation. At its core, the system basically breaks down tasks into multiple subtasks and then delegates them to smaller, more specialised models called "experts." The model then decides which task to delegate to which expert, based on its own predictions. Greatly simplified, you can think of it as a model that includes multiple more specialised models.

Again it sounds really good if you are a technical person, but for a translator or editor, it sounds truly bad. Basically, it means that one AI makes a prediction and allocates the sentence based on its own assumptions to a sentence “expert” which makes another decision. While this might be good at picking up jargon terms, it depends on the original literal translation for accuracy.

So if you are a new generation of the translator, forced to accept a “machine translation” and a pay cut. You will have to go back to the original translation anyway to see if the first subroutine made a mistake. It would be a lot quicker for you to just translate the thing yourself and accept the fifty per cent pay cut.

The other alternative is to wait until a company is sued, or loses a fortune after one of its machine-based translations stuffs up something important. Frankly, the technology is not there yet.