Our AI Positioning Statement

However, at the time of writing, AI is shaking up the translation industry. It is an understatement to say AI public COTS translation systems are disrupting and penalising the market and clients. It sacrifices quality for quantity. It undermines the English language (and all other target languages). And today, it is even jeopardising the employment of translators, both new and experienced, with some wondering why they spent all those years studying, and others wondering whether they will have work tomorrow. In short, AI is a cause of concern.

At a time when most stakeholders are making extensive, reckless and ill-considered use of AI public COTS translation systems, often at the expense of quality and data security, we believe it is worth taking stock of the efforts Infomonde is devoting to this highly topical issue.

We have benchmarked and benchtested all publicly available COTS AI translation systems, and none of them achieved a passing mark against our 8 quality criteria, the 4 human translation quality standards* and the 7 data security standards** applicable to our business area. Some COTS systems even scored minus 4 on a 10-point scale! Actually, none of these systems guarantees the security of clients’ data and the quality of the documents, simply because these tools are far from being fully developed and are being brought to market far too early, but also because, in many instances, they are not designed or developed for translation as we practice it in the context of bilingualism in Canada and in the rest of the world, nor for communications in general. Actually, some of these systems were not developed to generate high-quality translations or effective communications, but simply to meet basic information needs, and in some cases the requirements of US military intelligence. (Relevant passages were first identified in the source texts, then translated roughly, and finally retranslated by human translators.)

We have benchtested all the public COTS AI translation tools available on the market.Our benchmarking and benchtesting efforts have led us to the following conclusions:

1. Clients should be warned against using these public COTS tools, for quality and security reasons. All these tools are underdeveloped, incomplete and unfinished. The translations generated by all these systems are mediocre. These tools do not secure clients’ data. The unsecured use of these AI translation tools carries major risks: data compromission and leakage, in addition to a marked decline in target translation quality. Furthermore, these COTS tools frequently ‘fail to understand’ basic concepts, specifically because their developers are themselves unfamiliar with these concepts or unable to identify them, which distorts the meaning of the message and seriously misleads the reader.
2. There are no cost or time savings to be made using COTS AI translation systems when the aim is to produce high-quality texts. This is what we could observe on our test benches. Post-editing means that translators have to significantly rework, or even completely rewrite the AI-generated texts, which ultimately costs more than the translation itself. And when this post-editing is not carried out – which is very often the case – the quality falls far short of what is required.
3. It is possible to achieve considerable cost and time savings without compromising the translation quality and data security. Please read ‘OUR HI-AI TOOL’ section on this site to find out about the ways and means we have developed to achieve results fully compliant with translation quality and data security requirements.
4. All these AI public COTS tools are far from being fully developed and are purely experimental and speculative. The GAFAM (Google, Apple, Facebook, Amazon, Microsoft) and the unicorns of this world are investing colossal amounts (over 800 billion US dollars at the time of writing) in experimenting with these tools, which they must quickly sell to us in the form of licences in order to recoup all these investments.