AudioVisual Translators Europe (AVTE) issued a statement on generative AI which I found brilliant. I agree with everything that’s said in it, and its content could apply to any type of translation (audiovisual, literary, technical, legal, medical, etc.).
Every day, I see my beloved profession sadly turn into a sick industry, where linguists are financially exploited, end clients are fooled and quality is suffering.
Why? Because misleading information about artificial intelligence (AI) and machine translation (MT) is being circulated. It makes these technologies look much more suitable and harmless than they are.
To illustrate this, I wanted to share with you some key points from AVTE’s statement. I could have quoted the whole document given how spot-on it is, so I strongly encourage you to read it in its entirety here.
🔥How AI really works🔥
“[generative AI] utilizes massive amounts of data to predict the next word in a sentence based on the context provided by the preceding words. […] This is not a meaning-driven or purpose-driven translation like one done by a human.”
“Generative AI doesn’t come up with original concepts by itself. Instead, it re-elaborates (possibly copyrighted) materials, making artistic work less and less original.”
🔥How AI impacts on quality🔥
“relying on machine-generated text for translation dramatically increases errors of meaning, style and general readability. Even when they are accurate, machine-translated texts are poorly written, and being exposed to bad writing can be detrimental to the writing skills of both post-editors and end users.”
🔥How AI impacts on our society🔥
“Some propose a compromise in which audiovisual MT would be confined to lowbrow entertainment […] while high-calibre titles would still be handled by experienced linguists. […] Such distinction shows intrinsic bias and elitism: who is to decide on the value of a cultural product?”
“we must not overlook the massive energy consumption it demands, its impact on the environment, and the exploitation of ‘microworkers’, humans (often from poorer countries) undertaking repetitive, low-paid tasks to train the AI”
🔥How AI use can be highly unethical🔥
“Leading clients to believe human translators will take care of their works and then sell them MTs is unethical and in bad faith, even more so if the output is used to train the machine without the client’s authorization.”
〰️〰️
As stated by AVTE, “translators are willing to integrate new technologies in their workflow as they have been doing for decades”. Translators are not complaining about AI and MT just for the sake of it. They are complaining about misuse (if not abuse).
I don’t need to be told that technologies are essential to my work. I’ve known this for a long time. What I want instead is, like AVTE again perfectly puts it, “that they are used for improving human output, which is clearly not the case for the proposed applications of generative AI in translation”.
Thanks to AVTE for publishing this decisive document!