Differences between AI simulation translation and a human interpreter
Photo by Steve Johnson
As artificial intelligence is increasingly adopted by businesses, speech-to-speech technologies offering automatic interpretation (hereinafter referred to as “AI”, also known as AI simultaneous translation) have recently emerged, accompanied by promises of profitability and a high return on investment for companies that implement them.
However, it is clear that AI cannot provide the same cultural background, expert knowledge of languages, local information, mastery of the complexity of human oral communication, or professional experience as human conference interpreters.
Human interpreting can be defined as the simultaneous or consecutive real-time translation performed by a conference interpreter of an original speech into a target language, enabling the in-person or online audience to fully understand the original message in their native language.
How do we demonstrate that human interpretation is more reliable?
While it is obvious that human interpreters cannot match certain aspects of algorithms (such as working 24 hours a day, 7 days a week, without pay and without sleep!), it is worth highlighting some key elements that demonstrate that human interpretation is always more reliable and remains the only option in most communicative situations between humans:
Distinctive features
– Humans are naturally more capable of communicating than algorithms, especially when it comes to complex content that involves understanding specialised vocabulary, different speeds and accents, prosodic semantics, body language and facial expressions, nuances of meaning such as irony, humour, sarcasm, silences… and a deep knowledge of complex and highly specialised fields such as medicine or law.
– Humans are more precise with the vocabulary and terminology they use, as we always take cultural and linguistic nuances into account. Obviously, machines do not think, understand or correct themselves.
– Humans can guarantee the confidentiality of information. Cloud-based speech-to-speech technologies supposedly respect confidentiality, but how can LSPs (language service providers) and end customers really know if their data is being stored somewhere, shared with third-party servers, or misused or misshared now… or in three years’ time? Once again, can we be certain that these companies are not using our data to feed their algorithms in order to improve their linguistic models (and profit at the expense of our confidential information)?
– Machines lack humanity. In our fast-paced, technological urban societies, the human factor is, if anything, more important than ever.
– AI-powered simultaneous translation technologies are not reliable enough. These technologies are not mature enough to be effective in all markets or communication situations. The companies that market them want to be technological pioneers, which is perfectly legitimate. Making false promises is not. Customers should thoroughly test these technologies before compromising the effectiveness of communication or, worse still, losing prestige with their customers. AI constantly makes mistakes, producing results that are meaningless or completely inaccurate.
– Economic interests. Let’s not be naive. Startups and large corporations see AI as the goose that lays the golden eggs. It’s understandable; we all want to make ends meet. However, this technology is being marketed without being able to do what it is supposed to do. Worse still, it simply doesn’t work. And it’s not the international community of conference interpreters saying this; it’s the WHO, in this report, following a scientific analysis of numerous speeches. Our sincere advice to companies wishing to try AI tools is to contact an expert provider, such as CGS Translations, so that they can fully understand its limitations and risks and make an informed decision about when and how to use it.
Companies are trying out these programmes and realising that they don’t work, so they are returning to their trusted human conference interpreting providers.
Coexistence is possible, but with common sense.
Finally, it is possible to combine AI tools with human interpretation. This is something that professional interpreters, professional associations, and language service providers have been doing for several years to offer greater efficiency and productivity for everyone.
If you would like to try out these technologies, contact us!