A computer can certainly beat the majority of people at chess and, if you believe theoretical physicist, cosmologist, and University of Cambridge professor Stephen Hawking, the rise of the robot could have potentially very serious implications for the entire human race. Writing exclusively for RetailTechNews, Ed Bussey, CEO of Quill, explains why we shouldn't be handing everything over to the computers just yet.

Institutions such as PwC have also joined the chorus of voices heralding the rise of machines. It claims that more than 10 million UK workers could see their jobs performed by machines within 15 years, with roles that include repetitive or formulaic tasks at particular risk.

News of robots or artificial intelligence replacing humans appears to always make good fodder for headlines – but are these predictions wildly exaggerated? The answer may not be a simple ‘yes’ or ‘no’.

In roles where creativity and perceptiveness are required, the prognosis for humans is more optimistic. Focusing particularly on retail, the creation of online product descriptions in multiple languages is something that still requires a human touch – despite recent advances in machine translation.

For our report: ‘Man vs. Machine – Is machine translation ready for retail?', we asked 400 consumers to compare machine and human translations of the same product description copy, without knowing which version was which. They were asked which descriptions made them feel more likely to buy the products and which left them with the best impression of the retailer.

For both these metrics, the copy translated by humans dominated across all markets and verticals.

In fact, almost four-in-five people (79%) across the Chinese, Japanese, German, and French markets stated that the human-translated copy made them more likely to purchase the product in question. That’s not all: 80% of respondents agreed that the human-translated details gave them a significantly more positive impression of the retailer.

But, surely, with recent innovations in machine translation – in particular, the introduction of ‘deep neural networks’ that draw on vast reserves of language data to produce intelligent context-based translations, rather than word-for-word substitutions – we should expect to see machine-translated copy making a better impression on consumers?

Interestingly, our research does reveal that 50% of consumers find machine translations equally as easy to understand as their human-translated counterparts across categories such as fashion, homeware, and beauty. But, critically, the technology is still not good enough to convert browsers into buyers.

Fundamentally, consumers do not find machine-translated content as persuasive, because the technology still struggles with the more nuanced aspects of translation that really sell a product – such as capturing a brand’s unique tone of voice, navigating cultural idiosyncrasies, injecting flair through more creative methods like metaphor, and interpreting colloquialisms. These emotional components are vital to any product description if a brand is to engage its consumers emotionally, build trust, and inspire them to want to part with their cash.

Take the fashion vertical, where our respondents were shown machine and human translated versions of a product description of a silk shirt. The machine translation outperformed its human equivalent when it came to basic understanding of the content; but over three-quarters (77%) of people across all four markets agreed that the human version made them feel more likely to buy.

There’s also the question of SEO to consider: machine translation software will always produce the same output from the same source text; whereas different human translators will produce individually tailored and differing outputs, even if the source data is the same. For retailers who want to produce variations of product description copy for different channels – e.g. for marketplaces versus on their own e-commerce sites – machine translation simply won’t meet this need, producing duplicate content that is highly damaging for SEO.

Of course, there’s no denying that machines and AI are likely to become increasingly important in the field of content creation – and retailers should stay on top of these innovations. However, for the reasons described above, we suggest they should – for now at least – take a ‘humans in the loop’ approach.

This means using machine translation technology to produce a ‘rough draft’ of the content, speeding up the process – but also deploying a skilled, native-speaking human editor to review and amend the content to ensure it is bespoke, carefully tailored to the consumer audience and the brand’s tone of voice, and optimised to deliver the desired commercial results.

For the foreseeable future, at least, any retailer that relies exclusively on machine translation at this early stage of the technology risks directly damaging its bottom line.