Creativity is a quality we have all had to develop as a response to the COVID-19 pandemic, so this year’s Creative Language Conference also called for some creative repurposing as an online event via Zoom. While there are certain aspects of a physical conference which cannot be replicated in this way – the coffee and free pens – it is remarkable how we have adapted to virtual meetings as social and networking forums. The organisers, Crisol Translations and Retro Digital, deserve credit for scheduling networking breaks into the busy day, as well as managing timings impeccably.
AI, ethics and creativity
First up was Arwyn Bailey, ‘The Time Lord‘, on AI and visual language in photography – a topic unfamiliar to many of the participants, but a fascinating insight into how AI is used in many widely available image manipulation tools. Arwyn demonstrated how AI is not a threat to the creative industry, but can be a tool to enhance creativity. Throughout history there have been fears about technology destroying jobs, but in fact the nature of work has changed and new kinds of jobs have been created in response to technological change. Alwyn pointed out that ‘‘we have to consider how much society is prepared for things to go wrong before they go right’. This is the crux of the ethical issues posed by AI – we have to know the failure points of a system to enable repair, but this can come at human cost – such as the much-publicised accidents involving autonomous vehicles. We also have to remember that human error is a factor in many failures: there are examples of this throughout history, such as the SS Eastland disaster, which ironically resulted from new legislation on lifeboat provision following the sinking of the Titanic.
Bots and bias – AI in customer support and marketing
Tina Squire of Interact also addressed the issue of how AI will impact on our work, in particular within customer service. Tina shared Arwyn Bailey’s view that AI is leading to a shift in the way we work, rather than destroying jobs. The work of contact centre agents is actually becoming more skilled rather than less, as bots are now doing the menial tasks and agents are training them. A bot can handle the low-complexity parts of a conversation, but skilled human input is needed when an issue becomes more complex.
Sabina Jasinska of Just AI explored the ethical concerns of AI in marketing, in a thought-provoking session which covered topics from bias in AI systems to persuasion via anthropomorphic voice technology. We have moved away from the ‘Mad Men’ era when advertising was a creative industry: it has now become a data-driven industry, ruled by algorithms. We do not always recognise the hidden advertising which targets us via recommender engines and personalisation, as we largely accept this as a consequence of our use of ‘free’ platforms like Google and Facebook. Sabina’s talk also touched on the issues of algorithmic bias (which is really human bias!) and the possibilities of using AI for positive applications such as adaptive and assistive technology.
Linguistic creativity and transcreation
Those looking for more in-depth discussion on language technologies were not disappointed, as there were also a number of talks on more specialised translation and localisation topics. Giulia Tarditi led us through the use of source-free transcreation at Monese, the innovative challenger bank whose customers are largely non-native English speakers. It is important to them that the entire customer journey is localised and that they ‘speak their customers’ language’. Content is scored on the basis of its impact on UX, allocating more resources to strings with higher impact and using regular translation or machine translation for lower impact segments. Removing the source allows translators the freedom to work as copywriters – a blurring of roles which is becoming more apparent in many parts of the language industry.
Jamie Brown spoke about the importance of language expertise at what3words, a service which converts GPS coordinates into an easily memorable three-word address which customers can use to describe their location for a variety of purposes, including alerting the emergency services. The company’s ambition is to be ‘the global standard to tell where people are’. The three-word system has obvious challenges including eliminating homophones, ensuring each address is unique and avoiding potentially offensive words and phrases. What3words is currently available in 46 languages, with further expansion planned, and each address is created from scratch in a new language.
To round off the day, Gaetan Chrétiennot from Six Continents gave an overview of the machine translation (MT) landscape in 2020. There are now more words translated by MT in a day than by human translators in a year and neural MT has now achieved ‘human level’ – meaning average human level rather than the level of a skilled translator. The suitability of MT for any given piece of content depends on its predictability and creativity is not predictable, since it is not based on repeating patterns. There are particular issues with user-generated content, a category which is growing due to the rapid expansion of social media. Gaetan proposed a ‘human-optimised, machine first’ model of MT, using neural MT as an assistant for creative texts.
The conference may be over, but discussion continues in the LinkedIn group and many attendees have already connected on LinkedIn and Twitter – this conference was an excellent networking opportunity, with a wide range of participants from various sectors and locations. Definitely a worthwhile event for anyone interested in the intersections between AI, language technology and the creative industries.