INTERVIEW 'Human, AI translators complement each other', says Flitto CEO
.jpg?w=728)
By Kim Bo-eun
Flitto CEO Simon Lee
Earlier this year, human translators beat artificial intelligence (AI)-powered translation programs at a competition held at Sejong University.
But there are still concerns that AI translators will someday outplay their human counterparts, painting a bleak future for the profession. After all, last year, Google’s computer program AlphaGo defeated Go master Lee Se-dol in a historic match that signaled the possibilities of AI outperforming humans in various fields in the near future.
Simon Lee, CEO of Flitto, a global crowd sourcing translation platform, rejects this idea of intense competition between human and AI translators.
“Even professional translators rely on machine translations for rough drafts,” Lee said in an interview with The Korea Times.
“Human and machine translators will continue to support each other, and both sectors will grow rapidly,” he said.
Flitto was founded in 2012 with the aim of helping people overcome language barriers.
“The name was derived from ‘flit to,’ which means fly quickly from one place to another, like transcending language barriers,” Lee said.
Since it participated in Techstars London’s incubator program, the first Asian start-up to do so, in the same year it was founded, and won the Seedstars World Competition in 2014, Flitto has been carving a niche in the global translation market, which has an estimated size of $40 billion in 2016 and continues to grow.
The Flitto model
Flitto’s service model is based on the need for human minds in the process of translation.
Users can request translation for a simple slang or phrase that is unavailable on Google Translate, and instantly get answers from translators around the world.
The users can then select from the various answers based on the translators’ ratings.
While translators are required to pass a test to be able to provide these quick translations, they may still commit errors. In addition, the answer users select may not necessarily be the most accurate.
To this drawback of the service, Lee is unfazed: “Errors exist on every platform including Google and Naver.”
Lee, who has a good command of Korean, English and Arabic and is knowledgeable of French and Chinese, confidently uses Flitto to communicate with foreign employees and when traveling for business.
In addition to quick translations, Flitto also facilitates translations for longer documents, such as resumes, business presentations and theses.
Once users submit a request, professional translators, who undergo a more rigorous certification process, present their offers, outlining the cost and estimated completion time.
The users select the offer that suits them; and they can also communicate directly with the translators during and after the task such as to get an update on the task or to request revisions on the translated document, respectively.
“People who submit translation requests to conventional agencies never know who translates their work because the agencies conceal this information at all costs to prevent the customer and translator from dealing directly with each other,” Lee said.
In contrast, Flitto’s model presents customers with a wider range of available translators, and can establish trust between the customer and translator as their identities are known.
Currently, 7.5 million people in 173 countries use Flitto’s translation services in 18 languages, and around 1.2 million translators perform an average of 70,000 translations daily.
Users avail the services by purchasing points with credit cards, and translators can convert the points they earn into cash or coupons that can be used at convenience stores or coffee shops.
Flitto takes a 10 percent commission from the payments for professional translations, which is considerably lower than the market rate for agencies.
Its main source of revenue is sales of data collected from the platform to other businesses.
Status of AI in translation
According to Lee, major translation platforms such as Google and Naver employ the same technology, neural machine translation, which uses a greater pool of data to provide more accurate translations than statistical machine translation, an earlier technology.
What distinguishes translation services are the additional features they offer, which Lee explains using the case of burger restaurants.
“If burgers taste more or less the same, features such as a restaurant’s interior and perks, like free soda, will make the difference.”
Baidu, for example, offers translations for various types of content from around the world, such as cartoons, while Google uses “Word Lens” to automatically provide translations of words when pointing a smartphone camera at them. Naver’s translation service Papago facilitates context-specific translations by enabling users to set a specific location, such as a police station.
Flitto users can already get instant translations when they take pictures at the Museum of Modern Art in New York or Uffizi Gallery in Firenze, Italy. It is also getting ready to launch a service that will provide users translations and pictures of dishes when they take a photo of a menu.
Flitto is working with the Seoul Metropolitan Government to make the service available at the city’s restaurants.
Lee predicts translation demand for Arabic will grow but says currently, Arabic translation services are few. Thus, Flitto will also launch a service that will enable users to take pictures of Arabic text and get translations of the text, as Arabic characters may be difficult to type for non-speakers.