In a recent article, Eleni Vasilaki, Professor of Computational Neuroscience at the University of Sheffield, reminded readers that humans tend to be afraid of what they don’t understand. According to Vasilaki, some technological achievements surpass expectations and human performances are to the point to look unrealistic and surrounded by a ghastly mystery halo.
A common mistake is in considering AI applications singularly and fearing humans to be replaced. Singularity is near, but nearness is relative. Vasilaki points out that AI is task-oriented, while humans are versatile by nature. Human versatility comes from an understanding of the world, and this in turn is developed over years. No AI seems likely to achieve this understanding anytime soon. People seem to overlook how much the huge amount of data and computational power available today might be the reason for the success of today’s AI.
First Man has brought back memories of the debates around the utility of the space program prior to the launch of the Apollo 11 mission to the Moon in 1969. In a paper prepared for IAF’s meeting in Stuttgart in 1952, Wernher von Braun wrote: “When we are asked the purpose of our striving to fly to the moon and to the planets, we might as well answer with Maxwell’s immortal counterquestion when he was asked the purpose of his research on electrical induction: «What is the purpose of a newborn baby?»” Today, few seem to pay attention to the fact that the impressive technological development of recent years owes almost everything to the space program.
A by-product of the mission to the Moon was the belief that any technological achievement is possible and at hand, and this might be one of the reasons for the cyclical proposition of new technological hypes. As Isabella Massardo reminds, in the last decade, speech-to-speech technology has been a constant hype, while machine translation has reached the plateau of productivity. Blockchain, together with cryptocurrencies or alone, also has been a hype for a few years now. In 2017, blockchain was already on the verge of disillusionment. In 2018, blockchain (now for data security) is still a hype. Not surprisingly, among the emerging and rapidly accelerating technologies that are advised to be actively monitored as disrupting innovations for being expected to profoundly impact the way of dealing with workforce, customers and partners none is directly related to translation.
Indeed, democratized AI might make digital twins closer than blockchain, as hundreds of millions of things are estimated to have digital twins within five years. Actually, according to Gartner, blockchain “has the potential to increase resilience, reliability, transparency, and trust in centralized systems.” The keyword here is “centralized systems,” while it is now pretty clear that the magic word to sell blockchain is “decentralization”. Unfortunately, the decentralization of business models and processes is definitely not straightforward for most businesses. As a matter of fact, many are still trying to understand what blockchain is and how it works and, more importantly, how it can be utilized for mission-critical applications. Not surprisingly, Gartner anticipates that through 2018, 85% of projects with “blockchain” in their titles will deliver business value without actually using a blockchain. Also according to Gartner, “blockchain might one day redefine economies and industries via the programmable economy and use of smart contracts, but for now, the technology is immature.”
Even technology enthusiasts should better be cautious about the prospected use of blockchain in translation. Maybe, translation blockchain enthusiasts might answer a few questions and help clarify:
- How is blockchain supposed to solve the perennial problem of interoperability?
- How is blockchain supposed to have more professional translators to match demand?
- How is blockchain supposed to open up language platforms?
- How is blockchain supposed to guarantee security, confidentiality, and privacy?
- How is blockchain supposed to cut translation prices further?
- How is blockchain supposed to make translation quality quantifiable?
- Is the network for translation blockchain open?
- How is mining implemented, through PoW or PoS?
- Mining for cryptocurrencies requires huge investments; this is why it is rewarded with cryptocurrencies, which are negotiable. Are “tokens” negotiable too?
- Given the investment in tokens required, how can users be guaranteed against a lack of transparency and a possible crash?
Contrary to what has been happening where the introduction and implementation of blockchain is advocated or has been taking place, no one in the translation industry has been asking any of these questions, at least publicly or outloud, and obviously no answer has been given or anticipated so far.
Presenting interoperability as a dilemma still in 2018 means that the translation industry if far away from maturity. Since inception, the translation industry has been said to be on the edge of a massive change in how they receive and translate content. Changes have actually happened over the years, coming almost exclusively from outsiders. Major translation buyers have been imposing their own solutions to their own problems to their suppliers who, in cascade, have imposed these solutions to their own vendors. The fragmentation of the industry has effectively prevented the birth of any real industry standards, further encouraging this intrusiveness. Translation industry players have always been so obsessed with the risk of compromising their own little garden up to rejecting, if not hindering, where possible, any real standardization effort. Major players have been trying, in turn, to take advantage of any standardization initiatives, even those that they themselves advocate, to enforce their own models and maintain what they see, often wrongly, as a competitive advantage.
This attitude is in blatant contrast with any new methodologies, but has the reassuring effect of keeping players in a sort of comfort zone, allowing them to prevent any “resource dispersion” and contain any losses due to the inefficiencies ensuing from their immobility. This is also why the processes of most LSPs are optimized for small projects and why organic growth and a critical mass are so hard to achieve. Unfortunately, process efficiency comes from design and technical interoperability is effective only when technology matches processes, not vice versa.
Everyone working in the translation industry knows the problems permeating it. Listing them is barely a starting point towards a solution whatsoever.
How is “tracing a user’s history” supposed to be “increasing trust for the translator’s ability and capability?” How is the tracking of digital assets supposed to benefit their creators when blockchain in no way can guarantee ownership? A ledger is used to record transactions not to certify the ownership of the assets in each transaction.
Therefore, Kirti Vashee’s doubts here are well expressed: “Everybody involved in blockchain seems to be trying to raise money. The dot-com boom and bust also had, to some extent similar characteristics, with promises of transformation and very little proof that anything that was clearly better than existing solutions. I feel the problem description of the LIC initiative is clear in this overview but I am still unclear on what exactly is the solution. I would like to see examples of a few or many transactions executed through this blockchain to see how it is different and better before, I cast any final judgment.”
The translation industry is an intricate intertwinement of relationships between the businesses, players, publishers, analysts, and consultants governing its economy. In this context, the difference is made by who you know. For this reason, ignoring who Renato Beninatto is is tantamount to a lèse-majesté offense and it is not exactly clever for someone in a prominent position to ignore him or, even worse, pretend to ignore him, as Lionbridge’s CEO, John Fennelly reportedly did at LocWorld 38 in Seattle, even though or especially if he comes from another industry and a different experience.
The intertwinement of relationships that characterizes the industry has resulted in exclusive clubs that have their meetings at industry events. Each area of the industry has its own club, and each club has its governance. Occasionally, members of different clubs from different areas mingle, but generally clubs remain distinct. Some clubs are more numerous or powerful than others and their governance may be assimilated to a mafia, as a young and overly ambitious would-be analyst and consultant named it. He also did whatever it took to join it, and he made it.
As long as you are a member of one of these clubs and share its spirit and its policy, you can be sure that any initiative you take will not be hindered, far from it. No one will ever challenge you or even ask you any embarrassing questions.
For this very reason, though, the questions on the openness of the blockchain network and the negotiability of tokens are fundamental. Blockchain may have the potential to increase resilience, reliability, transparency, and trust in centralized systems, but the most powerful promise of blockchain is about decentralization. Being extremely clear on the openness of the blockchain network and on the associated protocols is paramount.
Clarifying the negotiability of “tokens” is equally crucial. Indeed, more and more often, “investment” is the other word accompanying cryptocurrencies, even though, in principle, they are not supposed to generate returns; after all, it’s just software. But they are used also to purchase goods having a countervalue in fiat money and are then negotiable. Bitcoin, for examples, can be converted into cash, using a Bitcoin ATM or a Bitcoin debit card or via an online service. Joining a token-based translation blockchain network would require an initial investment in tokens, whether on a barter exchange for data or in fiat money. If tokens are distributed by a centralized entity, this entity would most probably be asking people to purchase tokens. Even though any new users that would join the network won’t fund older users, the founders will end up being the richest ones guaranteed, as in a typical Ponzi scheme: The more people join, the more the founders will earn. And this is the only way they can make money. From nothing, as the only asset of founders is the network. Their net worth would be in fiat currency while the members of the network would not be able to cash their tokens after having bestowed their data assets to the network, and if the network crashes they might be dumped with nothing.
Finally, with merger or acquisition accounting for growth at 3 of the top 5 fastest growing LSPs for 2018 it is hard to believe that these will join the blockchain network anytime soon. And, by the way, there has always been only one man in black.
The comparison with the automotive industry and the car is definitely out of scale, but it is true that translators too use only a fraction of the many features available in any translation software tool. Also, the automobile is now a general purpose technology and the only possible comparison might be with the smartphone.
Yet, although “augmented translation” is just diverting marketing crap, if democratized AI will make any sense, it will help redefine the value of linguists rather than taking jobs away from them.
Arthur Clarke’s famous quote above explains why technology is outpacing our ability to comprehend what we can do with it. The next new thing in the translation industry will very soon be conversational agents and virtual assistants rather than blockchain.
Virtual assistants, aka chatbots or bots, already are or are going to be the bridge between technical documentation teams and customer support and power most of customer service interactions. Indeed, technical support is the most common type of chatbot content, and bots are said to be the new FAQ.
Technically speaking, there are two kinds of virtual agents:
- One kind is scripted. It can respond only to questions that it was programmed to understand.
- Another uses AI, so it can understand what the customer is telling it, and its knowledge grows the more it interacts with people.
The issue, today, is how to prepare, organize and structure content so that chatbots can use it.
Translation industry players, from each side of the fence, have learnt to reuse content, while CMS are still underused, especially for single-sourcing. The next challenge for content producers is to extrapolate answers to customer questions from a unified set of content modules delivered across channels, rather that creating new batches of (largely duplicated) content or recreating content by copying and pasting existing content from their CMS into a form that chatbots can use.
More technical authors will be needed accustomed to single sourcing through CMS. Will they be translators accustomed to leveraging past translations using TMs?
In fact, Microsoft has already issued a new chapter of its style guide devoted to writing for chatbots.
The main components of chatbots are four:
The “things” users are talking about with a chatbot; they can be inherited from taxonomy nodes in a CMS.
The goal of a user’s interaction with a chatbot; it can be mapped as content elements in a CMS and be defined as primary and alternate questions.
The (unique) questions or commands a user asks a chatbot.
The answers the chatbot returns to utterances; they can be defined in a CMS.
The coming future authoring skill consists in breaking existing content into smaller, modular chunks within CMSs, to achieve COPE (Create Once Publish Everywhere), the new holy grail.