TECHNOLOGY

A little over a decade ago, if you would have asked the average person what a chatbot was, they might not have known. Chatbots were a cheap wine — still unrefined — and the technology wasn’t celebrated. They were mainly used for customer support when companies were too economical to use people, and their sophistication was limited to guessing what plane information was being requested and providing a less-than-full-bodied regurgitation from their FAQ page (it was always better to just search the FAQ page). Basically, they were automated automatons presenting very much like robots — everyone hated them. In fact, I found these statistics published just last year showing everyone still hates them then:

 “What Do Your Customers Actually Think About Chatbots?,” Userlike.com, 2022″

Advertisement

Chatbots were not cool, not expected to understand you, and not expected to actually be helpful, other than to expedite a connection to a human agent who could “better assist you.” You can imagine irate customers rage-typing slurs just before they’re transferred to the “next available agent.”

Through the 2000s, as the internet evolved to being the primary place of working and buying, connective technologies that facilitated automation advanced, and so did digital marketing. This meant potential access to many more buyers than the sales team could ever reach or process on their own, but they were more sophisticated buyers. Marketing had to morph into “digital marketing” in order to adapt to the new reality: sellers are forced to be reactive — simply make their presence known online via content, SEO and advertising, and patiently wait to receive potential inbound opportunities. Chatbots were a response to this: Instead of leaving control of the sales cycle up to the buyer, chatbots were a tool to move the buyers down the sales funnel, eventually convincing the buyers to buy on the seller’s terms.

Chatbots, originally a marketing technology, were not well-liked, but they had become an increasingly valuable tool for businesses beyond marketing. In sales, chatbots tried to influence customers’ purchasing decisions with lively recommendations based on their preferences and previous purchases. In marketing, chatbots tried to entice customers and promote products or services earlier in the sales funnel. In customer experience, it tried to provide 24/7 post-sale support and handle routine queries, freeing up human customer service representatives to deal with more complex issues.

Interest in “chatbot” Google searches from July 2017 to January 2022, with “100” indicating peak popularity.

Simultaneously, evildoers leveraged the same technology to try and scam people out of money by, among other strategies, impersonating the very brands using chatbots to try and sell things legitimately. People hated chatbots even more. Why were companies still trying to use this rubbish? Because they were the cheapest way to get customer attention, manage incoming orders, claim 24/7 support, and handle the majority of simple customer queries — if they worked. The market for chatbots was expected to soar.

Advertisement

What happened?

Throughout all this, robotic process automation programs, like Power Automate and UiPath, were on the rise. Automated companies, like aggregators, niche search engines, and social media sites, began appearing, and experts began predicting the transition to fully autonomous industries, like real estate and other agency-based business models. Nobody was sure whether or not they would live to see the day their job would be completely automated by computers. Sure, maybe there were waves of worrisome translators when translation memory crashed the scene, only to ebb and flow again when machine translation resurfaced as neural machine translations (NMT), but it never happened (and it never will, but more on that later). If the research by OpenAI, and other research by Goldman Sachs, is to be believed, 66-80% of the workforce will be impacted by the implementation of chatbots doing their job — pretty much all desk jobs, including translators. Chatbots aren’t doing 100% of these desk jobs (except for 15 or 86 occupations, depending on whether you ask a human or a chatbot), but 25-50% of their workload. How is that possible from a chatbot marketing toy?

On a typical day, desk people, whether from home or at an office, carry out important typing and mouse-clicking tasks — usually copy-pasting and/or transforming data/information from one format to another between browser tabs or desktop applications. Interrupting these important tasks were even more important meetings (that could not have just been emails) with an attached transcription and summary notes, that ensure team coordination and spirit.  There is just no way a primitive chatbot can do this stuff, right? They can’t even understand what we’re saying, and they definitely can’t do anything but parrot things we’ve already told them. But what if they suddenly could seemingly understand and take orders?

The projected market distribution across several industries for chatbot usage.

Rapid advancements in artificial intelligence and natural language processing (technically called large language models, or LLMs) now allows chatbots to understand context, learn from past interactions, and provide more human-like responses in every language on the internet, including computer code, which allows them to actually do things that people at a computer can do — AKA robotic process automation. The one chatbot right now that can do all this well is chatGPT. This doesn’t make chatbots just chatbots anymore, it makes them robotic process automation applications with conversational UI — complete Digital Assistants (DAs). You tell it what to do, and it does it, even if you don’t know how to do it.

The potential for ChatGPT and other DAs is mind-blowing. The technology is so good, “conversational UI” could be ubiquitous in just a few years. Why have various input controls, navigational components, and associated containers when you can just provide an input for the user to tell you what they want to do or see? As the technology continues to advance, DAs will become even more sophisticated and capable of handling increasingly complex input, like your speech, sequenced directives, and camera-based vision. They will be a more integral part of businesses’ general operations and even likely become more human-like, with the ability to understand emotions, grasp humor, and build relationships with users based on their individual needs and preferences.

Google Trends data on “chatbot” searches, beginning with the launch of ChatGPT.

Chatbots are in favor

Sure, DAs can’t do everything perfectly, yet, but neither can people. Plus, like a fine wine, it’s explicitly designed — almost guaranteed — to improve over time (there are at least two new versions slated for this year). That’s better than some people we may know. The rise of DAs will have a harsh hangover effect on the business world. DAs will have become a valuable tool for businesses looking to actually improve customer experience, and even streamline internal operations.

ChatGPT can, or eventually will, be used for a variety of purposes, and not just sales-side jobs (customer support, lead generation, sales, marketing, and internal communication) but also production-side jobs (translation project management, vendor management, quality control and almost any internal management function within an agency). In fact, if you google “ChatGPT” and “translations” the results will include an automated ChatGPT-created and ChatGPT-run language company. ChatGPT is coming for the organizational guts of the language service providers (LSPs), not translators. Why? Because somebody will always have to be held accountable for the actual service (translations) — that’s where the ultimate vendor value is. LSPs offer value-added services that are now automatable, thanks to DAs. Why haven’t templates, automations, and chatbots completely replaced lawyers? Because somebody has to be held accountable for the actual legal documents and decisions in court. (Accountants, ironically, are not really held accountable, and are only filling out government forms, so I think they’ll likely be replaced.)

In this new wave of automated online businesses, if tools like ChatGPT can write advertising to lure prospective buyers, write website-marketing content, chat to buyers about needs, log quotes and projects, translate content, follow up with editors, modify content based on feedback, write emails and follow up with clients, then where does that leave an LSP and other agency-based businesses? Quality control or — in the worst case scenario of having to deal with an irate individual — customer service. The same person who created an autonomous website company years ago is the same person who has to maintain it today. But now they don’t need to know how to code, just monitor, prompt, and check if the bots are working to correct and prevent issues. That person, probably a linguist for our information arbitrage-based industry, needs the most sophisticated and advanced data dashboard in existence to track the number of data points going through their LSP, and our company focus at Zin Global is to build that dashboard.

What will happen to all the jobs in our industry? I think it’s safe to say that by allowing our translations to eventually be found online by LLMs, we’re training the next version of them to do our jobs for us. Interpreters in general are fine for now (give it a few years). Translators will end up doing more and more machine translation post-editing. LSPs that can afford to invest in building chatGPT tools, will — just to survive. Hopefully, there will be a marketplace for tools that help anyone build a chatGPT services website, so that all LSPs can enjoy the innovation.

Generally, technological innovation that displaces workers in one area of the economy also creates employment growth, but in a totally different area of the economy — engineering (you know, for making and maintaining the bots that took their job in the first place). In the past, this required retraining in computer programming, but ironically, these jobs are the ones most at risk of being replaced by bots themselves. So translators in our industry are going to need to become a bit of a coder to build out the tools they need, become versed in prompt engineering, monitor their dashboard, and check to make sure chatGPT did its job right.

I get some solace in the understanding that the language industry is not alone in this transition. Already, according to the ELIS 2022 Language Industry Survey, about one in three translators don’t earn enough as a freelancer — they’re technically part of the “gig” economy. Are we doomed to give way to a more sophisticated version of ourselves? Will we have to find another gig or industry? There are many times that technology has shook the foundations of a particular industry, for better and sometimes for worse, but one thing is certain: It has always provided new opportunities for advancement, revenue and learning for those who embraced it as early adopters. Maybe we need to adapt? As automation technology marches forward, we will be relegated to hiding behind the chatbot hoping not to be on the receiving end of an irate customer wanting us, the “next available agent.”

You’d likely be hard pressed to find which parts of this article were not written by ChatGPT. Truth be told, the article was authored by ChatGPT… partly… maybe.

Advertisement