Machine learning is taking over, but there are drawbacks attached
Machine learning has joined artificial intelligence (AI) as a buzzword across various industries. The onset of big data has companies analyzing information like they never have before. Machine learning can be categorized into three parts: data processing, model building and deployment & monitoring. According to TechCrunch, the model building is where the machine learning happens—an algorithm actually learns from training data to predict given input data. An examples of machine learning of you may be familiar with is Apple’s virtual assistant Siri, which approximately 700 million iPhone users have access to. Siri uses voice recognition and AI to respond to the user’s commands. Machine learning is also image tagging, such as Facebook’s feature of knowing your friend’s faces when you upload photos.
Machine learning is currently disrupting transportation. Nearly 10 million self-driving cars are expected to be cruising our streets by 2020. Car manufacturers like Tesla have been testing prototypes for a few years, and Google’s Waymo is working toward self-driving vehicles that “ make it safe and easy for everyone to get around.” In March 2017, Bernard Soriano, deputy director of the Department of Motor Vehicles told The Guardian the technology itself will perform better than actual drivers. “We needed to provide a clear path to completely driverless vehicles, because of the safety benefits.”
The drawbacks to self-driving vehicles are many. The technology to develop the machine learning is costly, so it could be awhile before self-driving vehicles are affordable to the majority of consumers. Security could also become a major threat in regards to hackers or just operating the vehicle for individuals that are not so tech savvy. And this, of course, is on top of the obvious skepticism around accident protection.
Machine learning is gaining traction in sales, marketing and communication. The technology often helps us communicate, such as that with wireless communications. It also helps us avoid each other — another classic example being spam filtering, which learns to classify an email as spam or something you’d like to read, based on the sender or subject. Companies are using machine learning tactics to get to know and communicate with their customers. The technology has proved to be so useful in sales forecasting that major tech players like Salesforce are investing in machine learning.
Another marketing and communication trend adopting machine learning are chatbots. Chatbots are computer programs that carry out conversations with people using lightweight messaging, but they often fail with localization strategies. Reason being, machine learning and AI can’t always predict where someone is from or what they will say. Chatbot acceptance is growing but 56 percent of customers still say they prefer to speak with a person over bot. This ends up being a major drawback of machine learning in communication. With many businesses looking to expand into international markets, it’s important to fully know your customer and be able to communicate with them. It’s because of this that many companies still use human translators for localization strategies because machine learning doesn’t always localize information correctly.
In fall 2016, Google introduced a new system for machine-assisted language translations, which takes advantage of deep neural networks to improve the quality of translations.
Since then, there have been multiple claims that this new and improved system will leave human translators out of work. And in many respects, NMT represents a significant improvement. However, contrary to recent publications, it still hasn’t closed the gap with human translation. In fact, in many instances, neural machine translation still makes significant errors that a human translator would never make, such as in the case of unusual words or phrases.
However, buzz surrounding machine learning and its application to translation isn’t completely unjustified. In fact, it highlights the tremendous potential for the integration of human and machine, promising a great future for both human translators and translation buyers alike. With tools like translation memory and other state-of-the-art translation technology platforms, machines has empowered translators and buyers with efficient, low-cost and high quality output.
Go global with Gengo’s people-powered translation platform.
or Contact us
Main image by Menno van Dijk via iStock.