The translation industry is not immune to the "invasion" of AI and other technologies. We are at a point where machine translation technologies are starting to produce acceptable quality work. While this is beneficial for optimizing budgets and growing business, there is still room for skepticism.
Machines are improving every day, but human intervention is still necessary to evaluate the viability of AI translations. This allows us to move away from a dependency on human translation and find a middle ground. The future of the translation industry belongs to those that can leverage AI translation at scale without sacrificing translation quality.
Where Does ChatGPT Come In?
ChatGPT has quickly become a source of interest in the business world, with its ability to generate anything from art to essays to scripts based on short questions and descriptions.
In translation, ChatGPT is a type of large language model (LLM) that uses machine learning to generate translations based on input from a user. This is not entirely opposite of Neural Machine Translations (NMTs), which most translation providers use as an AI tool to accelerate and automate translation jobs, but it is slightly different.
The accomplishments of the team at OpenAI, both for the technology itself and their rapid global expansion, are admittedly remarkable. Individuals and businesses alike are now trying to figure out how they can put this new super-tool to work to boost productivity and efficiency. Like Google's intuitive AI functions, ChatGPT also can translate content into other languages. But just like ChatGPT has garnered attention for providing false information and statistics, can you know for sure if its translations are correct?
Pros and Advantages: How is ChatGPT Useful for Translations?
The greatest benefit of using ChatGPT is its fast turnaround times. Human translation, even in tandem with machine translation, can take hours depending on the complexity of the project. ChatGPT translation occurs in a matter of seconds, even if it does contain inaccuracies.
ChatGPT is a free AI service that is accessible to anyone, individual or business. It's no secret that human translation comes with a cost, but you should be prioritizing quality over price when it comes to translating your website. Though ChatGPT is free, the cost of inaccurate and poor-quality translations could translate into lost business. It also means you will probably have to pay to correct your mistakes down the road. In the end, paying for high-quality work from a concierge-level translation partner will pay off more for your business
ChatGPT is affordable and easy to use. It produces comprehensive, mostly correct, and useful answers to any range of questions. Its user interface is simple to use and has catalyzed the mainstream adoption of AI everywhere. It's a big reasons why businesses are opting to skip professional translation agencies and rely on these AI based tools for translation.
Cons and Disadvantages: What are ChatGPT’s Greatest Flaws?
Translation quality. ChatGPT is an advanced free translation tool, but if you're looking for better quality at the same speed and price, Google Translate's NMT is still much better. ChatGPT does have the ability to produce translations, but despite its strengths, it is not a bigger or better solution than any NMT. While both NMTs and LLMs are able to generate a response to a prompt, the different training they each receive leads to different types of responses. So, where do LLMs like ChatGPT falter when it comes to translations?
1. It Isn’t Customizable
NMTs are trained and optimized to be predictive and accurate. This can be further enhanced through algorithmic learning and adaptive strategies. This helps NMT to read and translate content at a holistic sentence level, providing more accurate and personalized translations. It is unclear how a shared generic AI model can be adapted yet.
While ChatGPT can do translation at a quality level that is comparable yet inferior to Generic NMT models (e.g. Google Translate), it can’t compete with a well-trained brand-adaptive NMT, which will produce a better translation quality, much faster and much cheaper.
At MotionPoint, we have been training brand-adaptive NMT models for our customers for some time now. The AI learns from past translations to help customize things like word use, sentence structure, and intent. Within AI, our NMTs enable additional deep-learning and translation quality benefits like improved efficiency and scalability of website localization efforts and greater accessibility to a global audience. All of this has helped our customers experience very significant increases in the quality of output in comparison with the generic models that we started with.
2. It Lacks the Ability to Show Logic
ChatGPT's ability to understand context is limited. Hence, the model fails to answer many questions that a human might be able to. Take the example provided below. Though the question is posed as a riddle, it would be clear to any reader that there is a father named Bob, a son named Jay, and a son named John. The ultimate question is this: Who is Jay's brother? The answer is clearly stated in the first sentence, but ChatGPT lacks the basic ability to decipher the answer because it becomes confused as the riddle reads on.
This poses an issue for translation because not all languages translate word-for-word. If this tool lacks the ability to apply logic to a problem, it won't be able to consider structure, grammar, or syntax when switching from one language to another.
3. It Lacks Security
Even if you could use ChatGPT as a website translation tool, it neglects a very important part of translation: security. Simple AI technologies like ChatGPT cannot be used for certain industries where data security and privacy are of paramount importance, such as healthcare and hospitals, hospitality, and financial service companies.
When partnering with a translation partner like MotionPoint, we update you on how we complete our annual security assessments conducted by independent PCI SSC Qualified Security Assessors. We demonstrate ongoing practices that comply with PCI DSS and complete regular independent assessments to ensure they comply with HIPAA Privacy and Security rules, too. Failure to comply with some of these regulations can open a business to fines and lawsuits.
Here’s the Bottom Line
The future of translation belongs to those that blend with AI and offer flexibility on costs without sacrificing quality, by translating less content with humans which typically costs the most. ChatGPT has quickly become a groundbreaking AI platform that touts amazing new uses for technology. While it may not be the solution for translation itself, it boasts great opportunity for other aspects of the field. Where the LLM tech can be of use is regarding the various tasks around the edges of translation and localization, including how we prepare training data to improve the quality of output of NMTs.
MotionPoint’s suite of technologies is the answer to getting the highest quality translations at the right price. Our Adaptive Translation™ combines translation memory, AI, and algorithmic translation to produce human-quality translations while saving you up to 60%! Don’t fall behind on your translations; take advantage of all AI (and MotionPoint) has to offer.Last updated on April 24, 2023