ChatGPT is the biggest trend in technology right now, making waves across every software category and industry (even more so after the announcement of its integration with Bing). The language services industry is no different either, with the AI chatbot building its reputation as a potential translation tool to rival the likes of Google Translate.
In this article, we look at the pros and cons of ChatGPT as a translation tool, and whether it could rival the likes of Google.
What is ChatGPT?
ChatGPT is an AI conversational system developed by OpenAI, designed to engage in human-like conversations. Essentially, it is a chatbot that uses a huge database that allows it to formulate answers with real-world information.
The tool’s main interface lists three example questions:
- “Explain quantum computing in simple terms”
- “Got any creative ideas for a 10 year old’s birthday?”
- “How do I make an HTTP request in JavaScript?”
ChatGPT is designed to handle technical questions and provide clear, intelligible answers. Here’s what happens when asking ChatGPT to provide a definition of itself:
“ChatGPT is a conversational language model developed by OpenAI. It is built on top of the GPT (Generative Pre-trained Transformer) architecture and has been fine-tuned on a massive amount of conversational data to generate human-like responses to a wide range of topics and questions. ChatGPT can engage in casual conversations, provide information, and assist with various tasks, such as answering questions, translating text, and generating creative content.”
Interestingly, ChatGPT includes “translating text” as one of its potential use cases. Given all the hype this tool is generating at the moment, it is no surprise that language professionals have been putting it through its paces to see how good of a translation tool it really is.
How good is ChatGPT at translating text?
As things stand, ChatGPT is a text-only system that requires users to input a text prompt, for which it generates a text response. This is significant because a tool like the Google Translate mobile app can translate live speech and text in images through its Lens technology.
So, when travelling abroad and coming across a menu that needs translating, Google Translate will be the more effective tool to use.
However, if we limit the comparison to just translating text then this bears the following question: How does ChatGPT stack up against Google’s most advanced AI translation tool?
Let’s look at some examples.
ChatGPT vs Google Translate
We randomly picked a tweet from a non-English account to test ChatGPT and Google Translate. The tweet in question came from the Spanish news publication El PAÍS, in reference to the 2023 Premios Goya – the country’s most prestigious award ceremony for cinema and arts.
To get this tweet translated via ChatGPT, we copied and pasted the text into the following prompt:
“Translate ‘Descubre todas las actuaciones musicales, horarios, estrellas que entregan galardones, favoritos y un concurso: una guía de la 37ª edición de los Goya’ into English”
This is the response we got from ChatGPT:
“Discover all the musical performances, schedules, stars that award prizes, favorites, and a contest: a guide to the 37th edition of the Goya Awards.”
For the Google Translate test, we simply pasted the Spanish text into the translation tool – no prompt required. Moreover, this was Google Translate’s output, in English:
“Discover all the musical performances, schedules, award-winning stars, favorites and a contest: a guide to the 37th edition of the Goya Awards”
The responses are very similar when looking at the two outputs, but there is one key difference: ChatGPT translated the verb entregar as to award and Google Translate has translated this as to win or to receive.
ChatGPT is correct here with the official translations for entregar including: to hand over, to give, to present and to give out.
For better context, the El PAÍS tweet was posted before the event so the winners were unknown at that stage and time of the tweet. The article it linked to included the stars who would be presenting the awards for each category, as ChatGPT correctly interpreted.
So, in this particular example, the translation “award-winning stars” provided by Google Translate is not correct.
The only issue with this particular ChatGPT translation is that it reads a little awkwardly and the “stars that award prizes” lacks clarity.
For a more comprehensive breakdown of ChatGPT vs Google Translate, take a look at this excellent article from Metaroids.com, which includes several more examples.
Is ChatGPT better than Google Translate?
Early tests with ChatGPT suggest it is better at translating content into English than the reverse. Like most AI tools, it is also better at translating from some languages than others.
Going in the other direction, ChatGPT runs into some common problems. For example, when asked “How do I say ‘I’m hot’ in Spanish?” it makes the same mistake as Google Translate by suggesting “estoy caliente,” which actually means something of a far more erotic nature.
The correct answer, in this case, would be “tengo calor,” which directly translates as “I have heat” in English.
Google, on the other hand, has put a lot of training into specific languages, which shows when dealing with more complex language combinations – e.g. English to Chinese, Chinese to Vietnamese, Vietnamese to French, English to Greek etc.
This article from Makeuseof.com provides some interesting examples.
Either way, ChatGPT does an impressive job of translating text, and a better job than Google Translate in many scenarios – especially when translating from widely spoken languages like Spanish and German into English.
What does this mean for the translation industry?
AI translation is already a tool that is widely used by language professionals to do their job faster and to create the most cost-effective translation processes and workflows.
Any advances made in AI translation technology only partially benefit the language and translation industry, because if language professionals obtain the necessary knowledge in order to know when and how to use such tools properly (for example by always implementing the required review and editorial process to fix output mistakes), then quality and speed are guaranteed.
If you want to learn more about how to make the best use of language translation technology and of AI translation tools, get in touch with one of our experts today.