SPECIAL FEATURE: AI IN T&I, PART 1
Earlier this year both NAATI and FIT published position statements on the use of artificial intelligence (AI) in translation and interpreting:
Many more AI tools have been launched since we published Sam Berner’s article on ChatGPT and translation in our Spring 2023 issue (read it here), so we’ve asked practitioners, academics and researchers to tell us more about the topic. In Part 1, Dr Anna Gadd has compiled an overview of AI in translation; practitioner Nora Díaz explains how and when she regularly uses ChatGPT in her translation – and in her interpreting; and academic and researcher Michael Farrell reports on his recent research into translators’ attitudes to using generative AI.
As Director of Graduate Translation Studies at the University of Western Australia, Dr Anna Gadd is following the growing use of artificial intelligence (AI) in translation closely. She describes seven of the most popular AI models for us below.
… each model functions differently and has different strengths and weaknesses.
In 2025, translators have a wide range of ‘technologies’ at their disposal, some of which are computer-assisted translation (CAT) tools, while others are either AI proper or a combination of the two.
Second language learners, students and translators each have their preferred AI tools, chosen to suit the language pair(s), sector(s) and genre(s) that they work with, and there can be ample margins of difference in functionality, depending on these factors.
Different AI iterations are referred to as ‘models,’ and each model functions in its own way and has specific strengths and weaknesses.
Below is a list – by no means exhaustive – of some of the most common technologies available in Australia as I write. However, it is important to know that technologies evolve at an incredibly fast pace; by the time this article is published, therefore, it is likely that these will have improved, and new ones will have appeared on the market.
ChatGPT
When AI is mentioned, most people think of ChatGPT: the most well-known and widespread model out there, publicised by the launch of Sam Altman’s free research preview OpenAI at the end of 2022. ChatGPT – found at chatgpt.com – is a Large Language Model (LLM) which can translate from and into a large number of languages by means of a ‘(please) translate into xyz language’ prompt. This means you can tell the technology to edit the translation depending on your preference, and teach it how to translate in real time. It is being used daily by a staggering number of users, and is fast improving – supposedly – as each individual user prompts it towards their own perceived correctness. ChatGPT gives you the chance to dive into the nitty-gritty of translation and really tailor your target text to your own needs and/or your client’s requirements. As ChatGPT’s AI is constantly learning, the results for the same request will differ over time. As a user you can interact with the technology and use your own knowledge to help it learn, by guiding it towards your desired results, based on your own expertise. There are two main models of ChatGPT: a free version with public cloud storage, and a paid version (ChatGPT Plus) which offers the option of creating your own private cloud storage. Anything we upload onto the free version of ChatGPT is kept by OpenAI in the public cloud.
Gemini
After the launch of OpenAI and as a direct response to ChatGPT, Google invented Gemini – based on the LLM of the same name – in 2023. Gemini – found at Gemini.google.com – has very similar functions to ChatGPT. Users can provide feedback, and can even do so quickly via ‘thumbs up’ and ‘thumbs down’ gestures; and in theory, each response could improve the accuracy of following translations.
Quillbot
Quillbot – found at quillbot.com – is a lesser-known AI translator which can conveniently be added directly to the Google Chrome browser for free. You can input your text and compare source and target texts. It has a character limit of 5,000, and was created in 2017 by Anil Jason as ‘a way to help students still learning English improve their writing.’ Quillbot has an ‘AI humanizer’ – software that apparently makes AI-generated content sound and ‘feel’ more human.
Google Translate (newest version)
First developed by Google in 2006 as a statistical machine translation tool, Google Translate was updated to a neural machine translation (NMT) tool in 2016 – a drastic improvement in its capability. The current version of Google Translate has a large number of source and target languages, and a 5,000-character limit. When you hover over the target text with your cursor, Google Translate will highlight the same passage in the original, which allows users an easier comparison of source and target texts.
Microsoft Word (newest version)
In 2025, Microsoft Word is offering an in-suite translator function built directly into the processor and easily usable via a button. It can be used for direct translation between a number of languages including English, Italian, Spanish, French, German, Indonesian, Korean, Japanese, Chinese, Arabic and Portuguese. When a user clicks on this function, Word will output a new file containing the direct translation.
DeepL
Created in 2017, DeepL boasts that it is ‘the world’s most accurate translator.’ It has a lot of similar functions to Google Translate, but it builds on them. With DeepL, users can click on a segment to see other translatants (translation choices for specific words) and alternative translation options, and they can also create a summary translation (a useful tool for learners of modern languages). The free version of DeepL has a 3,000-character limit.
Trados
SDL Trados Studio was launched in Germany in 1984 as a language service provider (LSP); Trados stands for TRAnslation and DOcumentation Software. Trados GmbH was initially founded by Hummel and Knyphausen to complete a translation project for IBM. Over the decades SDL Trados has consistently been the industry standard CAT tool that can be used across several countries. Its suite is the most comprehensive on the market, and it now features inbuilt AI: Copilot, which is Microsoft’s AI model. Trados boasts some other great CAT features aside from AI, such as a translation memory (TM) and term bases (TBs). TMs are a key feature of CAT tools pre-AI which allow us to reuse previously completed translations, and to see how a certain word was translated by others working on the same project, while TBs enhance accuracy when translating sector-specific terms, and are especially useful when working on large sector projects, as occurs in the localisation industry. As Trados has been the leading CAT tool in the industry, and has now incorporated AI, it is very likely that it will retain its leadership in coming years.
Dr Anna Gadd is an award-winning language lecturer and translator with experience spanning three countries and nearly three decades. Dr Gadd launched the Spanish Department at the University of Western Australia (UWA) in 2017, and is currently UWA’s Director of Graduate Translation Studies and Convenor of Spanish Studies. She has also published extensively on translation studies and second language acquisition.
