From ChatGPT taking the world by storm, to perhaps taking over the world altogether (so say the AI developers themselves), should we be worried? Or excited? At least a little bit curious…
So, what are the benefits and limitations of Generative AI models like ChatGPT? How might we work with them as we ride the wave of changing times? Here at Dialogue, we’ve pulled together some helpful pointers for you to see just what AI & ChatGPT can do well and where there is room for development:
Machine Translation (MT) can process large volumes of text quickly and in multiple languages – ideal for time-sensitive projects. With a human post-editor, plenty of reliable MT engines such as DeepL, Microsoft Translator & Phrase NextMT can already be used for non-complex texts.
MT is not to be used for legal, medical or really creative texts – it struggles to preserve nuance, accuracy, cultural references and idiomatic expressions. ChatGPT is not yet able to translate accurately and without mistakes; instead, its strength lies in copy generation. It is, however, able to translate speech reasonably well. This is because ChatGPT has access to a wealth of speeches recorded and written online and is designed to imitate human speech in its answers.
ChatGPT can generate creative copy and marketing content from scratch in different languages without making mistakes. It is already a good tool for a linguist to use to help them come up with different creative ideas and options during the transcreation process; they simply need to feed and refine ChatGPT with the appropriate prompts.
The marketing copy it can generate still needs a human touch to add more spark to the generated text. In the future, machine transcreation with human ‘post-creation’ may well be a service – let’s see!
One-directional machine interpreting is great for getting the basic gist of speech at events or meetings where you just need a translation of what the speaker is saying.
It’s not so good for conversations or sensitive business meetings, as AI cannot facilitate conversation in the same way that human interpreters can. Check out this video to see AI interpreting in action! Machine interpreting is also not to be used in medical or legal contexts.
AI is really good at speeding up subtitling, transcription and voice-over processes and it already provides a valid, cheaper alternative. Automated, AI-generated services are just the job for any internal content, training & meetings.
Automated services are not good for creative films, games, ads, and marketing content for an external target audience. AI might speed up the transcription process, but it can’t create the style, humour and creative cultural input needed in these contexts.
Duolingo has teamed up with GPT-4 to create a language learning app where AI can correct the learner’s grammar and speech. The metaverse also allows you to learn the basics of a language in a creative, virtual reality!
Without human feedback, training & interaction you can only go so far with your language learning; who do you ask if you don’t understand something? What if it teaches you the wrong thing…?
The tech research giant, Gartner, reckons that Generative AI’s real impact will be really understood within the next three to six years; predicting that by 2025, 30% of marketing content will be created by generative AI with post human editing. So, let’s watch this space!
If you’d like to know more about the tech services we already offer, you can download this 1-page guide detailing our language & tech options to date. We’re watching this AI trend with interest and are adapting our services as we speak, whilst always ensuring that human touch.
We’d love to hear your thoughts on how you may be putting AI to use in your company – or not – and would be happy to chat through the language options available to you in this fast-paced, digital era! So, don’t hestitate to get in touch!