Why AI needs more language thinkers
| 5 Min Read
AI can now write code from a natural-language instruction. It can translate, summarise and answer questions in seconds. As it takes on more tasks that once required specialist knowledge, what skills should people build for tomorrow?
For National Coding Week, we spoke with Dr Tony Russell-Rose, who has been involved in developing the online MSc Computational Linguistics at Goldsmiths, University of London. His work spans natural language processing (NLP), information retrieval and human interaction with information. We discussed how NLP predates GPTs, how AI is changing coding, and why problem-solving, human needs, cultural understanding and language preservation belong in this conversation.
What skills should humans learn as AI advances?
Tony describes software development as a value chain. It runs from the mechanistic work of producing code to identifying the problem, understanding the human need and designing a solution.
AI coding tools can already handle much of the lower-level work. Tony sees this as democratising because people from different disciplines can turn ideas into working products.
Tony says: “The skill is understanding what to build. That’s partly a problem-solving activity and partly about understanding the human need we are trying to meet.”
People will need to decide when to delegate, when to intervene and when a convincing output needs to be questioned.
Are prompts becoming a new form of coding?
We increasingly tell machines what we want through everyday language. Tony agrees that prompting can be seen as a form of coding, up to a point. It queries an existing model and can quickly produce a useful first version.
How is coding different from prompting?
Prompting works with an existing model. Fine-tuning adapts it for specialised behaviour. Training builds or changes it at a deeper level and requires significant expertise and computing power.
Tony explains: “If you want to build something lasting, scalable and secure, you need to go further.”
Coding and software engineering address what happens beyond the first output. As language becomes a common way to instruct technology, understanding how language works may itself become part of technical literacy.
Did GPTs start the NLP revolution?
NLP has existed since the 1960s. Tony recalls how he used to introduce the field:
“I’ve been doing this for 20 years and for the first 19 years, nothing happened.”
Around 100 people worked in the UK field at the time, he recalls, and he knew almost all of them. Today, the field is vastly larger.
A milestone came in 2013, when Tomas Mikolov and colleagues introduced word2vec, a way of representing relationships between words mathematically. In 2017, Attention Is All You Need introduced the Transformer architecture that became foundational to modern large language models.
Tools once limited to specialists are now available to millions, creating opportunities for creativity and social good alongside new responsibilities.
Can computational linguistics help keep more languages alive?
AI can translate text, recognise speech and help people communicate across languages. These capabilities are not developing equally for every language.
Language technologies learn from data, so languages widely represented online give systems more material to work with. Context, cultural references and emotional meanings may still be lost in translation.
Tony believes computational linguistics can help preserve and promote languages and dialects. Technology can’t keep a language alive by archiving it, however. Languages live when people use them.
“There’s a danger that we’ll all become homogenised. AI and computational linguistics can play a role not just in preserving culture through language and dialect, but also in promoting it and ensuring that it stays alive.”
That means developing language technologies which work for more communities, and people with the technical, linguistic and critical skills to question whose voices those systems recognise.
Why do AI and ethics belong together?
For Tony, ethical judgement is integral to technical education. Developers must consider who a system serves, who could be excluded and what its consequences might be.
At Goldsmiths, he says, that responsibility isn’t treated as a standalone ethical exercise. It runs through the technical work.
To capture the distinction, he says, “Not just: are we doing things right? But: are we doing the right thing?”
This approach is reflected in the online MSc Computational Linguistics curriculum. It combines linguistics, Python programming, machine learning, NLP, interaction science and large language models. Students also examine fairness, bias, privacy, interpretability and responsible innovation.
Why study computational linguistics now?
Tony sees growing potential for creativity and social good in computational linguistics. The course develops technical competence, critical thinking, responsibility and an awareness of social justice.
The programme welcomes computer scientists interested in language and AI, along with linguistics, humanities and social-science graduates seeking technical skills. Tony says many students combine an existing interest in language with new computational capabilities. Students come from a range of disciplinary backgrounds.
Through project-based learning, students analyse language data, work with NLP and machine learning, and critically evaluate emerging technologies.
Tony describes the purpose of the programme as an opportunity to do something “interesting, exciting, worthwhile and responsible.”
How will language shape our interactions with machines?
Tony expects language to become even more important as AI develops. Future systems could take physical forms as companions, support systems or co-workers in unsafe environments. Many will need to communicate with us through language.
His conclusion is clear: “Computational linguistics will have a huge role to play.”
AI may produce the first line of code or answer. People must still decide what to ask, what to build, how to assess the result and how technology should respond to human needs. That is why AI needs more language thinkers.
Learn more about the online MSc Computational Linguistics at Goldsmiths.
AI in 2030: What do humans still need to be good at?
This October, our AI programme convenors discussed the skills you’ll need for tomorrow in the age of AI. The webinar explored how technical capability is changing, and why judgement, problem-solving, critical thinking, communication and responsible use of AI will continue to matter. The discussion brought together perspectives from AI & Data Analytics, Computational Linguistics, AI & Machine Learning and AI & Creative Practice.
#SkillsForTomorrow. Shaped @Goldsmiths.