It is common to say that the Inuit people have 50 words for snow. This is for many reasons a tricky statement, but above all the very idea of reducing human experience to a number is problematic. Human language is an expression of the unique culture, experience and society it represents; ideally a way to communicate and connect distinct experiences through mutual understanding. Nevertheless, throughout history, we have developed technologies to assist human communication by reducing the complexity of human experience into comprehensible form — data, numbers and systems. These technologies were deployed with many different purposes — some more genuinely seeking to create human bonds across distinct cultural experiences, others with less altruistic objectives. 

With the latest peak of language technology, AI and Natural Language Processing (NLPs), such as the Large Language Models (LLMs) and their chatbots, OpenAI’s ChatGPT, Google’s Gemini, Anthropic’s Claude and Meta’s LLaMA and Microsoft’s co-pilot, the human identity of communication is transforming. This raises questions — with what purpose are these new language technologies introduced into our lives? On what ideas of human communication are they based? Whose interests do they serve? 

The Global AI Race for Domination

There are many creative ways of supporting humans with language technology, including the ones based on AI. However, the AI language technologies dominating our everyday lives today reflect a ruthless competition between powerful states and industry conglomerates in the pursuit of dominance, profit and scaling for the sake of scaling. The recent story about a Chinese chatbot, DeepSeek, challenging the current US favourite, ChatGPT, illustrates this well. That is; the AI models leading the race might be effective and profitable at this very moment in time, but they have huge flaws, like intellectual property infringement, privacy invasion and censorship that have never been tackled.  They submit to the rules and parameters of the current geo-political AI race, which is mirrored in their design. 

As a stark warning, we should look to the history of colonial powers’ use of language technologies to convert the local human life, culture and complexity of the people living under their rule. During European colonialism that expanded across the world in the pursuit of world domination for centuries, local languages were at first sought to be repressed. Later for administrative and managerial purposes it was considered beneficial to educate the colonised populations. Part of this process involved religious conversion, which required translating the Bible into local languages. However, the language technologies of the time, such as dictionaries and grammars, reframed the Indigenous languages through European linguistics. The nuances of local languages were trimmed down, as were lives and cultures to fit the purpose of social control. 

The cultural impact was profound. Colonised people were denied not only their traditions, language and culture, they were denied everything that made them human as the Carribean-French author and activist Frantz Omar Fanon described it in 1961 in his book “The Wretched of the Earth“. As I quote in my new book “Human Power – Seven Traits for the Politics of the AI Machine Age,”¹ Fannon wrote: 

“(…) centuries will be needed to humanize this world which has been forced down to animal level by imperial powers.”²

AI News for Global Citizens

Imminent Readings

Eager to know more? Here, you will find a selection of articles from top newspapers, research publications, and leading magazines from around the world, exploring AI’s impact on language, culture, geopolitics, and economies.

Dive deeper

Human Power in the AI Machine Age

In “Human Power” I invite readers to reflect on the delicate balance between humans and the machine technologies shaping our world today. Exploring seven distinct traits of human power — Creativity, Intuition, Emotion, Life, Defiance, Love, and Wisdom — I argue that our human potential and characteristics must remain central not only in the politics of policymakers and regulators, but also in our own everyday lives — when we interact, relate and communicate with other people. 

Digital spelling and grammar correction tools and machine translation systems have been household items of every human communicator’s life for decades. We’ve learned to incorporate these tools in our everyday interactions with other people without much friction. But recently, generative AI systems that combine NLP with Machine Learning, and Deep Learning have swirled into the human communications sphere changing the rules of the game. And we have embraced these tools like no other language technologies in history.

In August 2024, ChatGPT alone was reported to have 200 million weekly active users.³ Professionals use generative AI for email writing, summarising content, report writing and marketing material. Students use it for essay writing, answering questions and researching topics. Increasingly, we see the polite, sterile language of the perfectly polished AI generated social media posts in our daily feeds.

Increasingly, we see the polite, sterile language of the perfectly polished AI generated social media posts in our daily feeds.

The NLPs and LLMs of the conversational AI bots — such as ChatGPT — function by converting human language into machine-readable data, analysing it, and generating new text. They discern patterns in the language data they are trained on, predict word sequences, count words, assign weights to words and relate them to others. They encode and decode words. They imitate human communication but compared to the complexity of human language and communication, these machine languages remain one-dimensional.  They are mechanistic, prioritising efficiency and system optimisation. 

Language, Culture and Power

In the 1990s British Jamaican cultural critic and sociologist Stuart Hall illustrated how dominant cultural powers are practiced and enacted through language systems that are “encoded” so to speak, with a “…dominant cultural order…”. As Hall wrote: 

“These codes are the means by which power and ideology are made to signify in particular discourses. They refer signs to the ‘maps of meaning’ into which any culture is classified; and those ‘maps of social reality’ have the whole range of social meanings, practices, and usages, power and interest ‘written in’ to them.” ⁴

We can already see the cultural identity of a global AI race encoded in the leading AI language technologies of today and the impact it has on human culture in fundamental  ways:

The first casualty of the “scaling for the sake of scaling” and for-profit business model behind current AI language technologies is the diversity of human language cultures. While LLMs can be used to promote and document Indigenous languages around the world, they primarily support widely spoken languages, leaving smaller less profitable ones behind.⁵ These systems excel in common tasks like search, email, and social media, and thus Indigenous communities increasingly adopt dominant languages for daily communication. This further marginalizes their native languages, leading to language loss.⁶ As indigenous language scholar Hēmi Whaanga warns: 

“With homogenization comes loss. It has been suggested that by the end of this century at least 50% of the world’s languages will face the prospect of death. Many if not a majority of these languages will unfortunately be Indigenous languages. When we lose a language, we lose the conduit to our linguistic and cultural ecosystem. If we lose those ecosystems, we lose our identity, our history, our culture, and ultimately, we lose our power. With the increase in the probability of this homogenization, will AI accelerate this change, this loss?” ⁷

Yet, there is another concern of a more general nature regarding these technologies’ impact on human culture that requires urgent attention. In Human Power, I argue that while we should not fear the extinction of the human species by artificial general intelligence (AGI) and killer robots as many prominent business people and scientists have warned lately,⁸ there are other ways of stagnating a species’ evolution.⁹ Consider the impact of AI on our human powers — our creativity, unpredictability, intuition and complexity. These qualities do not compute in any frictionless manner with the rigid and reductionist culture of AI machine language technologies designed for profits and optimisation.  

For instance, the cultural framework of the current NLPs is that it is always “complete”. There is no room for ambiguity when the system processes and represents human language; all data must fit into the classifications and categories of the model. Into the systems of meaning assigned to specific words, the priorities and classification of the system. But these systems are never “complete”, they do not represent the complete picture.¹⁰ Take a human’s age as an example. We are not just ‘old’ or ‘young’, we are ageing gradually.¹¹  

These systems excel in common tasks like search, email, and social media, and thus Indigenous communities increasingly adopt dominant languages for daily communication. This further marginalizes their native languages, leading to language loss.

Biases also creep into these machine-based language cultures often perpetuating harmful stereotypes. For example, a group of computer scientists found that word clusters created by commonly used machine learning models for online search engines, grouped titles like architect and financier as “extreme he” words, while words like receptionist and nanny were categorized as “extreme she.” ¹²  But they are much harder to detect when the model operates in a “complete” data structure, in which no meaning can be found outside the system. 

A Machine-to-Machine Love Hate Relationship

There is an ongoing battle between a culture of machines and a culture of humanity that needs urgent attention. However, the fear of missing out of the global AI race does not allow for much human contemplation. It may be our own demise if we do not take the time to make this right. We need to value and prioritise human culture from the early stage of development and deployment of AI language technologies, or human powers will be reduced. The communication between human beings is already transforming into one between machines with exchanges of perfectly sounding, flawless messages and phrases. Teachers are receiving impeccable essays written by AI from students, some are surely already using AI tools to screen and suggest grades on the same essays. We are prompted to “rewrite with AI” our helpless human social media post and messages turning feeds into unending simulated courtesy conversations. The word processing softwares that we write articles, books, theses and emails to each other do the same. It is hard to resist. To spend a little less time on the painful human writing process and delete those final human defects that are so evident in our own personal language. Because yes indeed human language is the mirror of each human individual’s personal history and culture, which does also include the struggle of being a messy human. 

It might not be long before human beings are no longer the originators of language, but the screws and bolts of a machinery.  Spinning, turning. Passing on the messages between the machines.

“Hello”, “goodbye”, “what can I do for you?”, “please sir, madam”.

Gry Hasselbalch

Co-founder of DataEthics.eu & Author

Gry Hasselbalch is an acclaimed author, tech critic, and scholar specializing in data and AI ethics, human rights, and politics, holding a PhD from the University of Copenhagen. She has shaped global AI policy, contributing to the EU’s High-Level Expert Group on AI and the EU’s International Outreach for a Human-Centric Approach to AI global diplomacy initiative. As co-founder of DataEthics.eu and a sought-after keynote speaker, she champions a humanistic approach that emphasizes the role of human power in the digital age.

1. Hasselbalch, G. (2025) Human Power – Seven Traits of Human Power, CRC Press.
2.  Fanon, F. (1963) The Wretched of the Earth (p. 100) (translated by C. Farrington), Grove Press (original work published in 1961). 
3. Roth, Emma (2024) ChatGPT’s weekly users have doubled in less than a year ChatGPT’s weekly users have doubled in less than a year. 
4. Hall, S. (1980) Encoding/Decoding. In S. Hall, D. Hobson, A. Lowe, P. Willis (eds.) Culture, Media, Language Working Papers in Cultural Studies, 1972–79, Hutchinson 118–27. An edited extract from S. Hall, Encoding and Decoding in the Television Discourse, CCCS stencilled paper no. 7. (Birmingham: Centre for Contemporary Cultural Studies, 1973). p. 123.
5. Pinhanez, C. et. al (2024) Harnessing the Power of Artificial Intelligence to Vitalize Endangered Indigenous Languages: Technologies and Experiences, [2407.12620] Harnessing the Power of Artificial Intelligence to Vitalize Endangered Indigenous Languages: Technologies and Experiences.
6. Hao, K. (2022, April 22), A new vision of artificial intelligence for people, MIT Technology Review, A new vision of artificial intelligence for the people | MIT Technology Review BK-. 
7. Wanga, H. (2020) AI: a new (r)evolution or the new colonizer for Indigenous peoples? in Lewis (2020, p. 35). 
8. In 2023, hundreds of prominent business people and other renowned figures, including OpenAI’s CEO Sam Altman, Google DeepMind’s Demis Hassabis and co-founder of Microsoft Bill Gates[8] in 2023 published a public statement with just one sentence: “Mitigating the risk of extinction from AI should be a global priority alongside other societal-scale risks such as pandemics and nuclear war.” Statement on AI Risk | CAIS. 
9. Bergson, H. (1914) Creative Evolution (translated by Arthur Mitchell), Macmillan and Co. (originally published 1907). 
10. Bowker, G.C., Star, S.L. (2000) Sorting Things out: Classification and Its Consequences. Inside Technology. | Cambridge. MIT Press. 
11. Alpaydin, E. (2016) Machine Learning | MIT Press. p. 51.
12. Bolukbasi, T., Chang, K-W., Zou, J.Y., Saligrama, V., Kalai, A.T. (2016) Man Is to Computer Programmer as Woman Is to Homemaker? Debiasing Word Embeddings, 30th Conference on Neural Information Processing Systems (NIPS 2016).