We’re living in an AI-powered world. The way we are communicating with technology and each other is changing. A recent study found that 40% of respondents were studying English to mitigate the impact of AI and technology on their jobs. English proficiency is emerging as a top asset in today’s business landscape. It is spoken by almost one in every five people on Earth and has solidified its position as the de facto global language.
However, it’s not just the language of business; English is also the internet’s primary tongue. That’s why it’s the learning ground for AI, and its main mode of communication, with AI developers typically training their models on English-language data. This makes it the most crucial “programming language” in AI, overtaking computer languages such as Python. With AI changing the nature of work and our daily lives, how can we utilize English to help us navigate the shifting tides?
Decoding AI
AI refers to a machine’s ability to emulate the behaviors and outcomes we typically associate with human intelligence. The most common form an average person uses today is generative AI technology based on large language models (LLM). These models can understand and generate human language text by processing vast amounts of digital data.
LLMs use prompt engineering, a concept that originated among English speakers and was developed in English. They are steeped in English from conception and still a lot of the language data, performance, accuracy benchmarks, and experiments, including training and refining, are happening in English.
LLMs have many positive applications across industries, benefiting businesses and employees alike, from lending the marketing department a hand in content generation to aiding the operations team with task automation, and providing the sales division with sentiment analysis. Automating tasks enables people to spend more time on the cerebral aspects of their professions, analysis, and strategic thinking.
Nevertheless, businesses integrating AI must implement robust AI governance to mitigate potential security, safeguard risks, and uphold ethical principles. For that reason, it’s vital to invest in accurate data sets to avoid risks around prompt manipulation that can provide inaccurate or misleading information. Implementing input filtering, conducting system audits, and recognizing AI’s current limitations are equally important.
Crossing the chasm
Despite LLMs behind popular chatbots being capable of responding in multiple dialects they still “think” in English. This bias stems from their training data, which favors English-centric concepts. Researchers have found the path of processing almost always passes via what they call the “English subspace.” Therefore, English proficiency helps to bridge the gap between individuals and the ever-evolving world of AI and technology as they can better understand and engage in these transitions.
