Prompts as a skill in a new kind of language

For millennia, we have communicated almost exclusively between humans and humans. We are now seeing the rise of people having generalised conversations with AI, shaping language again. As part of this, Martina Manfrin — a data science analyst from Italy — explores why ‘prompt engineers’ will be a key part of how we communicate with AI in the future.

For Martina, shared language between machines and humans has become an essential way to understand the world. We have primarily communicated with computers via clicks and keystrokes, articulating our conversations with humans in a digital world. We are transitioning further towards talking with AI directly, perhaps even using our own voices.

Martina is a computational linguist, meaning she researches, creates, and maintains models that help technology to better process human language. On some days, she works directly on software that helps devices understand human speech. On others, she might spend more time studying human speech habits to better understand recent language developments. Through her work, she is aware of the power of language to adapt and change to the pace of culture and human interactions. For example, she pointed towards how language evolved through the centuries in line with the times — we don’t speak in Old English anymore, though fragments exist in our lexicon today (such as ‘bōc’ which evolved into ‘book’).

Martina thinks that, like all tools, it requires skill. In the case of AI, the skill directly relates to language — the way in which we articulate the prompts we use to communicate with AI systems like Bing or Midjourney. The best-articulated prompts draw out the most relevant information to support our work. If done well, AI becomes a supportive tool for our own innovation, as a parallel track for our work.

It’s not lost on Martina that language is a core part of our prompts. We use our human language for instructions — and it is possible that our language itself will be shaped by AI. For example, the word ‘prompts’ has seen a large spike in use, according to Google Trends data. What other phrases will leap into the vernacular? Regardless, these changes are quite normal. SMS taught us to be more concise, probably AI prompting will teach us to be more word conscious.

Martina also commented that technology influences how we communicate with one another. We already migrated from faxes to email, then again to mobile devices and beyond. We can now communicate via avatars in virtual spaces if we wish to. We are now seeing a new dimension where humans are not just interacting with one another, but with machines directly — which will require a deft hand for strong communication between both parties.

This brings us to Martina’s prediction: eventually, there will likely be a role called the prompt engineer, who will be skilled in asking AI the right questions using specific words and sentences to generate efficient or creative responses. For example, there could be a prompt engineer for images, text, or coding, each with advanced skills in their respective area. This will enable us to work more efficiently, and some of the tedious tasks we used to delegate to junior members will be taken over by AI. As a result, we will need to adapt our working styles, and entry-level positions will require higher levels of preparation to remain competitive.

It’s a big prediction, but it’s one that fits into the mold of how humanity uses language to speak with one another. It also challenged how Martina saw AI as a conversation partner. In centuries past, memorizing poetry or playing chess were considered ‘exclusively-human’ abilities, but computer science has demonstrated otherwise. Today, we see the same phenomenon with creativity; our conversation partners are not necessarily human.

Perhaps our creative process is merging conversations with AI and people together, as part of how our conversations change. “This is fascinating, and it is also emblematic of the most advanced technologies,” Martina said.