A Couple Of Things– Caring for AI, and Texting Like We Don’t Care
As we train AI with thoughtful conversations, we’re simultaneously making our human ones drier, shorter, and less expressive. A reflection on two linguistic shifts shaping our conversation patterns.
The evolution of language is as fascinating as the evolution of many other elements we interact with today, and at times take for granted. The application of language has adopted different forms:
· Emojis substituting words and phrases.
· Shorter texts substituting long-drawn explanations.
· Succinct and recurring prompts substituting standard search queries on the internet.
It’s certainly an interesting point in time to reflect upon who’s getting better at training whom between humans and AI, and who’s better off identifying the gaps. I’m going down two conflicting yet distinct trends we’re getting to witness—the importance of conversation loops to build context and evolve AI’s output in the generative era, and the concept of ‘dry texting’. Read on!
Conversation is Everything
Today, every second post on one’s social media feed is some kind of prompt compilation, a learning course, or an AI-generated essay, book, music, video, or an image. A prompt can be viewed as a glorified search query within an AI app, which used to be a set of keywords on a conventional search engine till a couple of years ago. I was listening to Rick Rubin’s conversation with Perplexity AI’s Co-Founder and CEO, Aravind Srinivas, where the latter talks about how the role of the first prompt is just to get the discovery going. Perplexity’s attempt is to fuel the world’s curiosity with accurate, transparent, and accessible information. There’s abundance of information on one side, and there’s limitless curiosity on the other. The Large Language Models (LLMs) bring the two worlds together, thereby satisfying one’s quest for information. A prompt is not a mere search query; it’s just an opening statement. The subsequent prompts get you deeper into the field of information, or closer to the epicentre of the output you desire. The back and forth is what strengthens the context in human-to-human conversations, and it’s the same in the case of human-to-AI interactions. These conversations within the AI apps unlock larger value for every user—much beyond the scope of a standard search query.
Understanding, and making tweaks to our linguistic capabilities is important. It’s not as difficult as learning a new conventional speaking or software language from scratch, but it still requires effort to balance specificity with comprehensiveness. Vocabulary is also evolving alongside the syntax. I chanced upon this piece by Lance Eliot, which paints a scenario of a universe of finite and fake words becoming mainstream. The evolution of words used in common modes of communication is a constant. Today, we don’t use most of the words that were prevalent in the 16th and 17th centuries. However, as AI models are trained on existing data and human preferences, the responsibility of checking for typos, accurate usage, and factual existence of new words making way into the system rests with the architects of these models. I personally haven’t experienced alien words winning my approval, but as language evolves, some part of what and how we speak could need a double take. I’ve been using ‘please’ and ‘thank you’ in my prompts, and I don’t see them going away anytime in the near future.
Dry Texting and Conflict Management
I came across this Vox piece by Anna North, and the headline caught my attention, ‘Are phones making teens more conflict-averse?’ I learnt of the term ‘dry texting’ for the first time through this piece, but I have seen its application in communications across generations and situations over the years. I could never qualify the usage of ‘thx’ and ‘tx’ for ‘thanks’ with an apt word or phrase before today. This phenomenon is certainly a paradox to the conversation and context-rich world that we’re imagining and creating with AI.
Dry texting refers to a style of text communication where responses are notably terse, lacking in emotional cues, elaboration, or contextual richness. Typical features include:
Minimalist Responses: Single-word replies such as ‘K’, ‘yes’, or ‘no’.
Lack of Enthusiasm: Absence of emoticons/emojis, punctuation emphasis, or conversational prompts that invite further interaction.
Perceived Emotional Detachment: Recipients may feel the sender is disinterested or not fully engaged in the conversation.
This phenomenon is not necessarily indicative of negative intent; sometimes, it reflects the sender’s communication style, mood, or situational constraints such as multitasking or fatigue. Dry texting is not just limited to teens as a section of the population. It’s often defined by the cultural influences and preferred style of communication across the world.
It’ll be interesting to keep a watch on how the AI LLMs train on dry texts, decode sentiments and feelings, and factor them in their reasoning engines. So, will dry texts become a valid token, or a high-value currency, or a dying fad? Or would the LLMs end up being just as confused as the teens who dry text?
Listen Inn
‘Whatever It Takes’ by Imagine Dragon, from their third album, ‘Evolve’ released in 2017. The track has added a billion + views on YouTube, and ‘ feels like a great soundtrack to the world-consuming AI ambition, and a self-motivating war cry for the Gen Z.
I’m forever on the lookout for tunes old and new, You can check out my expanding Trove Of Tunes that I’m curating in a Spotify playlist.
Cheers,
Shri