Prompting Dialogue | Redefining the Label
Jonathan "Kyle" Hobson
UX Researcher | Human Factors | ASU Master's Graduate | Curating AI-Infused Human Experiences
In the age where artificial intelligence (AI) weaves through the fabric of daily life, transforming industries and personal interactions alike, the nuanced art of communicating with these intelligent systems becomes paramount. This thesis is an odyssey into the realm of human-AI interaction, a field poised at the brink of revolutionary change, where every prompt we feed into the system can either bridge or widen the gap between human intent and AI comprehension. As we stand at this critical juncture, it is imperative to redefine our approach to AI communication—shifting from mere command to a dialogue, from simple interaction to meaningful engagement.
This treatise weaves a tale where science and artistry dance, each discipline lending its voice, sculpting the space where human meets machine—not merely engineering, but an orchestra of mind, morality, and tongue.
As we stand on the shoulders of giants—from the agrarian revolutionaries who first tilled the earth to the digital pioneers who connected the globe—AI's rapid integration into diverse sectors is the latest chapter in humanity's relentless pursuit of technological mastery. This historical continuum underscores the urgency to enhance how we instruct and understand these digital counterparts. At the heart of this enhancement is prompt engineering—a term that, while technical, encapsulates the intricate dance of language and technology, marking a pivotal moment in our ongoing dialogue with innovation. This thesis argues that prompt engineering, or more broadly, the crafting of AI-human interactions, is not just a technical challenge but a multidisciplinary endeavor that demands insights from psychology, linguistics, ethics, and design. By advancing our understanding of prompt dynamics, we can forge AI systems that not only understand but also anticipate and adapt to the vast spectrum of human needs and emotions.
The exploration begins with a dissection of the evolving language of prompt engineering. As AI systems grow more sophisticated, so too must our prompts evolve from rigid commands to flexible, context-aware dialogues that cater to a broader demographic, including those traditionally marginalized in technology’s rapid ascent. This thesis systematically unpacks the layers of prompt construction, identifies the prevailing patterns that enhance interaction clarity and efficacy, and critically evaluates the usability of these systems across diverse user groups.
Moreover, this research aims to establish a taxonomy of terms and concepts that underpin effective AI-human communication, setting the stage not only for current understanding but also for the uncharted territories we anticipate exploring. With advancements on the horizon from brain-to-computer interfaces to AI-mediated interspecies communication, our exploration through surveys, interviews, and usability testing integrates both qualitative and quantitative methodologies to capture the essence of user experiences today and forecasts the transformative potential of tomorrow. The findings not only reflect on the current state but also pave the way for innovative approaches that could democratize AI usability.
Why is this discussion critical? As AI becomes an integral part of decision-making, its ability to understand and respond appropriately to nuanced human prompts becomes not just beneficial but essential. This thesis posits that a profound shift in prompt engineering can significantly enhance AI's utility, making it a more inclusive tool that serves a broader spectrum of society. By delving into this emerging field, the research encapsulates the urgency and necessity of developing communication strategies that are not only effective but also equitable.
In navigating this complex landscape, this thesis adheres to a multidisciplinary approach, drawing upon the vast reservoirs of computational linguistics, cognitive psychology, and interaction design. It challenges the traditional paradigms of AI interaction and proposes a future where AI tools not only respond to our commands but understand our intentions, context, and diversity.
This introduction sets the stage for a comprehensive examination of how we communicate with machines and how these interactions can be transformed to be more inclusive, accessible, and effective. By the end of this journey, the reader will not only grasp the critical importance of optimized AI-human communication but will also be equipped with the knowledge to participate in or critique this rapidly evolving field. Through this exploration, the thesis aims to illuminate the pathways towards a more connected and understanding digital future.
"AI is the most profound technology humanity is working on... exteriorizing our imagination. It is a technology that rewards experimentation." (Silva, 2024)
The Landscape of Prompting: Terminology, Roles, and Strategies
As we embark on an exploration of the intricate world of human-AI interaction, particularly through the lens of prompting and its pivotal role in AI communication, it is essential to establish a clear understanding of key terms that will recur throughout this discourse. At the heart of our conversation lies Artificial Intelligence (AI), an expansive field dedicated to simulating human intelligence processes by machines, especially computer systems. This encompasses tasks traditionally viewed as requiring human intelligence, such as visual perception, speech recognition, decision-making, and language translation.
In the context of our exploration, the focus sharpens on Generative AI (GenAI) and Large Language Models (LLMs). GenAI refers to AI systems specifically designed to generate new content, whether that be text, images, music, or any form of digital output, by learning from vast datasets. Within this subset, LLMs stand out for their ability to understand, generate, and interact with human language at an unprecedented scale. These models, exemplified by the Generative Pre-trained Transformer (GPT) series, are trained on extensive corpora of textual data, enabling them to perform a wide array of language tasks from translation to content creation, with a degree of fluency that often mirrors human capability.
For the purposes of this paper, every mention of AI, unless specified otherwise, refers to this domain of generative AI, including image generators, LLMs and GPTs. This focus stems from the current technological zeitgeist that sees these models at the forefront of AI innovation, driving forward the capabilities and potential applications of artificial intelligence in mimicking, complementing, and enhancing human communication.
In exploring the intricate world of communication within the sphere of the burgeoning field of human-AI interaction, it's essential to delve deep into the foundational element—the term "prompting" itself. The concept of "prompting" occupies a central position, serving as the linchpin in the dialogue between humans and artificial intelligence (AI) systems.?
The art and science of prompting, encompassing the formulation and delivery of inputs to AI, demand a nuanced understanding that goes beyond mere technical execution; it requires a synthesis of creativity, linguistic precision, and strategic thinking. This multifaceted interaction lays the groundwork for the discipline of human-AI communication, a field that transcends conventional boundaries to redefine the way we engage with technologies.
Prompting: The Quintessence of Human-AI Dialogue
At its essence, prompting, in the context of human-AI interaction, represents a dialogue—a series of exchanges with AI designed to elicit specific responses or actions. This process, akin to understanding how AI interrupts the subtleties of human language and thought, demands more than mere command inputs; it requires an understanding of context, the anticipation of potential responses, and the ability to refine and iterate prompts based on feedback. This dialogue is not just about instructing a computer program but involves a dance of minds, where the human's role shifts from a mere commander to a communicator, a guide, and at times, a teacher to the AI.?
"The world is made of words, and if you know the words, the world is yours" (McKenna, 1992)
As White et al. (2023) elucidate, interacting with an Large Language Model (LLM) “in general involves feeding it 'prompts', which are natural language instructions used to provide context to the LLM and guide its generation…” or transformation “...of textual responses." This definition underpins the essence of prompting as a dynamic dialogue, highlighting the critical role of natural language instructions in bridging the communicative divide between humans and AI. "Generative AI systems require context and detailed information to produce accurate and relevant responses" (Amazon Web Services, 2024). The essence of prompting, therefore, lies in the skillful blend of directing and conversing with AI, steering it through the complexities of human thought and language to achieve desired outcomes.
The significance of prompting, as evidenced by the surge in Google Trends data, is not just a reflection of growing curiosity but a testament to its fundamental role in advancing AI's utility and its integration into our daily lives. This surge, particularly notable from 2021 onwards, underscores an awakening to the potential of AI beyond the confines of structured command inputs to a more fluid, conversational engagement. Referencing the observed trends (see Figure 2.1.3), we witness not only an uptick in the term "prompting" but a skyrocketing interest in "prompt engineering," illustrating a collective movement towards refining the art and science of communicating with AI.?
Through this lens, each prompt emerges as a crafted narrative, a story told to the AI with the expectation of a coherent, contextually aligned response. White et al. (2023) further expand on this concept, noting that "...a prompt is a natural language instruction given to an LLM to facilitate its generation of requirements…”. This perspective illuminates the creative and analytical dimensions of prompting, where the design of each prompt not only directs but also inspires AI to generate responses that span the spectrum from functional outputs to creative expressions.
Expanding on this dialogue, it's crucial to recognize that prompting goes beyond mere technical interaction. It embodies a creative process, where the design of prompts can influence the direction and quality of AI-generated responses. Fu et al. (2023) states that "currently, prompt engineering for large models is a community-wide collective trial and error effort.” This emphasizes the creative aspect of prompting. An art akin to crafting a piece of literature, requires a deep understanding of language and model’s capabilities. Each prompt is a narrative crafted with intention, a story told to the AI with the expectation of a coherent and contextually appropriate response. It is here, in the careful construction and iteration of these narratives, that the true essence of prompting unfolds—a blend of analytical reasoning and creative thought aimed at bridging the human-AI communication gap.
These AI models, as Mozilla's AI Guide (2023) elucidates, are "designed to generate text that resembles human-written content," demonstrating an unprecedented ability to understand, generate, and transform language. The process of tokenization, transforming raw text into understandable 'tokens,' and the subsequent generation of contextually rich responses, showcases the models' prowess but also surfaces the critical need for guided interaction to mitigate the risks of cons such as “hallucination”—where AI generates convincing yet unfounded content.
Incorporating this understanding into the crafting of prompts highlights the dual nature of AI communication: a realm brimming with potential for creativity and innovation, yet necessitating careful navigation to ensure reliability and relevance. It underscores the essence of prompting as not just a technical task but a deeply intellectual exercise that blends creativity with analytical precision.
In essence, the act of prompting encapsulates a multifaceted interaction that extends beyond the functional to touch upon the philosophical. It raises questions about the nature of intelligence, both artificial and human, and the ways in which we can co-create knowledge and understanding. As we continue to explore and refine this dialogue, we delve deeper into the potential for AI to not just mimic human thought processes but to complement and expand them, fostering a symbiotic relationship between human creativity and AI's computational power. This exploration is not just about achieving technical proficiency but about redefining the boundaries of human-AI collaboration, making every prompt a step towards a more integrated and intuitive future.
Prompt Engineering: A Multidisciplinary Nexus
Prompt Engineering stands at the intersection of multiple disciplines, embodying a fusion of computer science, linguistics, psychology, and design. As defined by Amazon Web Services (2024), "Prompt engineering is the process where you guide generative artificial intelligence (AI) solutions to generate desired outputs... it requires detailed instructions to create high-quality and relevant output." This definition underscores the complexity and depth of knowledge required to effectively interact with AI, moving beyond simple commands to crafting detailed, context-rich instructions that steer AI towards generating meaningful, relevant responses.
"One common approach to interacting with LLMs is prompt engineering, where users design and provide specific prompt texts to guide LLMs in generating desired responses or completing specific tasks". (Chang et al., 2023)
Within the multifaceted realm of Prompt Engineering, the generative capabilities of Large Language Models (LLMs) present both unprecedented opportunities and notable challenges. These models, powerful as they are, often grapple with the intricacies of problematic training data—reproducing biases, misinformation, and toxicity embedded within their foundational datasets. This issue, coupled with their struggles to grasp the subtler aspects of human communication, underscores a critical aspect of the prompt engineer's role: bridging the gap between AI's current capabilities and the nuanced demands of human interaction.
Prompt engineers, therefore, navigate a complex landscape, where their expertise becomes crucial in contextualizing the AI's responses to more closely mimic human understanding. This involves accounting for nuances typically taken for granted in human conversations—such as the time of day, the location, even the tone of voice—elements that, while second nature to humans, remain beyond the AI's intuitive grasp. By meticulously crafting prompts that incorporate this contextual information, prompt engineers significantly mitigate the limitations inherent to LLMs, guiding them towards responses that are both coherent and contextually relevant.
Moreover, the challenge of LLM hallucinations—where AI generates convincing but potentially misleading or irrelevant content—becomes a focal point of prompt engineering. While these hallucinations can be viewed as a drawback, reflecting the AI's penchant for generating content devoid of factual accuracy, they also embody the creative potential of generative AI. Skillfully directed, these hallucinations can catalyze innovation, merging disparate knowledge bases to unearth new insights and discoveries. The artistry of prompt engineering lies in discerning which hallucinations can be harnessed for their inventive potential and which must be curtailed to maintain the integrity and relevance of the AI's output.
"Prompt engineering is necessary for LLMs due to their sensitivity to adversarial prompts, requiring careful design to ensure optimal performance" (Chang et al., 2023)
In essence, the prompt engineer's mission extends beyond merely mitigating the cons of LLMs; it is about enhancing their ability to serve as a conduit for creativity and discovery. By providing AI with a scaffold of meaningful context and guiding its generative processes, prompt engineers play a pivotal role in transforming potential pitfalls into avenues for innovation. This nuanced interplay between the limitations and possibilities of AI underscores the dynamic and critical nature of prompt engineering in the continuous evolution of human-AI dialogue.
The Evolution of Prompt Engineering
The rise of prompt engineering as a recognized discipline has been meteoric, a trend vividly illustrated by Google Trends data which shows a significant uptick in searches for "prompt engineering" over the past years. (See Figures 2.1.1, 2.1.2, and 2.1.3)
Figure 2.1.1: The Rising Trend of Prompt Engineering?
Trend Analysis of "Prompt Engineering" Popularity on Google Trends (2021-2024)
Note. Illustrates the trend data for the search term "prompt engineering" as sourced from Google Trends from 2021 to 2024. The data tracks the relative search frequency of the term, indexed from 0 to 100, where 100 denotes peak popularity. This visualization reflects the interest in prompt engineering over time. For a detailed exploration of the data, interact with the visualization. Chart created using Datawrapper, an online visualization tool.
Rise of Prompting
A significant surge in the popularity of "prompt engineering" is observed beginning in 2022, with the sharpest increase in searches occurring in January 2023. The trend peaks in July 2023, marking the highest recorded interest during this period. Following this zenith, the frequency fluctuates but stabilizes at a score averaging around 40, indicating sustained engagement with the topic. These findings from Google Trends underscore the rising relevance and escalating visibility of prompt engineering in recent years.
Figure 2.1.2: The Rising Trend of OpenAI
Trend Analysis of "Prompt Engineering" Popularity on Google Trends (2021-2024)
Note. Illustrates the comparative trend data for the search terms "prompt engineering" and "OpenAI" from 2022 to 2024, according to Google Trends. Popularity values are indexed from 0 to 100 for each term, where 100 signifies the peak of popularity. The visualization juxtaposes the search interest in both terms over time. For a detailed exploration of the data, interact with the visualization. Chart created using Datawrapper, an online visualization tool.
Rise of Prompting
Notably, search interest in "OpenAI" precedes the climb in "prompt engineering" searches, with initial spikes for "OpenAI" occurring independently. In the first quarter of 2023, coinciding with major advancements and releases from OpenAI, both terms begin to experience synchronized surges. Suggesting an interconnected public awareness of the brand and the concept it popularizes. Suggests that key events related to OpenAI's developments are significant drivers of interest in prompt engineering, reflecting the impact of technological breakthroughs on the associated terminologies' popularity.
Figure 2.1.3: The Rising Trend of Prompting
Search Trend Analysis: "Prompt" in AI and Non-AI Contexts (2022-2024)
Note. Illustrates Google Trends data for the terms "prompt engineering," "OpenAI," and "prompt" in both AI-related and non-AI contexts from 2022 to 2024. It differentiates the relative search frequency for "prompt" when associated with AI and when it stands alone, offering an insight into the term's popularity. Data values are indexed from 0 to 100, where 100 represents peak search interest for each category. The visualization captures the evolving interest in AI-related prompts as well as the overall term. For a detailed exploration of the data, interact with the visualization. Chart created using Datawrapper, an online visualization tool.
Rise of Prompting
Throughout the observed period, "OpenAI" and "prompt engineering" show prominent surges in public search interest, indicating growing awareness and engagement with AI technologies. Concurrently, the term "prompt" experiences an uptick in searches, of which approximately 40% are AI-related, as inferred from the search context. Notably, the term "prompt" sees significant search volumes outside AI contexts, suggesting a broader interest in the term that may extend beyond the technological domain.
This surge in the prompting signifies a growing interest and rapid growth in the technology as well as marks the term “prompt engineering” as the preeminent nomenclature within this new emerging field.?
Supporting this, survey data from the research indicates that 19.5% of 179 respondents identify with "prompt engineer" as their preferred professional label, further cementing its standing as the dominant term in the industry.
Figure 2.2.1: Preferred Labels in Prompting and GenAI Involvement
Distribution of Preferred Professional Labels in Generative AI and Prompting
Note. Represents survey results from "Exploring the Landscape of Prompting" survey conducted by Jonathan Kyle Hobson. Data are categorized by participants' self-reported labels in the field of prompting. Reflecting the diversity in professional identity within the community, revealing preferences for various levels of engagement and expertise in the field. For a detailed exploration of the data, interact with the visualization. Chart created using Datawrapper, an online visualization tool.
领英推荐
Preferred Labels
“Prompt developer” (21%)? emerged as the most selected label. With “prompt engineer” closely following at 20%, this label, along with “prompt architect” at 12%, demonstrates significant professional identification within the field. The term “casual prompter” (21%) and “prompt enthusiast” are noted but are viewed as reflecting a more casual or hobbyist involvement. The findings from this survey inform ongoing discussions about standardizing professional titles in the emergent field of AI prompting, highlighting a slight preference for “prompt developer” over “prompt engineer” and pointing to a potential shift in favored industry terminologies. Another term of “Prompt Designer” since this survey was identified as the preferred label for some. In the rapidly evolving domain of Prompt Engineering, we stand at a pivotal juncture that beckons a forward-looking perspective. As we reflect on the discipline's trajectory, informed by a meteoric rise in recognition as depicted by Google Trends, it becomes imperative to cast our gaze toward the horizon. The future of AI, particularly through the lens of DARPA's Launchbury (2017) Third Wave of AI, invites us to envision a paradigm of contextual adaptation. This forthcoming era, where AI synthesizes the rule-based logic and statistical learning from its predecessors, promises a landscape where AI systems not only grasp but also reason within complex, nuanced contexts.
This impending shift necessitates a profound reevaluation of the Prompt Engineering vocation. Traditional perceptions, predominantly grounded in the concrete and the technical, must expand to embrace the fluid and the contextual. The essential role of Prompt Engineers and their kin is poised for transformation. This evolution is not merely about refining technical prowess but about embracing the "three Cs" — creative thinking, critical thinking, and compassionate thinking. These competencies, crucial for navigating the nuanced interplay between human intent and AI capability, underscore a future where communication with AI transcends binary instructions to embody a dialogue rich with human context and subtlety.
Moreover, the envisioning of AI's future — replete with AI agents and swarms designed for task completion under human oversight — underscores the growing complexity and sophistication of AI interactions (Shapiro, 2023c). The limitations of current AI models, notably their struggle with autonomous reasoning and susceptibility to hallucinations, illuminate the indispensable role of human guidance. Prompt Engineers, therefore, find themselves at the crossroads of being architects of reason and guardians against misinformation, shaping AI outputs to align more closely with human expectations and needs.?
"These models are improving fast, the hardware is improving fast as well" (Shapiro, 2023c)
As we segue from discussing the evolution of Prompt Engineering, the narrative beckons us to confront the need for redefinition. The discipline, while rich with potential, grapples with perceptions of exclusivity and technical rigidity that may deter diverse talents from contributing. The imminent redefinition seeks to broaden the tent of Prompt Engineering, embracing a tapestry of skills that includes, but is not limited to, the technical. It is within this reimagined scope that Prompt Engineering can truly flourish, adapting to the complexities of AI's future landscape while championing inclusivity, creativity, and interdisciplinary collaboration.?
In line with this transformation, Maeda and Schillace (2023) note that “Text is like this kind of wire protocol now for sending things back and forth. You don't really need to do as much work to build these rigid near interchange formats.” This insight underscores the shift towards viewing natural text not merely as a medium of communication but as a universal protocol for data exchange, effectively reducing the necessity for constructing complex interchange formats (Maeda & Schillace, 2023). The transition towards viewing the natural language, or human language more broadly, as the new coding language, encapsulates a significant paradigm shift in AI interaction.?
This perspective is reinforced by predictions based on current trends and survey data, which suggest a future where the ability to articulate prompts in natural language will supplant traditional coding skills. The notion that complex programming tasks might soon be achievable through sophisticated prompting, without the need for conventional coding, heralds a transformative phase in AI's evolution—a phase where the barriers between human thought and AI understanding continue to blur.
Thus, as we transition to the discourse on redefining Prompt Engineering, it is with the acknowledgment that the future of this discipline is intrinsically linked to its ability to evolve alongside AI itself. The challenge lies in fostering a paradigm where Prompt Engineering not only guides AI but also cultivates a symbiosis between technology and the rich tapestry of human intellect and creativity. This journey towards a redefined Prompt Engineering is not just a pursuit of technical excellence but a voyage towards a future where AI and humanity coalesce to unlock unprecedented potential.
The Redefinition of Prompt Engineering: Navigating the Terrain
As we contemplate the future trajectory of prompt engineering, underscored by the insights garnered it's evident that the discipline is on the cusp of a significant transformation. This evolution demands not just a recognition of the term's burgeoning popularity but also a thoughtful introspection on its scope, inclusivity, and the breadth of skills it encompasses. Prompt engineering, as it stands, serves as a beacon for those at the nexus of technology and creativity, yet its nomenclature carries the weight of traditional engineering paradigms, potentially obscuring the rich, multidisciplinary essence at its core. Amazon Web Services elucidates, "Prompt engineers use creativity plus trial and error to create a collection of input texts, so an application's generative AI works as expected" (Amazon Web Services, 2024). This insight reinforces that prompt engineering extends beyond mere technical skills, encompassing creative capacities that are essential for effective AI interaction. The technical connotation of "engineering" may inadvertently gatekeep or intimidate potential contributors from diverse backgrounds, including those with expertise in creative fields, linguistics, or philosophy, who may not traditionally align themselves with engineering disciplines.
This all leads us to an imperative to redefine prompt engineering. As we arise from this critical juncture, where the discipline's growth intersects with the advancing frontier of AI capabilities, particularly as envisioned by Launchbury’s (2017) Third Wave of AI. This new phase of AI development, characterized by systems capable of nuanced understanding and contextual adaptation, necessitates a broadening of the prompt engineering umbrella to include not just the technical architects of AI's linguistic structures but also the creatives, the storytellers, and the philosophers who imbue AI interactions with depth, empathy, and a human touch.
“In English, we say, I met someone. In poetry, we say, my heart collided with its other half today” (Hanson, 2023).??
Amazon Web Services further supports this expansive view, stating, "Creativity involves generating new ideas, concepts, or solutions. Prompt engineering can be used to enhance a model's creative abilities in various scenarios" (Amazon Web Services, 2024). This call to action—to redefine prompt engineering—reflects a conscious move towards embracing a more inclusive and expansive view of what it means to guide and shape AI dialogue. It is an acknowledgment that the future of AI, with its complex tapestry of agents and swarms, requires a collaborative symphony of minds. Minds that can navigate the intricacies of human language, culture, and emotion, and translate these into the digital whispers that AI can comprehend and act upon.
Figure 2.3.1: Frequency of AI Interaction by Professional Label in Prompting
Frequency of AI Interaction Across Prompting Professional Categories
Note. Represents survey results from "Exploring the Landscape of Prompting" survey conducted by Jonathan Kyle Hobson. Data are categorized prompting professionals by the frequency of their interactions with AI technologies. The highlighted frequencies—more than daily, weekly, and occasionally—provide a snapshot of user engagement levels with generative AI. For a detailed exploration of the data, interact with the visualization. Chart created using Datawrapper, an online visualization tool.
Frequency of Interactions
Notably, “prompt engineers” exhibit the highest engagement with more than daily AI interactions, while 'prompt artists' show a significant daily involvement. 'Prompt enthusiasts' primarily interact with AI on a weekly basis, underlining a consistent but less intensive engagement pattern. The frequency of interactions suggests varying levels of integration of AI into different prompting roles and activities.
Figure 2.3.2: Experience of AI Interaction by Professional Label in Prompting
Professional Labels and Expertise Levels Among AI Prompting Practitioners
Note. Represents survey results from "Exploring the Landscape of Prompting" survey conducted by Jonathan Kyle Hobson. Data are categorized professional labels and their corresponding levels of expertise within the generative AI and prompting field. The data highlight the varied roles individuals assume and the spectrum of their experience. For a detailed exploration of the data, interact with the visualization. Chart created using Datawrapper, an online visualization tool.
Expertise of Interactions
This visualization accentuates the distribution of expertise levels—particularly 'expert professionals' and 'novice users'—within specific self-identified professional labels in AI prompting. For example, 43% of those labeled as 'AI software engineers' consider themselves expert professionals, whereas 'casual prompters' predominantly classify themselves as novice users. The contrasts and similarities revealed by these highlights draw attention to how professional identity is perceived and reported in the evolving domain of AI prompting.
Therefore, as we pivot from examining the evolution of prompt engineering, we are not just advocating for a semantic shift but championing a paradigmatic shift in the field. One that recognizes the indispensable contributions of a diverse array of professionals—psychologists, philosophers, designers, artists, and writers—who can enrich the AI landscape with their unique insights and perspectives. This diversity, far from diluting the engineering rigor, enhances it, fostering an environment where creativity and technical acumen coalesce to advance the frontiers of human-AI interaction.
By reimagining prompt engineering in this light, we prepare the ground for a more vibrant, inclusive, and dynamic field. A field that not only anticipates the technological advances on the horizon but also remains steadfast in its commitment to harnessing these advancements in a way that is fundamentally human-centric. It is through this lens that we transition to exploring the spectrum of prompting roles beyond traditional engineering, poised to uncover the myriad ways in which individuals from across the disciplinary spectrum can contribute to shaping the dialogue between humans and AI. This exploration is not just a journey into the technical mechanics of AI communication but a venture into the heart of what it means to interact, co-create, and evolve alongside our digital counterparts.
Beyond Engineering: The Spectrum of Prompting Roles
As we delve into the multifaceted world of prompt engineering, it becomes evident that the field transcends traditional engineering confines, ushering in a new era characterized by a broad spectrum of roles. This diversity reflects the evolving needs and approaches to AI interaction, as identified in the survey where:
This rich tapestry of roles underscores a shift toward more descriptive and inclusive terminology, capturing the comprehensive range of skills and activities integral to effective prompt engineering.
"The digital revolution is spilling over into the biotechnological revolution. We are patterning biology, which is also software. Everything can be upgraded, re-tuned, and transformed." (Silva, 2023)
Within the expanses of AI and prompt design, the emergence of non-traditional roles challenges the notion that engagement with AI is purely a technical endeavor. Illustrated in Figure 2.4.1, the distribution of industry professionals by technical and non-technical classifications reveals an insightful trend: individuals from non-technical industries show a marked preference for identifying as “prompt architects” and “prompt artists.” These roles, often adopted by designers, creatives, and other specialists, highlight the need for an expanded recognition of the array of skills contributing to the AI field.?
The figure points to a substantial 57% of non-technical industry participants claiming the title of “prompt architect,” indicating a significant departure from the conventional “engineering” label. This variance not only enriches the AI prompting landscape but also calls for a broader acknowledgment and incorporation of diverse professional expertise. It suggests an imperative to extend our understanding of AI interaction roles well beyond engineering, embracing a full spectrum of contributions that foster innovation and inclusivity within the industry.
Figure 2.4.1: Industry Distribution in AI Prompting Roles
Industry Distribution of AI Prompting Roles: Technical vs. Non-Technical
Note. Represents survey results from "Exploring the Landscape of Prompting" survey and Usability Tests conducted by Jonathan Kyle Hobson. Data are categorized participants by their identified roles within AI prompting and further distinguishes between technical and non-technical industries. It provides insights into the prevalence of specific prompting roles across diverse professional backgrounds. For a detailed exploration of the data, interact with the visualization. Chart created using Datawrapper, an online visualization tool.
Technical vs. Non-Technical?
Technical industry participants predominantly identify as “prompt developers” and “prompt engineers,” a significant majority within non-technical sectors align with the “prompt architect” role. Additionally, a notable portion of those in non-technical fields embrace “prompt artist” designations, suggesting a broader interpretation of AI prompting activities outside traditional tech-oriented domains.
Prompt Architects and Developers: Shaping the AI Landscape
Prompt Architects and Developers stand at the forefront of this evolution, embodying the confluence of creativity and technical precision. They are tasked with constructing coherent, adaptive frameworks for AI interaction, ensuring that prompts not only guide but also inspire AI behavior. Their work parallels architectural principles, where both form and function are meticulously balanced to achieve optimal outcomes.
Conversely, Prompt Developers embrace the dynamic nature of the field, focusing on the iterative refinement of prompts. Their role mirrors the software development lifecycle, emphasizing the importance of feedback loops and continuous improvement in enhancing the AI's understanding and responsiveness.
The Creative Frontiers: Prompt Artists and Designers
Prompt Artists infuse the technical realm of AI with creativity and imagination, crafting prompts that transcend functional directives to evoke nuanced and human-like responses. Their role celebrates the expressive potential of prompting, where innovation meets insight to drive forward the boundaries of AI interaction.
Casual Prompters: Democratizing AI Interaction
Casual Prompters and Enthusiasts, though not immersed in the professional nuances of the field, play a crucial role in its democratization. Their involvement highlights the universal applicability and relevance of prompt engineering, ensuring that the field remains accessible to a diverse audience beyond the confines of technical expertise.
Drawing parallels between casual prompting and ubiquitous digital tools like email and Word, the analogy underscores the evolving perception of prompting as a fundamental skill set. In a digital age where interaction with technology is an integral part of daily life, the ability to effectively communicate with AI through prompts is becoming as essential as traditional computer literacy skills. This democratization of AI interaction suggests a future where engaging with AI will be an intrinsic part of navigating the digital world, accessible to all regardless of technical background.
The Evolution of Prompting: Waves and Systems of AI
The development of Artificial Intelligence (AI) has been marked by distinct waves or systems, each bringing significant advancements in the technology's capabilities and its interaction with humans. For prompters, understanding these waves is crucial, as they not only represent the historical progression of AI but also highlight the evolving role of prompting in shaping AI interactions. As AI continues to advance, the field of prompting must adapt, embracing a multidisciplinary approach that incorporates the expertise of writers, philosophers, psychologists, and creatives alongside traditional technical professionals.
The First Wave (System): Logic-Based AI
The first wave of AI, characterized by static, rule-based interfaces, laid the groundwork for the emergence of prompting as a discipline. In this system, prompting primarily involved programming predefined conditional statements, or "if-then" rules, to guide the AI's responses. Prompters in this wave were primarily technical experts with a deep understanding of programming languages and logic structures. "The first wave was handcrafted knowledge; here, experts took knowledge that they had about a particular domain and they characterized it in rules that could fit in the computer and that the computer could study the implications of those rules" (DARPAtv & Launchbury, 2017). Enabling “reasoning over very narrowly defined domains but they have no learning capability and poor handling of uncertainty" (DARPAtv & Launchbury, 2017). While this system had limitations in terms of flexibility and adaptability, it marked the beginning of the journey towards more natural human-AI interaction.
The Second Wave (System): Adaptable AI and the Rise of Prompt Engineering
The second wave of AI, powered by advancements in Natural Language Processing (NLP) and Natural Language Understanding (NLU), brought about a significant shift in the role of prompters. "A big change was that they started to use machine learning... techniques that were probabilistic in the way that they handled information" (DARPAtv & Launchbury, 2017). As AI systems became more adaptable and flexible, capable of learning from vast amounts of data, prompters needed to blend technical skills with linguistic understanding. “Second wave systems are very good at being able to perceive the natural world... they're able to learn and adapt to different kinds of situations" (DARPAtv & Launchbury, 2017). In this wave, prompt engineering emerged as a discipline focused on guiding the AI's learning process through carefully curated datasets and feedback loops. According to Launchbury (2017) "these systems turn out to be statistically impressive but individually unreliable." Thus, prompters became not just programmers but also data strategists, ensuring that the AI's outputs aligned with the desired context and tone.?
The Third Wave (System): Contextual AI and the Emergence of Prompt Architects
The third wave of AI represents a leap towards models that can understand and engage in natural conversation, grasping the nuances of human communication. This “wave is about contextual adaptation... systems themselves will over time build underlying explanatory models that allow them to characterize real world phenomena." (DARPAtv & Launchbury, 2017) In this system, the role of the prompter evolves once again, moving beyond technical programming to become a conversation architect. Prompters in this wave focus on crafting prompts that leverage the AI's contextual understanding to create engaging, empathetic, and human-like interactions. This shift demands a multidisciplinary approach, incorporating insights from fields such as psychology, philosophy, and creative writing to design prompts that truly resonate with users.
The Interplay of Systems and the Future of Prompting
As AI continues to evolve, it is essential to recognize that these waves do not represent a linear progression where each system replaces the previous one. Instead, each wave builds upon and complements the others, creating a multidimensional landscape for prompters to navigate. For example, the rule-based structures of the first wave can be combined with the adaptability of the second wave to create more predictive and intuitive interfaces. Similarly, the contextual understanding of the third wave can be informed by the vast data processing capabilities of the second wave. Launchbury (2017) states that "we see the need to bring [the first and second waves] together... foreseeing a third wave around contextual adaptation."?
This interplay of systems highlights the need for prompters to develop a diverse skill set that spans technical expertise, linguistic understanding, and creative thinking. As AI becomes more deeply integrated into our daily lives, the ability to effectively communicate with and guide these systems will become an essential skill for professionals across industries.
Moreover, the democratization of AI interaction, as exemplified by the rise of casual prompters, underscores the importance of making the field of prompting accessible to a wide range of users. Just as email and word processing have become ubiquitous tools in the digital age, the ability to interact with AI through prompting will become an integral part of navigating the digital world. As stated by Schillace (2023) “we're still in the stages where the base models are getting better. In fact, often the next generation of the base model will be better than the last generation fine-tuned was. So like the models are continually getting better.” This presents an opportunity for prompters to not only advance the capabilities of AI but also to shape the ethical and social implications of these systems.
In conclusion, the evolution of prompting is intrinsically linked to the progression of AI through its various waves and systems. As we move towards more contextual, empathetic, and human-like AI, the role of the prompter must adapt, embracing a multidisciplinary approach that integrates technical expertise with insights from diverse fields. By understanding the interplay of these systems and the democratization of AI interaction, prompters can unlock the full potential of this transformative technology, creating AI experiences that are not only highly functional but also deeply engaging and accessible to all.
"We are living in a magical world. And if you could tie labs to our progress, it would look like our minds, our consciousnesses, the imagination spilling out of us." (Silva, 2024)