Prompt engineering is not the future – Updates on AI & Technology by Outreinfo
Prompt engineering, the optimization of textual input to communicate with large language models, has been hailed as a high-leverage skill and a job of the future. However, its prominence may be fleeting. Future AI systems will become more intuitive at understanding natural language, reducing the need for engineered prompts. New AI language models like GPT4 are also capable of crafting prompts, potentially rendering prompt engineering obsolete. Additionally, the efficacy of prompts is limited to specific algorithms and may not be useful across diverse AI models and versions.
A more enduring and adaptable skill for harnessing the potential of generative AI is problem formulation, the ability to identify, analyse, and delineate problems. Problem formulation differs from prompt engineering in its focus, core tasks, and underlying abilities. While prompt engineering focuses on crafting optimal textual input, problem formulation emphasizes defining the problem by delineating its focus, scope, and boundaries. Prompt engineering requires proficiency in a specific AI tool and linguistics, while problem formulation requires a comprehensive understanding of the problem domain and the ability to distil real-world issues. Without a well-formulated problem, even sophisticated prompts will fall short.
Problem formulation is a widely overlooked and underdeveloped skill, due in part to the disproportionate emphasis on problem-solving at the expense of formulation. This imbalance is illustrated by the misguided management adage, “don’t bring me problems, bring me solutions.” A recent survey revealed that 85% of C-suite executives consider their organizations bad at diagnosing problems. To improve problem formulation, four key components have been identified: problem diagnosis, decomposition, reframing, and constraint design.
Problem diagnosis involves identifying the core problem for AI to solve, or the main objective you want generative AI to accomplish. Some problems are simple to pinpoint, while others are more challenging. incentive, a company that has helped its clients formulate over 2,500 problems with an 80% success rate, attributes its success to its ability to discern the fundamental underlying problem. They often use the “Five Whys” technique to distinguish root causes from symptoms.
An example of successful problem diagnosis is the subarctic oil problem, which involved cleaning up subarctic waters after the Exxon Valdez oil spill. Incentive, collaborating with the Oil Spill Recovery Institute, pinpointed the root cause of the oil clean-up issue as the viscosity of the crude oil, which became too thick to pump from barges when frozen. This diagnosis led to a solution that involved using modified construction equipment to vibrate the oil and keep it in a liquid state.
Problem decomposition can improve AI solutions by breaking down complex problems into smaller, manageable sub-problems. For example, when tasked with implementing a robust cybersecurity framework, Bing AI’s solutions were too broad and generic to be immediately useful. However, after breaking the problem down into sub-problems such as security policies, vulnerability assessments, authentication protocols, and employee training, the solutions improved considerably. Methods such as functional decomposition or work breakdown structure can help visually depict complex problems and simplify the identification of relevant components and their interconnections.
Problem reframing involves changing the perspective from which a problem is viewed, enabling alternative interpretations. By reframing a problem in various ways, you can guide AI to broaden the scope of potential solutions and overcome creative roadblocks. For example, Doug Dietz, an innovation architect at GE Healthcare, reframed the problem of designing state-of-the-art MRI scanners to “How can we turn the daunting MRI experience into an exciting adventure for kids?” This led to the development of the GE Adventure Series, which dramatically lowered paediatric sedation rates, increased patient satisfaction, and improved machine efficiency.
Problem constraint design involves defining the boundaries of a problem by specifying input, process, and output restrictions for the solution search. Constraints can be used to direct AI in generating valuable solutions. For productivity-oriented tasks, specific and strict constraints can be employed to outline the context, boundaries, and outcome criteria. For creativity-oriented tasks, experimenting with imposing, modifying, and removing constraints can allow for exploration of a wider solution space and discovery of novel perspectives. For example, brand managers use AI tools like Lately or Jasper to produce social media content at scale, setting precise constraints on length, format, tone, and target audience. When seeking originality, however, brand managers can eliminate or modify formatting constraints to generate unconventional content.
Honing skills in problem diagnosis, decomposition, reframing, and constraint design is essential for aligning AI outcomes with task objectives and fostering effective collaboration with AI systems. While prompt engineering may hold the spotlight in the short term, its lack of sustainability, versatility, and transferability limits its long-term relevance. Overemphasizing the crafting of the perfect combination of words can even be counterproductive. Instead, mastering problem formulation could be key to navigating the uncertain future alongside sophisticated AI systems, potentially as pivotal as learning programming languages was during the early days of computing.
