Missing puzzle piece to LLM Enterprise Augmentation
Since early last year, when we led the development of an enterprise-level GenAI-as-a-service platform, we have understandably been bombarded with questions like “What are the art of possibles for …” or “Can LLM do …”
In this blog post, we will dive into a critical skill that will enable you to answer all these questions better — computational thinking. By the end of this blog post, you will have the answers to:
- What is computational thinking?
- Why is this relevant to developing LLM use cases?
- What is a four-step process that allows us to inject computational thinking into LLM use case development?
Computational thinking is a problem-solving framework that algorithmically breaks down a task into what I like to call atomic tasks. It involves designing a step-by-step algorithmic approach to solving a problem, identifying similarities and inefficiencies, and evaluating the relative importance of each step.
Imagine cooking a dish.
Recipe sourcing, grocery shopping, ingredient prep, cooking steps, and dishing are the atomic tasks that become an individual unit of action.
Breaking a complex task down into these atomic tasks allows us to see the process clearly. If you know that you will need to add sauce later down the line, get it ready ahead so that it is within arm’s reach. If you need an ingredient sooner, maybe put it closer to you. You will no longer need to rush through finding an ingredient only to realise that you have skipped a step.
Computational thinking also allows us to uncover transferrable knowledge. If we have to dice chives for garnishing, and we have to dice celery for cooking, too, the two steps require the same skills and tools. Can we then group it to reduce inefficiencies?
This framework of thinking also enables us to plan the amount of time and effort needed for each step for better predictability & robustness. How long does it take to heat a pan, chop your ingredients, prepare the sofrito, and cook your pasta…