Encoding Knowledge
No. 07 | May 10, 2024
I am an analyst by training. At this point in my career, I believe that I am also an analyst by nature as well. Work, habit, training, structure, goals, and outcomes collapse into one frame. That frame is easy to describe and less easy to achieve: creating the clearest possible view of the present so that others can enact their most compelling vision of the future.
I am not alone as an analytical person in being fascinated by artificial intelligence - for what it is is, for what it can do, and for what it might be. It is an important part of my work today, and certainly one of the most dynamic parts of my work. It is a whirlwind of growth and attention, a multi-dimensional agent.
AI is one of the most immediately apparent change drivers in our global energy system. It is a major driver of compute demand, and therefore of power demand. I believe it is also one of the critical levers for changing our climate and emissions future by changing our engagement with information.
For one element of AI capabilities, that information is ‘unstructured data’. Not neat tables in rows and columns, but rather words, text, pages, documents. This sort of information is almost aggressively opaque, even to experts. It is historically the domain of human eyes and brains. Now it is also within the purview of large language models.
Proceedings and rulemakings and dockets of regulatory activity work according to their own logic, with their own style, and for their own purposes. Accessing this unstructured date with speed and scale requires multiple steps of structuring, starting with indexing, moving to classification, and then ranking. At that point, it is usable but not accessible, so to speak, which is where large language models play a role.
LLMs act as a bridge between database and human language. Or to put it another way: humans work to encode knowledge into software; LLMs allow us to ask it questions, and knowledge speaks back to us.
At Halcyon we have kicked off our content series on decarbonization, AI, information, and platform-building. The words above are a sample. The links below contain much more.
