Short-form insights you can read in under 5 minutes.
My work involves taking decades of software engineering experience and applying it to domains where I have to build knowledge quickly. I rely on AI Tools, like the ChatGPT Large Language Model, to get my work done. Here’s a detailed experiment in prompt writing to make sense of the collision of regulatory, power industry, and legal terms of art. #generativeai #regulatorycompliance #businessintelligence
The most reliable use case for large language models (llm) I’ve found is culling through dense contexts of data and writing to produce answers based on prompts created by domain experts who then use that as input to their own data and experience informed analysis. It can drastically reduce time while leaving the human firmly in the loop.
Came up with a topic to propose for a journal article in my 30 min. walk to a conference. I need to get outside more.
Great conversation today with Neil Katz about how EyeLevel.ai’s tools improve the validity and relevance of responses from a Retrieval-augmented generation (RAG) LLM solution. Nothing more true than bad data produces bad results and that context improves relevancy. Nothing more familiar than structured metadata to help a machine create meaning.
It’s not that I’ve become more outgoing per se, it’s that my evolving acceptance of myself has enabled me to better disappear into conversation with others.
Just built ken-bot. A vector database in pinecone chunked from my public writings using langchain as the toolset. This makes the context for my openai request my own words. Now I can ask Ken-bot what I think about things.
I have come to believe that people, for the most part, are trying to get through their day. The cumulative effect is often cruelty. That we can work on. And, yes, there’s self-righteous ignorance and sociopathy. I’m privileged to not have to have to break bread with that.
Starting to reengage with software projects and finding myself tripling down on the fundamentals: drive to outcomes for stakeholders AND their customers , make progress visible, and build mastery so you are capable of making tradeoffs that don’t saddle a project with debt.
I want to learn more about the challenges and opportunities in the distributed energy resource sector. One experiment I’m considering, is target some successful outbound messaging we’ve done around our LLM crash course* to a segment like microgrid operators. This in parallel with the longer term effort to immerse ourselves in their communities of interest.
Blame doesn’t help solve a problem. It might help one survive the fallout, if the problem wasn’t important in the first place. But if that’s true, better to have found a different or more valuable slice of the problem to tackle, then rallying from a place of hope and pragmatism.
Some days, I just accept the day for what it is.
The Agility debates go on but, let me ask, which is faster, a 2001 Bugatti Veyron or a 2023 Rimac Nevera? Answer: depends who’s at the wheel. A professional can beat both in a Chevy Spark if the people operating those magnificent beasts don’t know how to drive.