What are the foundational principles and technologies behind AI and LLMs?
What are the differences between closed and open source LLMs
How can AI and LLMs be leveraged to enhance public service delivery?
Should we buy vs build vs adopt?
Should we go with GovTech/OGP products, build ourselves, or approach a vendor?
If we build, how can we avoid irreversible decisions that lock us into (soon to be) obsolete paths?
How can AI contribute to more efficient and effective public service workflows?
What are AI agents and what are the benefits/costs of agentic workflows?
What training or skills are necessary for public sector employees to work effectively with AI?
3. Objectives
Sort through the AI hype from the keepers
Understand what it takes to conceptualize, develop, & deploy AI products, e.g. compute infrastructure, toolchain
Know what everyone not on LLMs is missing out
Spot opportunities to do things much faster and easier with LLMs
4. My Observations
The AI genie is out of the bottle, there's no return to life before it
Given the ongoing hype train for AI, it's tempting to use AI for the sake of it, but that'd would be a huge L
Focus on solving a meaningful problem, not the hype
Long tail of training/finetuning yourself, building UIs, and maintaining is too much investment
AI user interfaces and workflows are not necessarily (and can often be contrary to) good UX
Simple applications of ML might be far more useful for everyday problems, as is straightforward improvements to how we organize info, design websites instead of pushing for a chatbot