A friend and former colleague passed along an article arguing that AI would have a bigger impact on government than the private sector. That may or may not be true but regardless there is a massive frontier that LLMs open in how we adjudicate causal claims, set policy and navigate our incredible complex, "pretzel palace" like maze that is regulatory and legal processes in a place like California. Here's an AI summary:
The article discusses the uneven distribution of skepticism towards artificial intelligence (AI) across various sectors, highlighting a general underestimation of its transformative potential for governments compared to its applicability in the business world. Stephen Bush points out that while there's considerable optimism about AI's role in creative fields, exemplified by Hasbro's CEO Chris Cocks' remarks on leveraging Dungeons & Dragons' extensive content, there are underlying tensions related to copyright and intellectual property that could hinder AI's widespread adoption in the private sector. The costs associated with AI, from intellectual property issues to the financial and environmental costs of running powerful processors, may limit its transformative impact in business.
However, the article argues that states have a unique advantage in leveraging AI due to their control over copyright laws and access to vast amounts of public policy-relevant data. Governments can potentially automate bureaucratic processes without the legal and cost barriers that private companies face. For instance, the UK government is improving data collection and storage, leading to significant efficiencies in public healthcare, as seen in trials at Chelsea and Westminster hospital. Similarly, in the US, AI is being used to expedite the planning process within the Department of Housing and Urban Development. The piece suggests that while AI may not revolutionize job roles on a broad scale, its capacity to transform government operations and services is substantial and deserving of more attention.
That broader claim comparing the public and private sectors may or may not be true. With respect to the public sector specifically, there does seem to be a great deal of untapped potential in a place like Los Angeles. Here is a list of LLMs that would be excellent to see.
A GPT that answers questions about CEQA approvals and the status of various projects.
A GPT to enable local residents to understand the permitting parameters for activities like local movie filming that impact their lives and literally get on their lawn.
Relatedly a GPT that goes answers questions about permitting requirements based on the project type and location, ideally built upon a computable dataset of parcel level zoning / land use information and various development fees and requirements. Much of that is supposed to exist in machine readible form by law but doesn't due to poor enforcement mechanisms.
A GPT that goes through all the CA open data and provides FRED style charts on demand. One thought is to have a crawler go to each of the pages and inventory what’s in the datasets. I wonder if it could count and categorize things.
A GPT that answers questions about the local economy, business stops and starts and other comings and goings like public meeting notices and agendas.
A public finance GPT that leverages state GF and agency and then also the panoply of local agencies. Something that's not just LLMs but uses the wolfram alpha plugin would be pretty apropos here for number crunching. The layer cake of federal, state, and zillion of local governmental funds deserves some too cheap to meter intelligence thrown at it.
Some of this would likely just require a simple crawl and then a text dump into a bag of words for some prompt engineering. So not entirely trivial though not that crazy a lift.
