This is a rejected op-ed (or at least none of the major papers in California I sent it to bothered to respond and say no-thanks, it could be none of them even looked at it). Might as well post it on personal blog and have a few hundred folks read it.

Recently Gov. Newsom released a letter of interest (LOI) for different tech companies to propose how the state could use GenAI (generative artificial intelligence) to help with California’s homeless problem. The rise in homelessness is a major concern, not only for Californian’s but individuals across the US. That said, the proposal is superficial and likely to be a waste of time.

A simple description of GenAI, for those not aware, are tools to ask the machine questions in text and get a response. So you can ask ChatGPT (a currently popular GenAI tool) something like “how can I write a python function to add two numbers together” and it will dutifully respond with computer code (python is a computer programming language) that answers your question.

As someone who writes code for a living, this is useful, but not magic. Think of it more akin to auto-complete on your phone than something truly intelligent. The stated goals of Newsom’s LOI are either mostly trivial without the help of GenAI, or are hopeless and could never be addressed with GenAI.

For the first stated goal, “connecting people to treatment by better identifying available shelter and treatment beds, with GenAI solutions for a portable tool that the local jurisdictions can use for real-time access to treatment and shelter bed availability”. This is simply describing a database — one could mandate state funded treatment providers to provide this information on a daily basis. The technology infrastructure to accomplish this is not much more complex than making a website. Mandating treatment providers report that information accurately and on a timely basis is the hardest part.

For the second stated goal, “Creating housing with more data and accountability by creating clearer insights into local permitting and development decisions”. Permitting decisions are dictated by the state as well as local ordinances. GenAI solutions will not uncover any suggested solution that most Californian’s don’t already know — housing is too expensive and not enough is being built. This is in part due to the regulatory structure, as well as local zoning opposition for particular projects. GenAI cannot change the state laws.

For the last stated goal of the program, “Supporting the state budget by helping state budget analysts with faster and more efficient policy”. Helping analysts generate results faster is potentially something GenAI can help with, more efficient policy is not. I do not doubt the state analysts can use GenAI solutions to help them write code (the same as I do now). But getting that budget analysis one day quicker will not solve any substantive homeless problem.

I hate to be the bearer of bad news, but there are no easy answers to solve California’s homeless crisis. If a machine could spit out trivial solutions to solve homelessness in a text message, like the Wizard of Oz gifting the Scarecrow brains, it would not be a problem to begin with.

Instead of asking for ambiguous GenAI solutions, the state would be better off thinking more seriously about how they can accomplish those specific tasks mentioned in the LOI. If California actually wants to make a database of treatment availability, that is something they could do right now with their own internal capacity.

Solutions to homelessness are not going to miraculously spew from a GenAI oracle, they are going to come from real people accomplishing specific goals.

If folks are reading this, check out my personal consulting firm, Crime De-Coder. I have experience building real applications. Most of the AI stuff on the market now is pure snake oil, so better to articulate what you specifically want and see if someone can help build that.