LLM links, 3/19

Agent-based economic models; Moshe Koppel on AI and Judaism; Brian Chau on a pro-AI manifesto; Ethan Mollick writes his book on AI with help from LLMs

Tyler Cowen points to a paper by Benjamin S. Manning, Kehang Zhu, and John J. Horton that uses large language models to help populate the simulation technique known as agent-based modeling. In an article I drafted that will come out later this year, I anticipated this development. But I do not see it as anything earth-shattering.

Moshe Koppel writes,

I will make two main points. First, drawing on work being done in my own AI lab in Israel, I will show how AI can provide tools that benefit Judaism by making Jewish texts and ideas more accessible. Second, I will suggest ways in which Judaism might, in return, offer models for purposeful and meaningful living, even as ubiquitous AI threatens to attenuate some of our deepest social and moral attachments.

We discussed some of his AI work last month. Later, we turned to the topic of the differences between Israeli youth and American youth.

Brian Chau writes,

Artificial intelligence is a process, not an object. It is not oil, shovels, or images. It is more like mathematical equations, programming languages, or speech. Artificial intelligence is a set of statistical methods used to turn information and energy into output, like new images or emails. It is a wide research area that includes simple methods a single software engineer could manually write in one night and billion-dollar models such as OpenAI’s ChatGPT. 

He describes a new think tank and manifesto to fight clumsy regulation of AI. Perry Metzger is the founder, and Brian is the staff. Brian talks with Metzger here.

Ethan Mollick writes,

It gave me advice on what to cut and simplify in one chapter, which I acted on, but also a warning: Your attempt to infuse humour , particularly in the AI joke section, is commendable. However, there is a risk that it may come across as forced. Remember, dear Ethan, humour is a spice to be sprinkled sparingly, not poured. Pretty direct feedback (and the addition of British English is a nice touch). But AI editors, unlike human ones, can be safely ignored on some topics, so I kept the jokes (you can let me know if that is a good idea or a bad one).

He makes much better use of LLMs than I could. He has practiced a lot more.

Note: I schedule my posts about a week in advance, on average. I think this is good, for many reasons. But in the case of machine learning, reacting to last week’s news can seem particularly out of date. Sorry, but it is still going to be my minhag.

substacks referenced above:

@

I, Cyborg: Using Co-Intelligence

@

The Alliance for the Future Manifesto

@

Perry Metzger: Announcing Alliance for the Future