David Weinberger's Blog, page 2

Do LLM’s store what they’re trained on? A reason to think not.

If you ask chatGPT to give you the first paragraph of Moby-Dick, it gets it exactly right. This and other such examples are sometimes taken as evidence that it has kept a copy of that book in memory, a potential copyright violation. The makers of these Large Language Models say they don’t, or at least […]

AI’s idea of knowledge

Traditionally in the West we define knowledge as a justified true belief. But the experience of knowledge usually also requires understanding and a framework of connected pieces of knowledge. Guess what machine learning lacks: understandability and a framework from which its statements of knowledge spring. We might want to say that therefore ML doesn’t produce […]

A Quick Guide for Academics Writing for a Broad Readership

via Midjourney, with some tweaks by the author. CC-0. (The prompt is at end of the article.) So, you’re an academic, a researcher, a scientist … a serious person. But you’ve decided to write a trade book — a book for the general populous — to amplify the effect of your work. Excellent! I edit […]

Three introductions to Jacob Collier

After a lot of work, study, and silent prayer, I’ve chosen three works to help you fall in love with Jacob Collier the way I have. Unless you already have. It happened by accident to me. I came across one YouTube of him — I don’t remember which one — and then fell into another, […]

The V.A.N.C.E. System of Voting

America is being run “by a bunch of childless cat ladies who are miserable at their own lives and the choices that they’ve made and so they want to make the rest of the country miserable, too.”  — J.D. Vance I am truly excited, right down to my authentic work boots, by the tremendous response […]

Limiting AI’s imagination

A Large Language Model (LLM) such as ChatGPT or Gemini has a setting called “temperature” that controls the randomness of its responses. The higher the temperature, the more random and diverse the response will be. So I gave it a try with two different prompts. PROMPT #1: I want you to answer the following question […]

gMail losing autocomplete? Here’s a fix.

Every few weeks, gMail has stopped auto-completing addresses as I type them into the “To:” box. Rebooting Chrome does doesn’t help. Rebooting my MacBook Pro doesn’t help. Clearing my cache doesn’t help. The obvious remedy of typing people’s addresses manually requires that I remember their addresses. That  has only gotten less plausible as I’ve aged. […]

What data would ChatGPT add to train a weather model

My prompt If you were going to train a model to predict rain showers, what data might you use that is currently not being widely used? ChaptGPT 4o Many advanced weather models incorporate a variety of data sources, but some of the more innovative and less traditional data types mentioned are not commonly used in […]

chatGPT knows its letters

A friend on a mailing list pointed to the massive palindromic story “2002” by Nick Montfort and William Gillespie, posted on Feb. 20. 2002, AKA 20-02-2002. It’s 2002 words long. And while it’s not exactly a great read, it is an amazing accomplishment. So, I asked chatGPT to confirm that “it’s”2002” is an accurate palindrome. […]

chatGPT admits it assumes a Western, Northern culture

What follows is a long-ish exchange with chatGPT 4.0 asking it about its cultural assumptions. This is not in doubt, so why did I bother? Because it beats working. The salient yet obvious admission is in chatGPT’s last response. I have bolded and italicized it, so it will be easier for you to find it, […]