Blogging on and off since 2003
I keep reading pieces which dismiss concerns about the environmental impact of AI data centers. They typically couple a description of the best-possible data center practices (regardless of whether these are actually being followed) with simplified comparisons of power and water usage of data centers and urban areas. What bugs me is that these justifications Continue reading "Experiential vs…
This is a long, depressing, blockbuster of a piece. Back in the 1960s I worked on Labour Party political campaigns as part of their youth group, the Young Socialists. My friends and I were definitely on the left, but still mainstream; it was a shock for me to arrive at university in 1969 and encounter Continue reading "The Coroner s Report on the Labour Party"
Personal anecdote, inspired by something in social media: In September 1975 I was a postgrad student at the University of Newcastle-on-Tyne, doing research on operating system command languages for my (never-completed) PhD. I was tapped as one of the three rapporteurs for a multi-day seminar on programme correctness proof with Djikstra. The format was simple; Continue reading "Correctness"
I just asked Microsoft Copilot if it would be possible to eliminate the anthropomorphic aspects of my interactions with the system. This was it s response: Yes — it is possible to interact with Copilot in a way that minimizes or eliminates first‑person language, avoids anthropomorphism, and uses “Copilot” as the only form of self‑reference. The Continue reading "From chat to service provider "
I recently had a lively online discussion about chatbots and liability, and it quickly emerged that one source of disagreement was that many people in the US tend to think that US precedents are all that matter. And this is obviously not the case. For example, the nearest thing to black letter law is the Continue reading "Chatbots and liability: thinking internationally"
It seems to me that today s LLMs try to solve all classes of problems with the same set of mechanisms. We need modular architectures that use different techniques for different aspects of problems. For example, consider an office productivity situation in which a network of agents is scheduling meetings, transcribing the speech content of video Continue reading "More on AI . from epistemology to…
[My inputs in italics; edited for length.] I feel that a fundamental weakness in current LLMs is their inability to resolve conflicting training data based on experience. Here s what I mean. Suppose an LLM is trained on a data set that includes document A, containing X is a Y , and document B containing X is Continue reading "A conversation with Copilot about AI learning."
When I m filling in an opinion poll, I am usually asked which party I support. And when I choose Democrat, the next question is Do you consider yourself a strong Democrat or a weak Democrat? I always choose Strong, because #reasons, but what I want to say is, I m a Roosevelt Democrat, because of fundamental Continue reading "I ve decided that I am a Roosevelt Democrat. And you should be too."
Tim Bray nails it: I’m not sure the profession I joined last century would attract me today. Why? This is @GordWait on Mastodon: “At our office, we are noticing a huge drop in Comp Sci co-op applications. The next generation is convinced there’s no future in programming thanks to AI hype.”
For many years I ve collected model buses, aircraft, and trains, mostly from the UK. The process was always pretty simple: the vendor would discount the price because VAT wasn t applicable, add a shipping cost, and send the item on its way. It usually arrived within two weeks. Shipping was obviously more expensive than within the Continue reading "Buying from overseas in Trump s America"