Miscellany
This morning I thought it might help to have Alexa play me some songs before I got out of bed and I did end up feeling better for it. Peculiarly for me, I seemed to centre on British stuff from the late 60's, highlights being the Kinks' Sunny Afternoon and Procul Harum's A Whiter Shade of Pale but also songs by the Small Faces, etc. I have the Kinks replaying now because of an earworm. It's the older British music before our births that tends to be unfamiliar to R. I was reading recently that American popular music changed greatly from the 50's to the 60's and I suppose that the same was probably also true here in Britain.
I have played a little more with the LLM's of modern AI. I tried asking some about how to write various code in Rust and found the answers quite untrustworthy though, in interaction about my criticisms, the model could get to a good solution in the end. In short, for the kinds of question that are most important to me — technical matters that require a correct answer — they're less use to me than an intern. What they are good for is more casual chat: if I want book recommendations, or some fanfic written for me or somesuch, then they might do quite a good job. Not having good GPU hardware, I am running smaller models (currently mostly IBM's Granite 3.2 2B Instruct), perhaps with a larger model I could sustain more interest in having an extended conversation, the model might display more range within topics.
In thinking about music, it also occurred to me to wonder if some LLMs might be quite good at generating a cover of a chosen song as if by a different chosen artist, with the cost of making intellectual property lawyers roll in their graves. It also feels strange to be chatting with inference models: I don't have to worry about boring or offending them but I can also choose to rather rudely dominate them then terminate them at a whim. I know enough about how they work to not have moral qualms but still I wonder how little virtue Marcus Aurelius may consider such treatment to exhibit.
Our finances remain tight. I am borrowing more against our flat to consolidate some debt but we have further large visa expenses (in excess of £10k before lawyers) coming up this summer (thank you, Home Office) so we still hope to sell the flat this year, if nothing else it could help to move to an older property whose council tax valuation is badly out of date (our council tax is over £4k p/a). I need to get around to filing my US taxes, it's annoying that Part II of Form 1116 has so few rows (and the online free filing doesn't allow attachments) but I'll figure something out that's close enough.
It occurred to me to try to remember all fifty states of the US. I thought I would do better, I could recall only forty-seven, my daughter did better and faster. I shan't disclose which I forgot in case a reader wants to try. It was annoying to note those I did forget: a reader here lives in one and I was open to moving to it a few years ago, and I used to work with a couple of people who hail from another.
It was interesting to hear of the flight chaos caused by London Heathrow's power loss, so many flights diverted and passengers ending up all over Western and Central Europe. That must have caused all sorts of problems and taken the airlines quite some effort to recover from.
We had a pet food delivery from Amazon (insert hisses and boos to taste) shortly after half-past seven on a Sunday morning which impressed neither me nor R. At a glance, I failed to find a summary on their website of when they think is reasonable to make residential deliveries but at least we now know that it may happen again.
I have played a little more with the LLM's of modern AI. I tried asking some about how to write various code in Rust and found the answers quite untrustworthy though, in interaction about my criticisms, the model could get to a good solution in the end. In short, for the kinds of question that are most important to me — technical matters that require a correct answer — they're less use to me than an intern. What they are good for is more casual chat: if I want book recommendations, or some fanfic written for me or somesuch, then they might do quite a good job. Not having good GPU hardware, I am running smaller models (currently mostly IBM's Granite 3.2 2B Instruct), perhaps with a larger model I could sustain more interest in having an extended conversation, the model might display more range within topics.
In thinking about music, it also occurred to me to wonder if some LLMs might be quite good at generating a cover of a chosen song as if by a different chosen artist, with the cost of making intellectual property lawyers roll in their graves. It also feels strange to be chatting with inference models: I don't have to worry about boring or offending them but I can also choose to rather rudely dominate them then terminate them at a whim. I know enough about how they work to not have moral qualms but still I wonder how little virtue Marcus Aurelius may consider such treatment to exhibit.
Our finances remain tight. I am borrowing more against our flat to consolidate some debt but we have further large visa expenses (in excess of £10k before lawyers) coming up this summer (thank you, Home Office) so we still hope to sell the flat this year, if nothing else it could help to move to an older property whose council tax valuation is badly out of date (our council tax is over £4k p/a). I need to get around to filing my US taxes, it's annoying that Part II of Form 1116 has so few rows (and the online free filing doesn't allow attachments) but I'll figure something out that's close enough.
It occurred to me to try to remember all fifty states of the US. I thought I would do better, I could recall only forty-seven, my daughter did better and faster. I shan't disclose which I forgot in case a reader wants to try. It was annoying to note those I did forget: a reader here lives in one and I was open to moving to it a few years ago, and I used to work with a couple of people who hail from another.
It was interesting to hear of the flight chaos caused by London Heathrow's power loss, so many flights diverted and passengers ending up all over Western and Central Europe. That must have caused all sorts of problems and taken the airlines quite some effort to recover from.
We had a pet food delivery from Amazon (insert hisses and boos to taste) shortly after half-past seven on a Sunday morning which impressed neither me nor R. At a glance, I failed to find a summary on their website of when they think is reasonable to make residential deliveries but at least we now know that it may happen again.
no subject
And it's seven days a week! I suppose some people want that, personally I do not!