Hark, I hear the tinkle of falling pennies
(Apple News link)
stocks.apple.com/AeC0ueEjFRvG...
@aboutel.bsky.social
Kiwi in Suisse Romande. Have practised law and taught philosophy. Wondering whether to do WSET 4.
Hark, I hear the tinkle of falling pennies
(Apple News link)
stocks.apple.com/AeC0ueEjFRvG...
The thing is, thinking and working through ideas is one thing, writing essays is another. Students will talk about philosophy into the wee hours because thatโs more fun (and less pressure) than writing it down. I suspect that being baffled by this distinction is a sign of a productive philosopher.
18.02.2026 17:24 โ ๐ 0 ๐ 0 ๐ฌ 0 ๐ 0(I decline to call them by their gaslighting French name)
18.02.2026 07:58 โ ๐ 0 ๐ 0 ๐ฌ 0 ๐ 0Today in Schweizerische Volkspartei.
18.02.2026 07:58 โ ๐ 0 ๐ 0 ๐ฌ 1 ๐ 0Ironically I am more optimistic about this than OpenAI is. โFundamental statistical and computational limitsโ wonโt stop AI getting the facts rightโif we manage to add knowledge to it. But that will involve going beyond LLMs.
15.02.2026 22:40 โ ๐ 0 ๐ 0 ๐ฌ 0 ๐ 0It looks like theyโve found quantitative relationships between โhallucinationsโ (factual errors in generative output) and other kinds of machine-learning error, which is interesting if right. But itโs not why machine learning makes errors!
15.02.2026 22:37 โ ๐ 0 ๐ 0 ๐ฌ 1 ๐ 0Itโs good for various reasons that theyโve acknowledged the bottom line here. But โfundamental statistical limitsโ are not the problem. LLMs are just not in the truth business.
15.02.2026 22:32 โ ๐ 0 ๐ 0 ๐ฌ 1 ๐ 0This, and that this wonโt be changed by more training data or more processing power. Itโs a matter of architecture, not scale.
15.02.2026 04:35 โ ๐ 1 ๐ 0 ๐ฌ 1 ๐ 0And itโs true that knowledge is a thing neural nets can do, viz us. Maybe I will be a urprised again. But neither theory, nor practical results, so far give *any* indication that mere hyper-scaling is going to get us there.
14.02.2026 23:59 โ ๐ 1 ๐ 0 ๐ฌ 0 ๐ 0In that mindset, the objection that stable, updatable representation is a *different kind* of performance and wonโt be conquered by just-more-training-data looks like a failure of imagination, if not Searleite biological chauvinism.
14.02.2026 23:54 โ ๐ 1 ๐ 0 ๐ฌ 1 ๐ 0Iโm trying to understand the AI optimist position. The best I can come up with is that the startling progress of generative AI in the early 2020s led people to think that remaining performance limits (hallucination) would be crashed through just as quickly.
14.02.2026 23:52 โ ๐ 1 ๐ 0 ๐ฌ 2 ๐ 0I know this will annoy some, but imo philosophers *on their staff* who are helping them push this โmaybe theyโre already minds ๐คโ stuff should not be treated seriously as philosophers.
I know the field has to include holders of absurd views by definition. But not actual sophists!
Right. No sympathy. But the argument that tax rises are counterproductive because, er, half a dozen named people left in response is pragmatically self-refuting.
12.02.2026 08:43 โ ๐ 2 ๐ 0 ๐ฌ 1 ๐ 0My guess is what weโre seeing here is Apple torn. They want to release an AI assistant, but (unlike everyone else in the space) they canโt bring themselves to release it until itโs reliable. Which it canโt be, because LLM.
No obvious way outโฆ
www.macrumors.com/2026/02/11/s...
Yeah if a news story gives the *individual names* of the rich people leaving a country of 70 million, it is not describing an exodus.
giftarticle.ft.com/giftarticle/...
I donโt like most of those, but I still think digital watches are a pretty neat idea.
04.02.2026 23:57 โ ๐ 2 ๐ 0 ๐ฌ 0 ๐ 0The sadly neglected paradox of the Truthteller
xkcd.com/3201/
Hi perhaps you would like a little break from the news in the form of an essay about my trip to Centralia, Pennsylvania, the town thatโs been on fire since 1962โa place where surprisingly I found a great deal of hope, wonder, and mystery.
13.01.2026 23:06 โ ๐ 495 ๐ 170 ๐ฌ 23 ๐ 31Said Trump on 5/1/2021
06.01.2026 13:56 โ ๐ 0 ๐ 0 ๐ฌ 0 ๐ 0Two gondolas are about to pass each other in the mountains above Churwalden, Switzerland
New Yearโs Day in Switzerland (above Churwalden)
01.01.2026 12:13 โ ๐ 0 ๐ 0 ๐ฌ 0 ๐ 0May have said this before, but I am amused that hidebound Cambridge, with its 100% closed-book end of year exams, finds itself on the cutting edge.
28.12.2025 08:50 โ ๐ 1 ๐ 0 ๐ฌ 1 ๐ 0how wild is it tho that the fuckin king of england is the one clapping back at white supremacy and labour is the one pushing it
25.12.2025 15:55 โ ๐ 43 ๐ 8 ๐ฌ 1 ๐ 0See also the Swiss โUnion dรฉmocratique du centreย ยป, which is more honestly named across the Rรถstigraben as the Schweizerische Volkspartei.
23.12.2025 04:28 โ ๐ 2 ๐ 0 ๐ฌ 0 ๐ 0Couldnโt tell you if itโs an adverb or what, but it should be โmerryโ: itโs parallel to โkeep you merryโ. www.etymonline.com/word/rest last entry
17.12.2025 10:07 โ ๐ 0 ๐ 0 ๐ฌ 0 ๐ 0My peeve is that this is the gateway version of โI know about the germ theory of disease but also I know better.โ Good luck RTS.
Wearing warm clothes is better than declining vaccines, obvs. But the anti-vax Facebook memes have to work on something.
www.rts.ch/info/sante/2...
I hate this. GDP per capita didnโt regain 2007 levels until 2015. There should have been a rebound, not just โsome growthโ. Osborne is celebrating that he only choked the economy and didnโt garrotte it.
14.12.2025 15:33 โ ๐ 10 ๐ 0 ๐ฌ 0 ๐ 0Lineup of Grover Zampa wines for a tasting at their Bengaluru winery
Proper Bangalore wines at Grover Zampa.
14.12.2025 08:23 โ ๐ 1 ๐ 0 ๐ฌ 0 ๐ 0Iโm normally that โApple know what theyโre doing, get used to itโ guy, but now I know the Truth About Alan Dye, I am reverting liberally.
14.12.2025 03:42 โ ๐ 0 ๐ 0 ๐ฌ 0 ๐ 0So, El Fasher looks like one of the worst single event atrocities to happen this century so far, but the evidence has come in such a trickle it seems to only now be getting verified
13.12.2025 03:50 โ ๐ 2098 ๐ 1169 ๐ฌ 39 ๐ 135Guilt by association, presumably. If you mean, what is their official reason, Iโd say guilt by association โand you canโt be too careful about tarrismโ.
12.12.2025 10:36 โ ๐ 0 ๐ 0 ๐ฌ 0 ๐ 0