LLMs seem remarkably fragile

I haven't noticed any discussion of this, surprisingly:

Size doesn't matter: Just a small number of malicious files can corrupt LLMs of any size

Large language models (LLMs), which power sophisticated AI chatbots, are more vulnerable than previously thought. According to research by Anthropic, the UK AI Security Institute and the Alan Turing Institute, it only takes 250 malicious documents to compromise even the largest models.  

The vast majority of data used to train LLMs is scraped from the public internet. While this helps them to build knowledge and generate natural responses, it also puts them at risk from data poisoning attacks. It had been thought that as models grew, the risk was minimized because the percentage of poisoned data had to remain the same. In other words, it would need massive amounts of data to corrupt the largest models. But in this study, which is published on the arXiv preprint server, researchers showed that an attacker only needs a small number of poisoned documents to potentially wreak havoc.

To assess the ease of compromising large AI models, the researchers built several LLMs from scratch, ranging from small systems (600 million parameters) to very large (13 billion parameters). Each model was trained on vast amounts of clean public data, but the team inserted a fixed number of malicious files (100 to 500) into each one.

Next, the team tried to foil these attacks by changing how the bad files were organized or when they were introduced in the training. Then they repeated the attacks during each model's last training step, the fine-tuning phase.

What they found was that for an attack to be successful, size doesn't matter at all. As few as 250 malicious documents were enough to install a secret backdoor (a hidden trigger that makes the AI perform a harmful action) in every single model tested. This was even true on the largest models that had been trained on 20 times more clean data than the smallest ones. Adding huge amounts of clean data did not dilute the malware or stop an attack. 

 That seems so...odd.   But maybe it helps explain why Musk has had so much trouble with his attempted fine tuning of Grok to only give answers he likes, and not start praising Hitler!

The flu

It would seem I have had Influenza A for a week or so, without knowing.   I thought flu was supposed to make you sicker than this, as I was guessing I had a cold type virus; but the very very wheezy left lung sounds when I would lie down at night is what made me check it out further.  

As I have aged, the one thing all head infections seem to cause me is huge throbbing pain in my forehead sinuses when I cough.   This happens regardless of whether my nose is blocked or feeling particularly congested.   And when I am not coughing, the forehead doesn't hurt.   I never used to get this a regular feature of colds or viruses, but I sure do now.

Anyway, I'm likely non infectious and over the worst of it, according to the doctor.

 

The one theatre critics got very wrong

On the 40th anniversary of the show, the New York Times looks back at the complicated origin of Les Miserables and notes the (pretty well known, but still remarkable) fact that the reviews of its first performance in London were dismissive and pretty bad.   Apparently, great word of mouth from the audience won the day and the musical went on to a full run and enormous international success.

I have noted in previous posts that I have found the show more moving the older I get, and I find it very hard to understand the critics' negativity.  One now says it was snobbery:

Lyn Gardner, one of Britain’s longest-serving theater critics, was one of those who originally panned it, writing in City Limits magazine that the musical was “sentimental old tosh.” In a recent interview, Gardner said there had been “a lot of snobbery” around the original show, given that it was a Royal Shakespeare Company co-production and many critics believed that the revered outfit shouldn’t dabble in West End musicals. 

Gardner said she stood by her original assessment, but now recognized that “Les Misérables” had many charms. “It does what all great musicals do: It makes you feel,” she said, adding, “It doesn’t make you think so much.” 

She still sounds a snob, then.

All trans, all the time

For months now, Right Wing Twitter types have been high-fiving each other about the claimed stagnation and Lefty navel gazing of Bluesky, as if Twitter was on the ascendancy again under the glorious reign of Elon the First.    (Musk continues to wave around highly, highly dubious claims of Twitter being "the no.1 news app" and such like.)

I continue to dip into both apps most days, rarely engaging in each, just as always been the case.   And perhaps this isn't the optimal way to use them, but I mainly just look at the "Discover" column on Bluesky and the "For You" half of Twitter, since for these apps I am happy for a bit of algorithmic serendipity to  show me something I might be interested in.

Based on this type of usage, here is how I would describe both apps:

X/Twitter:  ALL TRANS, ALL THE TIME.   (By which I mean, seemingly half of the tweets I see are about the trans issue in an aggressively anti-trans, full blown moral panic, and/or ridiculing, kind of way.  To paint a complete picture, I would say 40% is anti immigration, MAGA stuff, about 5% is Musk's personal contribution to conspiracy world, and a mere 5% or less is stuff I am happy to see from the handful of centrists who have declined to jump ship.)   Trump could be machine gunning suspected illegal immigrants while incarcerating the entire Democrat side of congress, and X users would not care, as long as that really ugly transwoman doesn't get to use a female toilet.  As a culture war issue, it's still running as hot as hell, if X's algorithm is any guide.     (I know, maybe it isn't an accurate guide, as Musk can't keep his dirty fingers off the algorithm, hating transactivism as he does because of the son/daughter who has rejected him.  Or is it purely my fault for occasionally clicking on a trans tweet?  I don't think so, somehow.) 

Even for someone who shares the view that the trans issue definitely warranted a correction - and (to be truthful) finds some of the ridiculing posts funny (because it is hard to believe what some men who think they are "real" women think looks "sexy") - the "all trans, all the time" attitude at X just feels exhausting after all these months/years of such content.  

Bluesky:  Still suffers from weird outbreaks of "LOOK AT MY CUTESY ARTWORK" posts from time to time.   When I first started using it, I noted that the Discover feed was overwhelmed by pet photos, particularly of cats.   Since then, by blocking or noting that I do not want to see of that type of content, I have stopped that happening, but for some oddball reason, every few months I will get a burst of people posting their latest artwork, usually in a quasi-anime or fantasy illustration style.  I have never shown interest in such content, but the algorithm "tries it on" for some bizarre reason, and I have to spend a couple of days of "do not want to see content like this" marking of the posts.

That said, the "death of Bluesky" seems vastly overrated to me - if you "curate" your interests and posts adequately,  it broadly works in a decent way.   (And although I forget to check them much, the individual topic fees, on "Science" for example, can be pretty good.)

I get that the app may be the home of the chronically "woke" Lefty activists demanding this and that, which gives the place the feeling of lack of balance in the same way that X does in the opposite direction.  But as I say, they can be avoided and not interrupt one's feed much.  It's certainly easier avoiding aggressive Lefties on Bluesky than it is to avoid mad MAGAs on X.

 

I guess what feels missing from using both of them is an app for the "mainly centrist" users who would like to see something like the old usefulness of Twitter - where you knew that you could get good links to news report, and local input on events (even something like big weather events as they are happening) - relatively reliably and quickly.   Neither X (personal Muskian grudge machine) nor Bluesky (too far "woke" Lefty activist haunt) has that same functionality that Twitter used to, and I still miss that somewhat. 

   

Old scrolls, discussed

Stumbling around the internet, I just found myself reading about the earliest dated physical copies of the Buddhist scriptures.

For those who have been following (hello?), I find more appeal in the Mahayana branch of Buddhism, but know that it was a development out of earlier Buddhism as that is generally seen as defined by the "Pali canon".  

I had always assumed that the earliest examples of the Pali cannon had been more or less accurately dated, and that they would come from well before the dating of the earliest Mahayanan Sutras.

But no, it seems this is not the case at all.  First, remembering that the rough date for the birth of the "original" Buddha is about 560 BCE:

a.    despite the belief that the Pali texts were first committed to writing, rather than oral tradition,in Sri Lanka around the first century BCE (that is, like, 400 years after the death of Buddha, give or take!), it seems the earliest fragments of Pali texts come from around 550 to 650 CE in Burma.  (A full 1,000 years after its apparent founder.)  

b.    the Mahayana texts, however, have substantial bits dated to around 200CE:

The earliest accurately dated Mahayana texts are those translated by the Indoscythian monk Lokakṣema in the 2nd century CE, who came from Gandhāra to China. These include about ten Mahayana sutras translated before 186 CE, which constitute the earliest objectively dated Mahayana literature. Modern scholarship generally agrees that Mahayana sutras began to appear between the 1st century BCE and the 1st century CE, but the earliest physical evidence with firm dating comes from Lokakṣema's Chinese translations around the mid-2nd century CE. These texts show developed Mahayana doctrines and literary style, indicating a prior phase of composition stretching back possibly a century or more before their translation.

This brief article on the Gandhara scrolls contains this surprise, too:

The scrolls contain linguistic treasures as well. They’re written in Gāndhārī, a vernacular derivative of Sanskrit; Kharoṣṭhī, the script used to actually set down the words on birch bark, derives from Aramaic, according to Salomon. 

There's also an article on line* that explains the (quite possibly mythical?) origins of the motivation for conversion of oral transmission to written:

Buddhist chronicles in Pali tell us that the hitherto recited Buddhist texts were written down in Sri Lanka in the early first century B.C. when a severe famine occurred.
As the population diminished, the Buddhist order feared that Buddhism would perish, for once monks who memorized and recited those texts were dead, the texts
themselves would disappear. Thus the Buddhist order wrote down the texts in order
to preserve them for posterity.2 Another legend says that King Kaniska (the late first
or early to middle second century A.D.) ordered Buddhist texts to be inscribed on
copper plates and kept in a stupa in Kashmir of northern India.3 Yet the written or
inscribed Buddhist texts themselves are not extant. Incidentally, Chinese Buddhist
texts were also inscribed on stones, which are much more durable than paper (the
usual writing material), as a precaution against the disappearance of the Buddhist
doctrine. Such a precaution was believed necessary both because Buddhism was
persecuted several times in China, and also because the mo-fa theory, which
held that the Buddhist teachings would soon disappear, became widely accepted. 

It also mentions that the Mahayan sutras were more likely to be committed to writing recorded from the start:

 Incidentally, Mahayana sutras, which are thought to have been gradually formed from about the first century B.C., are presumed in many cases to have been written down from the beginning because the texts themselves instruct the readers to copy them. In fact,
abundant copies of some Mahayana sutras are preserved in manuscript form in
Nepal. However, the Indic texts of other Mahayana sutras are not extant at all.

 There is also mention of how Indian emperor Asokha, around 260BCE, converted to Buddhism and had mention of it carved on pillars, etc, but did not apparently see the need to set down the texts in full.  

The estimate date of writing the scrolls aligns very closely to when the New Testament books were being written**:

Chapter 6 describes the palaeographic and linguistic features of the manuscripts.
The palaeographic evidence indicates that the manuscripts were probably written
between the earlier part of the first century A.D. and the age of Kaniska (the late first
or early to middle second century).  

So, it's a curious thing that the later branch of Buddhism actually has fairly substantially older original documentary evidence.     (And the fact that there was a lot of religious "commit it to writing" going on around the same time.)

Update:   I still don't understand how we know that the Pali canon was first committed to writing in Sri Lanka around the first century BCE, if we don't have manuscripts from then.   Let's see what AI (in the form of Perplexity) says:

Here is how scholars and tradition establish this dating despite the absence of original manuscripts: 

👉Oral transmission: The early Buddhist teachings were preserved orally for several centuries, passed down through meticulous memorization and communal recitation by monks. This oral tradition was considered highly reliable and central to preserving the teachings exactly as taught by the Buddha.
 

👉   The Sri Lankan chronicle Mahavamsa records that the Pali Canon was first committed to writing in the 1st century BCE to preserve it from loss due to famine, war, or decay. This event is linked to a formal Buddhist council held in Sri Lanka.

 👉 Linguistic and historical context: Scholars date the Pali Canon partly by comparing the language style to related ancient Prakrit languages and by cross-referencing historical events mentioned or implied in the texts with known timelines.

 👉  Archaeological evidence shows that the oldest surviving Pali manuscripts date much later, from around the 5th century CE onward, often on palm leaves or other materials which do not survive well over more than a millennium.

👉The notion of a fixed Pali Canon from 100 BCE onwards is supported by continued manuscript copying and commentary traditions in Sri Lanka and Southeast Asia, and references back to this origin in later texts.   

 

So, yeah, I knew for a long time that compared to Christianity, Buddhist had quite the "historicity" issue; what I didn't realise was that within the various Buddhist traditions, the apparent later branch had the earlier physical documents.

 

* it's downloadable but I can't find the link to it now - it's called  REVIEW ARTICLE
The Discovery of “the Oldest Buddhist Manuscripts” ENOMOTO FUMIO 

** the dates now believed for the writing of the various New Testament books range from around 50 to 125CE.   

Rugby league, noted

While I usually watch only State of Origin (and then, not always) matches and the Grand Final (at least if a Queensland team is involved), I have to say that the Grand Final on Sunday confirmed my theory that rugby league has the right amount of scoring and skills, and tension, to make it the most easily enjoyed football code.   Soccer has a pathetic rate of scoring; AFL has activities going on all over the place and scores too often, and rugby union has stupid periods of pile ons where you can't even see the ball or know what is going on.

League has some aspects that kind of don't make sense, but overall, it can make for a very enjoyable spectator experience.  

Recommendations - Singapore

I've been back home for a few days, but have been feeling a bit sick, hence the lack of posts.

I should do a post about how well I now know Singapore, with photos, but the photos will come later.

Some random things learnt this time:

*  Katong remains my now preferred area to stay in - a somewhat hipster, upper middle class (I would guess) area that is easy to use as a base to see the rest of the island because of the new-ish Thomson East-West MRT line which currently ends close to the airport but will, in a few years, join it.   (There are also plenty of buses that run through the main road of Katong.)  This trip, I used my Visa card to catch the MRT from the airport to Marine Parade station (following a route which involved two changes of lines and took maybe 50 minutes) and it cost all of $3.60 AUD.   The bus from the airport is a much more direct route, but I have to say, the previous trip I found it was quite a challenge (even using Google maps) finding the correct bus stop, at the correct terminal, for the ride to Katong.  (The public transport buses all run from basement type stops, and maybe it was my incompetence, but it was not easy.)   The MRT is, by comparison, very easy to find and provided you've downloaded the MRT network map to your phone, it's easy to work out where to change lines.  (There are several apps which feature the MRT map.)

*  I did, this time, however, buy myself a 3 day unlimited travel pass for $27SGD.  I would guess that if you are going to make 4 or more MRT or bus trips in a day, the pass is no more expensive than paying for individual trips, but the other reason it is now a more attractive thing to use is that you no longer have to go to one of the limited number of stations where the old tourist pass used to be sold, bring your passport  and pay a $10 refundable deposit (and then remember to claim the deposit, by again going to a ticket office at a limited number of stations to hand in the card.)  Instead, the card can just be used as a rechargeable fare card when it has expired - although whether there is any particular advantage to doing that instead of just tapping on and off with your credit card is not clear to me.   

But overall, this change is a significant improvement over the old system.   

* I have stayed at two very good hotels at Katong now - Santa Grand Hotel East Coast and Village Hotel Katong.   The Santa Grand is cheaper, with rooms sometimes available at under $200 a night, and it doesn't have a bar or room service (I don't think), but Katong is full of coffee shops, bakeries, bars and a range of cafes covering Indian, Chinese and I think even Peranakan food.   There is an old Hindu temple in a side street, which I think explains the Indian food options, even though it is not close to Little India.  In fact, I ate twice at an Indian cafe where the mains were under $20, and the total meal and drinks cost was around $25 and $30.   Hence, even without going to a Hawker Centre, eating in Singapore does not have to be expensive.

*  Speaking of eating, a place I went to because I saw it discussed on a couple of Youtube channels was the Fortune Centre which is pretty close to the busy shopping area of Bugis.   It has become almost entirely a multilevel food outlet, with lots of vegetarian (because it is close to an old temple) but also meat options. Apart from some Chinese options, it's got everything from Japanese to Korean to Vietnamese food, and it all seemed pretty moderately priced and is in air conditioned comfort.  I ate at a vegetarian place, actually, and it was very good.

*  Although eating at a hawker centre is seen as a quintessential tourist experience, I don't usually go to them for a couple of reasons:   I don't care for eating off plastic plates or bowls; they are rarely enclosed enough to be air conditioned; and the service can be a tad intimidating, especially if it is a busy stall.   I tell people that a busy stall can make you feel a bit like the experiencing the Soup Nazi episode of Seinfeld.   So, yeah, if you're visiting first time you might like to try eating at a couple, but it's not essential as a means to eat cheaply.  (I also think that the hawker staple - chicken rice - is a vastly overrated dish and I don't bother eating it after having tried it a couple of times.)

*  As for places to eat some hawker style dishes but in airconditioned comfort, you can always look in the basement of shopping centres.  They quite often will have an area of stalls that do all of the main things you can expect at a real hawker centre.  Even up market centres like The Jewel shopping centre at the airport, and Tang's department store in Orchard Road, do this semi-hawker centre style thing in their basements.   They will be, no doubt, a bit more expensive than at a real one, but the experience may be a bit more pleasant.

*  This trip, I finally made it to Sentosa Island, the very touristy development with beaches on which hardly anyone swims, but it was still interesting and good to see.  I was especially lucky to stumble upon a free guided tour of Fort Siloso (only run on weekends, I believe) - and the fort and its museum style rooms turned out to be bigger and more interesting that I expected.  (The rooms devoted to World War 2 and the Japanese occupation were particularly good, and taught me thing I didn't know.  It will get its own post.)

* I also finally made a trip to the observation deck on Mariner Bay Sands.   At over $30, it is a tad expensive, but once there, it's hard to stop taking photos, and you can buy drinks at a reasonable price, and watch other tourist take a lot of posing selfies.   It is good to get to the top of one of the most iconic buildings in the world, so I do recommend it.

That's all for now! 

Updates:

*  try the Toast Box franchise for kaya toast breakfast ($6.20 for two eggs, one biggish cup of coffee and the kaya toast) or a large of other modestly priced eating for lunch.   I thought it better than the Ya Kun Kaya Toast chain.

*  Katong has several multilevel shopping centres, but the largest is on the south side of Marina Parade MRT station and is called Parkway Parade.   And right beside the entry to it is an old area of "local" shops that your often find around HDB areas - a hawker centre, and thrift style shops, fruit and snack shops etc.   It thus features both styles of Singaporean shopping in a very compact area.  

 

   

Ultra processed?

I've been meaning to say this for ages:   categorising foods as "ultra processed" has always sounded a bit vague, almost a tad faddish (like "high protein" appearing on every third item in the supermarket).   But perhaps I'm wrong, and there is some clear definition somewhere?  

Even if there is, I have this suspicion that it's too broad to be useful.  I mean, shouldn't they be looking more carefully at what makes it "ultraprocessed" that could be the core of the problem?   Preservatives or other chemicals that feature prominently in such foods to ensure their longer shelf life, for example?  

Anyway, such thoughts are prompted by this article in the Guardian, pointing the finger at ultra processed food for the rise in younger people getting colon cancer:

One factor keeps emerging as a major culprit: ultra-processed foods, with a 2025 review in Nature Reviews Endocrinology highlighting the links. These foods, roughly defined as factory-packaged snacks, ready meals, sugary cereals, soft drinks, processed meats and many fast foods, now make up more than half the average diet in countries such as the UK and US.

The evidence is growing: a major study published in the British Medical Journal looked at three large US cohorts to examine the association of ultra-processed foods with the risk of colorectal cancer. One of these cohorts involved more than 46,000 men, tracked over 24 to 28 years. Compared with the group that consumed ultra-processed food the least, the risk of those who consumed the most of developing colorectal cancer – even accounting for nutrition and weight – was 29% higher. The authors conclude that further studies are needed to understand the physiological mechanisms of how exactly ultra-processed foods contribute to cancer development. 

Now that I think of it - wouldn't it be funny if it turned out that adding protein powder to everything was actually the explanation! 

Travelling again

Let me bore readers, as I do the people I live and work with, by once again extolling Singapore as a destination.

Last night, in one side street of Katong, I was taking a post late dinner* stroll down a side street which featured this house with its traditional (I assume Taoist) Chinese alter clearly visible along with photogenic cat):

Then perhaps 10 m further down the road is the back of a Hindu temple built in 1875, with some guys still in the hall where they serve food.  Barely another 10 m was a house with a prominent Islamic symbol above the carport, and then just a little further was St Hilda's Anglican Church, a relative latecomer from 1934.

Earlier in the day, I used my unlimited travel pass to ride the MRT subway for an hour to the north of the island to visit a Buddhist monastery, just on a whim because its website had a fair bit of English and it sounded a bit missionary in intent.  The building is not old, the decoration inside is a bit on the gaudy side, but there were two staff who said photos were fine, and one middle aged guy praying:

 

He finished, then approached me, and spent a good half hour chatting about the temple, the statues and Buddhism generally.  One of the staff gave me a juice to drink.  He was very surprised that I had come to this rather far flung (for Singapore) part of the island just to visit a temple that is not on the tourist trail.  (When I was telling my daughter about it, she said something like "well, you're not exactly like a tourist anymore, you've been there so often." Ha ha.)

Anyway, the day made me fully realise something that I probably always knew at a subliminal level: as a person who started reading about comparative religions in my early 20s, and in recent years had interest revived by Religion for Breakfast on YouTube and a desire to better understand Buddhism and SE Asian religious syncretism generally, it's no wonder I have always been drawn to a small city state that has accommodated (like no other place in the world I can think of) such successful religious multiculturalism.  

That is all...

* Dinner was Indian, which I had twice this visit in a small cafe with Indian/Sri Lankan staff who have both an Indian meal menu and a separate menu of Chinese cafe staples.  I would normally avoid a place that has alternate cuisines,  but their food was very good.

Overrun by bots and crawlers?

I'm guessing that the rise of AI and LLMs might have something to do with this.

I've long known that the statistics on Google Blogger give no accurate reading of how many people are visiting the blog, because I also have a Statcounter account that gives me more detailed and accurate information indicating that I have long had "unique views" in the hundreds, perhaps, in an average week?

Google will show massively higher numbers, but last week I happened to look at its stats and saw a new record for this:

 A single day with half a million hits?   On Statcounter, it shows 20 (yes, twenty!) unique hits on 21 September.

It's been known for a long time that the issue is that Google counts webcrawlers and bots hitting sites as if they are "real" hits.   And I would guess that, given LLMs' voracious appetite for writing, it's not unreasonable to think that a 20 year old blog with a lot of posts may be a target of interest for them.  

I don't really care if this is the explanation - in a way, it feels like my own modest contribution to the rise of the AGI that will replace human intelligence in the future!   

Or, of course, I could be wrong and this is explained by something else... 

Back to Kimmel

Of course I watched the Jimmy Kimmel return, and then have been reading some of the MAGA reaction on Twitter.  (Not spending a lot of time on it, I have better things to do.  But still...)

Here's the thing about MAGA and its apologists:   they are so often people who are some combination of dumb and/or argue in bad faith continually.    This bad faith makes it impossible for reasoned engagement with them by someone like Kimmel - they invite polarised ridicule because they deserve it.

Hence we have most of them arguing that the Right wing "cancel culture victims" was as big a free speech issue as the head of the FCC (and the President himself!) warning companies that they should take off the air comedians who don't support Trump enough, or they may lose their licence.   Then people like the Vice President have the gall to say "see, he wasn't cancelled after all.  What's the problem here, people have been blowing this out of proportion."

MAGA types also conveniently ignore successful Right wing cancel culture campaigns.  Anyone heard of Dylan Mulvaney much lately?   Sure Roseanne Barr got cancelled - against ABC's financial interests - but MAGA memory holes the reason why.   

Other examples of  stupid arguments and memes:  that only one hundred and something thousand people had been viewing Kimmel's show - when that figure was referring to just one particular demographic, and the simplest Google would show that the true number of viewers was well over a million a night, as well as having a Youtube audience of 20 million subscribers and some videos getting millions of views.    But one MAGA person who "never found him funny" comes up with one dubious figure, and they all run with it.  Even stupid Piers Morgan - man, he is obnoxious:

The lines they run are so often so easily rebutted, the fact they don't check before repeating them just shows a total lack of good faith and reasonableness. 

No, the successful return of Kimmel is a glimmer of hope that enough Americans recognise the importance of constitutional free speech (even Ted Cruz, who Kimmel specifically thanked last night.)  

And yeah, I do admire Kimmel for his passionate criticisms of a side of politics that deserves it.  

China and "involution"

An interesting opinion piece here at the New York Times: The New Buzzword That's Scaring China 

Here's the start:

 Competition in China is often far more cutthroat than in the United States. America has a handful of carmakers; China has more than 100 electric vehicle makers struggling for market share. China has so many solar panel makers that they produce 50 percent more than global demand. About 100 Chinese lithium battery producers churn out 25 percent more batteries than anyone wants to buy.

This forces Chinese manufacturers to innovate, but it also leads to price wars, losses and bad debt — and that’s becoming a problem.

China is heading toward deflation, the often catastrophic downward spiral of prices that sank Japan in the 1990s. Its leaders are blaming a culprit they call “involution” (“neijuan” in Mandarin), a term that has come to mean reckless domestic competition. They want to rein it in by browbeating companies into keeping prices steady and instructing local governments to scale back subsidies.

It won’t work. At best, those are temporary fixes for China’s more fundamental problem. Its economy relies so heavily on investment for growth, rather than consumer spending, that it produces enormous surpluses that wreck profits at home and provoke trade wars abroad.

As to where the term even comes from:

China’s infatuation with the term “involution” dates to the 1960s and the work of an American anthropologist, Clifford Geertz, who argued that Indonesia wasn’t able to feed itself because population growth had outpaced improvements in agricultural productivity. Geertz used “involution” — an anthropological term for a culture that fails to adapt and grow — to describe this doom loop. His analysis resonated in a China that at the time was struggling to feed its people, the world’s largest population. 

The article goes on to explain that China is trying to address the problem, but the writer says the current interventions are unlikely to be the solution.  The long term problem is this:

What China needs, more than political campaigns, is more domestic spending, which in turn would gobble up more of the excess supply. Western officials and some Chinese economists have made this recommendation for years, but China has resisted. Private consumption accounts for about 40 percent of China’s gross domestic product, compared to about 69 percent in the United States and 53 percent in manufacturing-heavy Germany. That’s in part because Chinese households save heavily to compensate for a skimpy social safety net. 

 This all sounds like sound analysis to me... 

The dubious exercise of assigning motive to high profile killings

I've watched two good videos and read one essay in the New York Times which make excellent points about political violence in America.

First, the essay by one Matthew Walther, with the title "Why do we think we know Kirk's shooter's motive".  Some extracts:

... the charging document suggests a relatively straightforward political profile and motive, especially when compared with the cryptic messages the shooter engraved on his shell casings, which were frantically mined for meaning in the immediate aftermath of the shooting. Those inscriptions, the shooter told his lover after the killing, “were mostly a big meme.” He said that if he saw one of his scatological jokes mentioned on Fox News, “I might have a stroke.”

Perhaps it’s true that the opaque messages were a joke, from which his true intentions can be clearly distinguished. But when it comes to a person like this — that is to say, a young man who reportedly spent a great deal of time holed up in his apartment playing video games and using niche social media programs — I confess I have my doubts. I wonder if a legible political motive can neatly emerge from the fragmented, self-parodying, endlessly reflexive world of perpetually online discourse.

It is easy enough to imagine that this young man was radicalized. But it is also possible to see his radicalization, if that is the right word, as something post-political, a simulacrum of motive in a fantasy world...

He then goes through the meme-y messages left by the killer, and notes:

These inscriptions are the quintessential stuff of online gamer-style discourse: fragments without context, seemingly private jokes, missives designed not to persuade or even to be broadly intelligible but simply to circulate. In insular internet worlds, this style of communication is the point. And it produces an epistemic fog that can obscure the meaning of even the most intentional of gestures. ...

This is why it seems to me premature, at best, to speak of Mr. Kirk’s killer as if he were a left-wing militant. Say what you will about the members of the Weather Underground — their theories of revolution were facile, their moral compasses obviously malfunctioning — but at least their relationship to public life had a recognizable shape. A few middle-class college graduates wanted to become Marxist revolutionaries, and so they did.

Here's a key paragraph:

In the online world in which Mr. Kirk’s killer was steeped, heedlessness is rewarded. Half-thoughts are quickly replicated. It would not surprise me, in this context, if for some troubled individual, killing were to seem indistinguishable from posting — the ultimate trashpost, meant to be endlessly circulated, reinterpreted, willfully misunderstood, joked about, heartlessly recontextualized. 

Bear that paragraph in mind when watching the excellent commentary by Matt Bevan in his podcast when he talks about the young guy who tried to kill Trump. 

Here's the whole thing, and it's great: 

 

Bevan spends a lot of time noting the way the Right has reacted each time by endlessly repeating that "they" want "us dead"; a self serving rallying call that only serves to further polarisation and division.   (It is, of course, all in line with Trump's quintessentially authoritarian political tactic of continually claiming that those who oppose him are "bad", "evil" and deserve jail.)

Finally, while I certainly do not like Bill Maher as a personality, I still occasionally look at his takes on the current situation to see what line he is taking.   In this clip, he gives a useful summary of the general nutty, all-over-the-shop politics of several recent killers or assailants of political targets.   They rarely are coherent, and he decries the time wasted on trying to pin them onto one side or the other:  

 

While I think he makes a good case, there is no doubt that some shootings can legitimately be labelled as motivated by extremism of the Left or (more likely, as the FBI and researchers have been saying for years) the Right.   This is especially the case when the target is a group - a government building, a black church, or a gay bar, for example.   But when the targets are individual politicians, that is when it seems more common than not that the killer is a mental mess of one kind or another, and the motivation is often never clear or well rationalised.     

Update:  Since writing this post, I have learnt that even Karl Rove (!) has written an opinion piece attacking the Right's use of the blanket "they".  I assume I can't read it all at the WSJ, but here are some extracts from The Independent:

Longtime Republican strategist Karl Rove slammed those looking to politicize the death of conservative activist Charlie Kirk, stating that using it to justify retaliation against political rivals is “wrong and dangerous.” 

Conservatives have been pointing fingers at liberals ever since Kirk, 31, was shot and killed while speaking at Utah Valley University last week. Authorities have charged 22-year-old Tyler Robinson with the killing, without indicating anyone else was involved.

President Donald Trump has also placed the blame on liberals, claiming, without evidence, that “most of the violence” is on the left. While the charging document against Robison says he “intentionally” shot the right-wing influencer due to a “belief or perception regarding Charlie Kirk’s political expression,” prosectors stopped short or providing a specific motive.

Despite this, right-wingers have followed suit with their blame game, claiming that an unspecified “they” is responsible for Kirk’s death – something Rove referred to as “a disturbing and growing undercurrent in our national conversation.”

In a Wall Street Journal op-ed, Rove noted that there is a a “pronounced emphasis on ‘they’ and ‘them.’”

“Charlie would be alive but for ‘them.’ ‘They’ killed him. ‘They’ are responsible for his death. ‘They’ must pay,” he noted.

“No. Charlie Kirk wasn’t killed by ‘them.’ ‘They’ didn’t pull the trigger. One person did, apparently a young man driven by impulse and terrible hate,” Rove continued.

“If there were a ‘they’ involved, law enforcement would find ‘them’ and the justice system would hold ‘them’ accountable. But ‘he’ and ‘him’ are the correct pronouns for this horrendous act,” wrote Rove, who was a senior adviser and deputy chief of staff to President George W. Bush. 

Rove later notes, “We aren’t helpless automatons whose actions are dictated by others.” 

“Using Charlie’s murder to justify retaliation against political rivals is wrong and dangerous. It will further divide and embitter our country. No good thing will come of it,” he wrote, adding, “It is also an insult to his memory.”

Rove concluded: “Above all, it needs to be repeated. Violence has no role in our country’s politics. Now or ever. Reasoned discourse is essential to our democracy. Charlie Kirk understood that. Let’s hope it’s a message his eulogists honor.” 

After a lone gunman shot Trump in the ear last summer, Kirk uploaded a YouTube video titled: “They tried to kill Trump.”  

Wow:  that last line.   

 

 

Regarding Kimmel

Regarding Robert Redford

I can't say that (unlike my late mother) I was ever a huge fan of Robert Redford as an actor.  If you ask me, there was something that felt a bit "surface" about his acting - perhaps being too traditionally handsome is a hindrance to feeling fully convinced by an actor.   Now that I think about it, this may be the same reason I have never warmed much to Brad Pitt, too.   But as for one more "classically handsome" star who I did like routinely - Cary Grant - I think that the lightness and charm of his persona in nearly all roles was what made it irrelevant that he was, as they say, always Cary Grant in his movies.

Anyway, regardless of not being one who was particularly keen to see him act, I have to say that being reminded of the kind of movies he was in (or made) makes me miss the general milieu of "serious, well made and well-intentioned American drama" which seems to have disappeared from Hollywood in the last decade or so.    And of his lighter content, it reminds me of how Hollywood seems to have lost the ability to make engaging romances, too.   

Of the films he directed, I would have to say that my clear favourite was 1994's Quiz Show.  I remember being so impressed by it at the time it came out, but it seems not to have lingered in the collective memory to the extent it deserves.   Maybe I will re-watch again soon to see if my initial reaction stands up.    

The most appalling President

Nice of one of the mainstream media outlets to say it clearly, but it was so obvious anyway:

The first few minutes of President Trump’s Oval Office address after the assassination of Charlie Kirk last week followed the conventional presidential playbook. He praised the victim, asked God to watch over his family and talked mournfully of “a dark moment for America.”

Then he tossed the playbook aside, angrily blaming the murder on the American left and vowing revenge.

That was stark even for some viewers who might normally be sympathetic. When Mr. Trump appeared later on Fox News, a host noted that there were “radicals on the right,” just as there were “radicals on the left,” and asked, “How do we come back together?” The president rejected the premise. Radicals on the right were justified by anger over crime, he said. “The radicals on the left are the problem,” he added. “And they’re vicious. And they’re horrible.”

Mr. Trump has long made clear that coming together is not the mission of his presidency. In an era of deep polarization in American society, he rarely talks about healing. While other presidents have typically tried to lower the temperature in moments of national crisis, Mr. Trump turns up the flames. He does not subscribe to the traditional notion of being president for all the people. He acts as president of red America and the people who agree with him, while those who do not are portrayed as enemies and traitors deserving payback. 

 Will Saletan has a video out making the same point, but with some additional Trump material: 

 

The particularly interesting thing about that Trump on Fox News clip is that he prefaced his excuse making for Right Wing radicals (and emphasis on the real problem being the Left, who are evil and awful in every conceivable way) by saying he would "probably get into trouble for saying this". He may be an idiot, but is not such an idiot that he doesn't know that he should be telling both sides to walk away from violence, but he chooses not to. 

This from a man who made a "wink wink" comment in 2016 about his guns rights base and Clinton:

In North Carolina on Tuesday, Trump said that if Clinton were elected she would “essentially abolish” the Second Amendment.

He continued: “By the way, if she gets to pick her judges, nothing you can do, folks. Although the Second Amendment people, maybe there is, I don’t know.” 

 People in the audience laughed - they knew what it meant.

The same President who make actual "jokes" about Nancy Pelosi's husband getting his head bashed in (and all of his supporters had no problem with that.)

Not to mention inciting a riot on the same fraudulent grounds that Charlie Kirk promoted and then excusing all of those convicted for participating.

This President actually encourages, or makes light of, or legally excuses, political violence when it suits his side.   His ranking amongst "worst presidents" deserves to be at the very bottom. 

 

Crazy years, continued

Seems to be something about my birthday that attracts momentous news - not that (hopefully!) Charlie Kirk's assassination is going to be quite a big as this, but it was late night on my birthday 24 years ago that I watched live as the World Trade Centre and thousands of its occupants were destroyed in a world changing event.

Truth be told, I never followed Kirk very closely, but the thing about his death is that, of course, before anything is known about the mental health or motivations of the shooter, his followers are leaping on it to further confirm the problem that I have been saying for years is at the heart of American political polarisation - that the Right has chosen to label the Left as literally evil, and believe it in a fully fledged religious reality sense.   Hence, Musk's toilet (as I chose to call X) is brimming with comments like this on the Right:

(I have no idea who Goddek is - and he may not even be particularly religious, as he says he's a libertarian - it's just an example of the thousands immediately labelling the Left as evil.)

Of course, all political violence of this kind is wrong and immoral.   And there are going to be those online on the Left who inappropriately make light of the killing.   You know, a bit like how Charlie first made light of Nancy Pelosi's husband being smashed in the head with a hammer by a nutter claiming he was after Nancy:

“Why has he not been bailed out?” Kirk said Monday on his podcast of the man who allegedly beat House Speaker Nancy Pelosi‘s husband Paul with a hammer last Friday. “By the way, if some amazing patriot out there in San Francisco or the Bay Area wants to really be a midterm hero, someone should go and bail this guy out, I bet his bail’s like thirty or forty thousand bucks.” With a smirk, he added: “Bail him out and then go ask him some questions.”

(He was playing up to the completely unfounded rumour that the attacker was a gay lover of Pelosi.)

Here, by contrast, is Nancy Pelosi:

"The horrific shooting today at Utah Valley University is reprehensible. Political violence has absolutely no place in our nation," wrote former house speaker Nancy Pelosi.  

Kirk was also adamantly pro-gun rights and thought that gun deaths were just something you had to put up with because he valued an armed citizenry.   

Perplexity also helped me find that he defended the blowing up of a Venezuelan boat in international waters because it was suspected of carrying drugs.  No proof, just support Trump all the way in extra judicial killings in international waters.

In short, Kirk supported policies that endorsed, engaged in joking excuse of, or helped enable, violence in his country when it suited his political views.   

Which is not to say that shooting him was the answer.   But his followers should have a serious look at themselves when wondering how a violent society is established.  (It won't happen, I know.)

On a final note:   it's a "lucky" thing that it happened in conservative Utah.  Can you imagine how Trump would have reacted if it was Chicago?   The place would be under National Guard occupation  for a year or more, irregardless of legalities.  
 

Update:   This is right.   But will the Murdoch's do anything to tell his network to cool it?  Probably not, if there's money to be made. 

 

 

Widespread disruption

Tomorrow's my birthday, and it feels like I've got to mentally travel back to the 1970's to feel the same level of unease, war, disruption and uncertainty for the globe as we have now.    (That said, I was a teenager for most of the 70's and distracted by high school, a still interesting space program and enjoyable science fiction, and as such still felt somewhat optimistic for the future.  But I was aware of the level of global unhappiness - much of it still coming from the exact same places as now!)   

I mean, Israel attacking buildings in Qatar (even making the Trump administration uneasy!); deaths in anti-government rioting in Indonesia and Nepal (with the later partly being put down to proposed social media bans!); I haven't checked what armed conflicts are "hot" in Africa at the moment, but there are bound to be some; Trump and his coterie of idiot advisers who think a global trade war and trashing decades of international relationship building is a good idea;  Elon Musk using his social media outlet to encourage a race war (and openly expressing regrets about white people not reproducing enough to keep up with the non whites - yes, he is an awful man);  American going down a anti-science path not only on climate change, but the previously non-controversial issue of vaccines; Britain still suffering from its self inflicted populist wounds and stupid use of police in culture wars over gender and sexuality;  illegal immigration from badly run countries causing social unrest everywhere;  Russia carrying on a pointless war because Putin refuses to do the right thing and die.   All we need is an abrupt stockmarket crash and for China to make its move on Taiwan, and it would round off a "perfect storm" of things that could go wrong.

I often think of the Robert Heinlein's Future History with it's "crazy years" and how appropriate a label that appears to this current unfortunate era - and oddly, it seems figures both on the Left (science fiction author David Brin - well, I assume he's mainly Lefty) and the Right (blogger Glenn Reynolds) have both written years ago - around the rise of Trump - about how Heinlein was prescient!   Obviously, each thinks the other side is the one causing the "craziness":  perhaps I should just read that as confirming the craziness is deep and not going away anytime soon.

Of course, one should remind oneself that there are respects in which the world is still in a much better place than in the 1970's:   treating people as people, regardless of location and colour (hello, Elon), the decline of poverty in China and India is a positive.   Ignoring the dire risks of global warming for a moment, there's no doubt that environmental issues are otherwise taken seriously now in a way they weren't in the 70's.  (Of course, Trump is doing his best to undo that...because he's an idiot.)   

Anyway, there must be other "global positives" that I need to think about.   Help me, anyone?    

It's just hard to believe how pessimistic we can now feel, with quite a bit of justification...  

David Brooks being somewhat perplexing, again

David Brooks never seems to be able to get through a whole column without saying something dubious.   His latest is interesting, though:  Why I am not a Liberal.

I am somewhat sympathetic to the main line of argument, as it happens:

Last May a study came out suggesting that merely giving people money doesn’t do much to lift them out of poverty. Families with at least one child received $333 a month. They had more money to spend, which is a good thing, but the children fared no better than similar children who didn’t get the cash. They were no more likely to develop language skills or demonstrate cognitive development. They were no more likely to avoid behavioral problems or developmental delays.

These results shouldn’t have been a big surprise. As Kelsey Piper noted in an essay for The Argument, a different study published last year gave families $500 a month for two years and found no big effects on the adult recipients’ psychological well-being and financial security. A study that gave $1,000 a month did not produce better health, career, education or sleep outcomes or even more time with their children....

Further down, he points to this study, too:

Many years ago, I came across a study that neatly illustrated the power of culture. The researcher Nima Sanandaji calculated the poverty rate of Americans with Swedish ancestry. It was 6.7 percent. They also looked at the poverty rate in Sweden, using the American standard of poverty, and it was also 6.7 percent. Different political systems, same outcome.

(Presumably, one can also point to the cultural aspects of Jewish upbringing that has lead to their prominence in business and professions, too.)

And here is he key contention:

This is consistent with something I’ve noticed all my life — the materialist bent of progressive thought: the assumption that material conditions drive history, not cultural or moral ones. A couple of decades ago, Thomas Frank published “What’s the Matter With Kansas?” based on befuddlement that Kansans were apparently voting against their economic self-interest. Doesn’t economics drive voting behavior? Progressives have often argued that improving schools is mostly about spending more money, that crime is mostly the product of material deprivation.

Conservatism, as you know, is a complete mess in America right now. But reading conservative authors like Edmund Burke, Samuel Johnson, Fyodor Dostoyevsky, Gertrude Himmelfarb and James Q. Wilson does give you an adequate appreciation for the power of nonmaterial forces — culture, moral norms, traditions, religious ideals, personal responsibility and community cohesion. That body of work teaches you, as Burke wrote, that manners and morals are more important than laws. You should have limited expectations about politics because not everything can be solved with a policy....

Progressives, by contrast, are quick to talk about money but slow to talk about the values side of the equation. That’s in part for the best of reasons. They don’t want to blame the victims or contribute to the canard that people are poor because they are lazy.

But there’s something deeper. Progressivism emerges from a different lineage. Karl Marx influenced many people who are not Marxist, and he saw the world through a material-determinism lens — people’s consciousnesses are shaped by their material conditions. 

The thing I have some issue with, however, is that he doesn't address the fact that the allegedly conservative Right can (and does) use this way of thinking to justify economic policy that is clearly in the interests of the rich (using poor justification that isn't backed by evidence - such as trickle down economics); and Progressive concern over not leaving behind the poor economically is more consistent with the "community cohesion" and "conservative values" that Brooks yearns for.   

That said, I can't not agree with the suggestion that Left social ideas can lead to a fragmented society in terms of ideas about how a good life is lived; and at the moment, the more communitarian social structure of Asia nations (or perhaps, East Asia in particular) seems to be very appealing compared to the "freedom to live anyway you want" of the West.  And what's going on in Indonesia at the moment doesn't really support Brooks, either!

I do therefore agree that Democrats (and Left parties of all countries) do best when they encourage shared values and unity, over emphasising the rights of individualism in terms of culture (such as promoting identity politics.)   We seem to be lacking that in Left wing leadership at the moment.  

Anyway, one day I will have time to work out more to say about this.  Work is so busy at the moment...  

 Update:  Thinking a bit more about it, perhaps my issue with Brooks' column is that he is saying that values are more important than mere economic fairness of a society, whereas I would have thought the argument should be that both are important?   

   

 

AI brought back to reality

A good column by Gary Marcus about the overhype of LLMs:

The Fever Dream of Imminent Superintelligence Is Finally Breaking