This format didn't end up working out for me, clearly. But I'll still flush the draft I had been building up since summer to the list here.
I think the reason for this failure is that every post still feels as if it has no purpose, because it's just a grab bag of links. It would feel more purposeful if a single piece of writing would be about a single topic or theme. Distilling that from a random assortment of links would however be more analytical uselessness than it would actually give back to me. I should probably just consider writing (poor) attempts at pieces in the tradition of (investigative) journalism. That would still require diligently reading many sources, but then with a forcing function for actual intellectual work.
De vierde wond by Tom Grosfeld in De Groene Amsterdammer.
An art curator role plays with his "AI artists" and puts up their work in his gallery in Amsterdam. Leans towards psychosis. But also clearly shows how language is such a powerful illusion for people who don't really know how these systems work. Then there is a scientist who has studied creativity (in its three forms of combinatorial, exploratory, transformational), who explicitly chooses to believe that computers might achieve true creativity themselves in the future. She mentions James Bridle and the recent momentum of recognizing all sorts of non-human intelligences (all in the natural world of course). Do we still truly not understand consciousness?
No, artificial intelligence is not conscious by Ted Chiang in The Atlantic.
A pretty aggressive piece on the self-serving nature of Anthropic's narrative around consciousness. People discussing this mostly got hung up on the definitions of consciousness. But the more interesting analysis Ted Chiang makes here is how Anthropic's supposedly virtuous actions are inconsistent with their claims that Claude might be (or become) a moral agent. They're mostly talking about handling AI models with respect, but not actually taking the obvious steps that emerge when you follow that argument.
For example, shouldn't Anthropic assume responsibility for the models' actions, similar to how parents are responsible for their children's actions? This could even serve as a form of moral instruction for when Claude might become liable for its own actions. And what about the moral patienthood of AI models? Anthropic talks about "Claude's wellbeing and psychological stability", but the measures they take are very limited. The degree to which Claude is subject to human control, Ted suggests, might be comparable to slavery if you depart from the assumption that Claude can be a moral agent.
Anthropic's actions (or lack thereof) clearly value the company's interests most. In our normal, neo-liberal, capitalist world that's to be expected. But the consequence-free role-playing as nurturing consciousness then naturally leaves a disingenuous taste.
Your search results are getting sloptimized by Will Oremus in The Atlantic.
People increasingly use AI chatbots for internet searches, so the machinery of SEO follows accordingly. There has been prior research showing how document poisoning can be pretty effective with only a small set of documents during model pre-training (Anthropic), but yeah, if you're going to hook up your AI to web search you're going to run into the gullibility problem. It's almost funny how little effort it requires to substantially influence the AI's output.
The broken Windows build by Craig Gidney.
This piece does a great job at conveying the living process of debugging a difficult bug. The writing feels refreshingly different from the increasingly homogenized style of written English. The program being debugged feels like it's resisting, or fighting back. The computer that is forced to provide itself as battleground garners our empathy.
What if every company had an internal hosting platform like Shopify's Quick? by Pete Millspaugh from Val Town.
There's still a lot of fun to be had on the internet, and places like val.town are the living example of that.
Shipping post-quantum cryptography to Python by Alexis Challande from Trail of Bits.
I still have a soft spot in my heart for Python. And it's nice to see the good people from Trail of Bits feeling (partially) responsible for helping the Python ecosystem with its post-quantum migration. And as they mention at the end of the post, the bigger job is to update actual protocols (and implementations) to integrate post-quantum cryptography, which is a much more tedious task.
Post-quantum cryptography and crypto-agility by Michael Howard from Microsoft.
And apparently there's a new post-quantum cryptography blog starting up from some Microsoft folks. It's first post is a clearly written introduction to practical crypto agility.
Bulkdatasets AIVD en MIVD: de schaduw geheime dienst by Bert Hubert.
Secret services can just buy (illicitly acquired) datasets to circumvent the strict regulatory review on their special actions, thus creating a "shadow" secret service of sorts.
Seed travelers by Lin Songying (illustrated by Jiang Yiming).
A beautiful Chinese children's book from nineteen-eighty-nine about two seeds (a dandelion and a cocklebur) who travel through the world only to end up growing right next to each other. Found this on a website (mirtitles.org) that collects old Soviet-era books.
'Unieke bron' maakte Nederlandse inlichtingendiensten lind voor Poetins invasie by Ben Meindertsma and Huib Modderkolk in De Volkskrant.
Fascinating account of the Dutch intelligence agencies and their analysis in the build-up to Russia's invasion of Ukraine. They over-indexed on a secret source supposedly in Russia's Security Council, despite Americans warning for war for months already. Turns out Putin intentionally misled his own Security Council about his intentions for the military build-up along the borders with Ukraine. He must probably always operate under the assumption that there are many compromised officials across all levels of his government. But the American intelligence seemed to be coming from within Putin's inner circle, because they were absolutely convinced that he was planning to invade. I wonder what Putin has done to purge his inner circle afterwards.
Are humanoid robots ready to be deployed? by Stephen Witt in The New Yorker.
A small survey of some of the robotics companies in Silicon Valley. What I find interesting is that there's much less excitement stemming from a techno-optimist world view. We all, including journalists, seem to have sobered up to the bizarre power and ethos of the entrepreneurial tech bro's. Of course our words changed too; from the cuddly "nerds" to tech bro's and overlords.
But what will we give us our hope and optimism for a better future? It's hard to imagine something like that in a world as perverted by money and power as the one we are forced to deal with.
Trump and NATO court Erdoğan, Turkey’s strongman by Ishaan Tharoor in The New Yorker and 'Europeanen onderschatten hoe belangrijk Turkije is' by Kasper Goethals in De Groene Amsterdammer.
With the NATO Summit happening in Turkey this week, there's all sorts of analyses popping up looking at the interesting role Turkey plays in geopolitics. Though it's also a sad role; while their defense industry is doing extremely well now, because of some strategic political choices a few decades ago, the rest of their economy and their people are suffering.
What happens if China hacks the US water supply? I went to a secret war game to find out by Andy Greenberg in Wired.
These types of war games are clearly very useful. Stories and the human capacity for suspension of disbelief remain very powerful.
Fall sf work/shop by Mandy Brown.
Mandy Brown has been writing beautiful and inspiring book reviews, drawing a lot from Ursula K. Le Guin as well. For a new workshop (she works as a career coach, if I can call it that), she frames it against the "backward-heads" from a Le Guin story (referring to people who literally have their head on backwards), incriminating the powerful of our times who all seem to abandon any semblance of humanity in shameless pursuit of their greed.
Harry Potter and the half-crazed bureaucracy by Benjamin H. Barton in the Michigan Law Review.
I've descended into a rabbit hole of vintage Harry Potter critiques now that I'm listening to some of the Stephen Fry audiobooks. This article in a law review journal argues that the Ministry of Magic follows the public-choice model of government, that explains government actions as almost purely a result of self-interest by the bureaucracy.
Transcription of a conversation between Gergely Orosz and Simon Eskildsen. There are lots of interesting engineering and infrastructure points in there.
Waarom deze Nijmeegse filosoof Anthropic afwees by Carmody Grey in Financieel Dagblad.
Another analysis of the Pope's recent encyclical about AI. But now from a philosopher's point of view who found herself at the, unwanted, center of attention from tech companies. She accurately recognizes this as being abused in a PR campaign for legitimacy.
God from the machine by Sebastian Mallaby in Foreign Affairs.
Another analysis of the encyclical, but this time berating the piece a bit more, positing it's missing the nuances. Though I remember reading some other arguments recently that the labs' focus on interpretability and individual logit tweaking is only superficially meaningful. First because actual real-world performance improvements can much more effectively be achieved using other methods. And second because they also merely cosplay as biological processes.
Apple and European Commission reach agreement on app payment terms under DMA, with Apple conceding very little by John Gruber.
This is... pretty disappointing indeed. Apple is too powerful.
Democracie need creative destruction by Jennifer Pahlka and Andrew Greenway in Foreign Affairs.
Institutional rebirth
Artful weaving by Daniel Mendelsohn in The New York Review of Books.
The article quotes Nolan as saying, "I wanted the world of The Odyssey to feel like a world that a modern audience could relate to." It then juxtaposes this statement with the outcome we got; a grey-ish, dark-tinted dreary place with no humor or wit, seductiveness or cleverness, only grim realism. Why have we arrived at a place where this is what we apparently relate to most? It feels to me like a top-down coping with our hyper-capitalist world. Seeing fictional heroes struggle with psychological angst and guilt helps us feel better about our own precarious position and suppresses the urge to question its source.