Monday 1 July, 2024


A phlegmatic mutt, after a visit to the vet.

Quote of the Day

“The English way is a committee — we are born with a belief in green cloth, clean pens, and twelve men with grey hair.”

  • Walter Bagehot

(A quotation I found in my friend Bill Lubenow’s new book, Secular Foundations of the Liberal State in Victorian Britain, which I’ve brought with me to France and am currently enjoying.)

Musical alternative to the morning’s radio news

Van Morrison | Connswater


Long Read of the Day

 How to Fix “AI’s Original Sin”

Regular readers will know that one of the topics that really occupies what might loosely be called my mind is Generative AI, particularly Large Language Models (LLMs) like GPT-4, Llama, Claude, etc. I read an awful lot of commentary and argument about the technology, much of which is unhelpful or tangential in one way or another. But every so often I come across an essay that is informed, insightful and wise.

This essay on “How to Fix AI’s ‘original sin’” by Tim O’Reilly is a shining exception to the above rule. So I commend it to anyone who is interested in finding ways of managing and harnessing a powerful and important technology. Tim does it by focussing on the way the tech depends on appropriating the intellectual property of others without recompensing the owners (its ‘original sin’). And he proposes a way of overcoming that problem in a way that might be good both for society and for those who build the technology. What I particularly like about it is that in his business he actually implements the ideas proposed in the essay.

The essay is long, and clearly written but I recognise that it’s not for everyone. But I felt an obligation to draw it to your attention.

Closing the Stanford Internet Observatory will edge the US towards the end of democracy

Yesterday’s Observer column:

The arrival of the internet, and particularly the web in the 1990s, started the process of radical fragmentation that has brought us to where we are now: instead of public opinion in the Gallup sense, we have innumerable publics, each with different opinions and incompatible ideas of what’s true, false and undecidable.

To make things worse, we also invented a technology that enables every Tom, Dick and mad Harry to publish whatever they like on opaque global platforms, which are incentivised to propagate the wildest nonsense. And to this we have now added powerful tools (called AI) that automate the manufacture of misinformation on an epic scale. If you were a malign superpower that wanted to screw up the democratic world, you’d be hard put to do better than this.

Fortunately, scattered through the world (and mostly in academia) there have been organisations whose mission is to conduct informed analyses of the nature and implications of the misinformation that pollutes the online world. Until recently, the Stanford Internet Observatory (SIO) in California was one such outfit. Among other things (it was the first to out Russian support for Donald Trump online in 2016), it raised China spying concerns around the Clubhouse app in 2021, partnered with the Wall Street Journal in a 2023 report on Instagram and online child sexual abuse materials, and developed a curriculum for teaching college students how to handle trust and safety problems on social media platforms.

But guess what? After five years of pioneering research, it has been reported that the SIO is being wound down…

Do read the whole thing, if only to find the sting in the tail.


Something I noticed, while drinking from the Internet firehose.

  • AI imagines the Roman Empire Link

Cod video, but might be of interest to the ghost of Cecil B. De Mille.

 This Blog is also available as an email three days a week. If you think that might suit you better, why not subscribe? One email on Mondays, Wednesdays and Fridays delivered to your inbox at 6am UK time. It’s free, and you can always unsubscribe if you conclude your inbox is full enough already!