Monday 15 March, 2021
Temperamental optimists
Fishermen on the beach at Cley-next-the-Sea in Norfolk.
Quote of the Day
”This person was a deluge of words and a drizzle of thought.”
Peter De Vries
Remind you of anyone? (Hint: Carefully tousled blond with a posh accent.)
Musical alternative to the morning’s radio news
Voi Che Sapete | Marianne Crebassa | Le nozze di Figaro | Dutch National Opera
Utterly sublime nonsense. And if you have time, compare it with the 2012 Glyndebourne version.
Long Read of the Day
Margaret Mitchell’s letter about the ‘firing’ of Timnit Gebru
This letter relates, in a way, to my column in Sunday’s Observer about Google’s ethics-theatre. Dr Mitchell was co-director with Dr Gebru of Google’s Ethical AI team.
This is how it begins:
When diving deep into operationalizing the ethical development of artificial intelligence, one immediately runs into the “fractal problem”. This may also be called the “infinite onion” problem. That is, each problem within development that you pinpoint expands into a vast universe of new complex problems. It can be hard to make any measurable progress as you run in circles among different competing issues, but one of the paths forward is to pause at a specific point and detail what you see there.
Here I list some of the complex points that I see at play in the firing of Dr. Timnit Gebru, and why it will remain forever after a really, really, really terrible decision.The Punchline. The firing of Dr. Timnit Gebru is not okay, and the way it was done is not okay. It appears to stem from the same lack of foresight that is at the core of modern technology, and so itself serves as an example of the problem. The firing seems to have been fueled by the same underpinnings of racism and sexism that our AI systems, when in the wrong hands, tend to soak up. How Dr. Gebru was fired is not okay, what was said about it is not okay, and the environment leading up to it was — and is — not okay.
Do read the whole thing.
(Thanks to Sheila Hayman for reminding me of it.)
Doc Searls: the Eventual Normal
Characteristically thoughtful reflection on what might lie ahead:
There will be a new normal, eventually. It will be a normal like the one we had in the 20th Century, which started with WWI and ended with Covid. This was a normal where the cultural center was held by newspapers and broadcasting, and every adult knew how to drive.
Now we’re in the 21st Century, and it’s something of a whiteboard. We still have the old media and speak the same languages, but Covid pushed a reset button, and a lot of the old norms are open to question, if not out the window completely.
Why should the digital young accept the analog-born status quos of business, politics, religion, education, transportation or anything? The easy answer is because the flywheels of those things are still spinning. The hard answers start with questions about how we can do all that stuff better. For sure all the answers will be, to a huge degree, digital.
The vaccine programme had one key thing Test and Trace didn’t. And it wasn’t money
Terrific piece by Robert Colville in the Sunday Times. It’s paywalled but Charles Arthur has the key extract in his wonderful Overspill.
Something almost no one outside government appreciates is that the British state, like all its modern counterparts, is essentially a collection of databases. Throughout the pandemic, its policy successes have largely come where there are good databases, and its failures where there are not.
The furlough scheme worked because of PAYE. The expansion of universal credit relied on the existing benefits system. The “shielding list” of vulnerable patients was compiled by blending six data sets from NHS Digital.
Good data is also the secret sauce of the vaccination rollout. The jabbers could move seamlessly down the age and risk cohorts, because GPs had the appropriate patient lists. There have still been huge challenges in distributing the vaccines and tracking down the unregistered, but the data gave us an enormous head start.
The central problem with Test and Trace, by contrast, was that it didn’t have a database. When the pandemic hit, Apple and Google developed a joint framework for contact-tracing apps, which would ping you if someone you met later tested positive. But they wouldn’t let your phone share those details with the government — hence Matt Hancock’s abortive attempt to develop a homegrown alternative.
The trackers and tracers therefore had to map out the nation’s social network from a standing start, getting individual contact lists from every person who had tested positive to find out who else needed testing and quarantine. Public Health England even managed to lose 16,000 cases because it built its database with a stone-age version of Microsoft Excel and the file grew too large.
So the key discriminator between success and failure was not (as I had assumed) public sector vs (bloated and incompetent) private contractors, but who had a database and who didn’t.
The Fantasy Island that is ‘Global Britain’
This from Jonty’s Blog hits the nail on the head:
“The Brexit fantasy combines brilliantly in the final story from the weekend. The government is being urged to permanently station a frigate in Australia as a “warning” to China. That is a warning to a country with a navy which is expanding so rapidly that it is building the equivalent of the whole French navy every three years. What possible difference would one frigate make, except to hack off Beijing? The country we apparently want a trade deal with. “
Reminiscent of the famous story about Stalin who, upon being told that the Pope was opposed to something he was doing, inquired “And how many Divisions has the Pope?”