In the end, Elsevier retracted Kostoff’s anti-vax article, along with a pro-ivermectin study in the same issue that was similarly statistically-challenged. (It was that ivermectin study that led me to discover the issue in the first place, via scite.ai)
But not before his article dominated media and social media references to the Journal for months; and the author parlayed his peer-reviewed work into a DailyClout essay that was even more extreme, and did a tour on the social media anti-vax circuit. Thousands of people spent time debunking this nonsense, including a dozen on PubPeer alone. Millions of people saw references to it on social media.
The editor-in-chief who regularly published his own articles (or added himself as author to articles in his journal) stepped down as EIC, but continues to edit other toxicology journals and publish research at a healthy clip of three articles a month. Global understanding of COVID-19 is advancing steadily, with no further confusion or misdirection whatever. Everything isfine 🐶🔥
Comments Off on Kostoff, reprised: peer review secured again, everything is fine.
The progression from normal cells, to cells that can form a detectable mass, to outright cancer, is called malignant progression.
2. 90% margins
Industries with 90% or higher profit margins (often: marginal profit margins, where there was some up-front cost doubling as barrier to entry and hand-waving excuse for continuous rent increases) are all deeply inefficient and non-competitive. That should be what you (or any economist) would suspect, yet people continue to say things like “I’m not actually against the 95 percent profit margins or even caps if the market for broadband were competitive. Unfortunately…”
The rise of these industries eat collective surplus and productivity, and funnel the fruits of new technology into the hands of organizations that think this sort of resource allocation is healthy. This gives them ample resources to expand their work, into new markets and topics, and to train new industries to adopt their techniques. 90% margins become 99%, until all available shared resources are captured by this network. In other words: cancer.
Here is the head of ISOC, convincing himself and others that a well-meaning private equity firm will not unreasonably raise rates for use of their namespace monopoly. “Given registries must announce price increases for renewal 6 months in advance, and domains can be registered at current prices for up to 10 years, any operator seeking to increase prices dramatically would certainly lose customers without producing any increased revenue.”
This is not so. Renewal rates are quite price-inelastic (it costs > 100x the annual registration cost to change one’s domain on all sites and materials, and breaks existing links). Incentivizing people to hurry up and register for 10 years at once would produce a surge of revenue, not a decline. New domains can have prices raised with no warning, which would simply raise new domain rates for TLDs across the industry: likely bringing in more revenue as well as support from other registries (.org / .net /.com are among the few TLDs that can unilaterally affect industry rates)
a quick aside while travelling, remembering also the Horowitz lab printer that printed a strange page while powered off, after I slept next to it
I like to get the lightest machines I can, and slide them into thin bags. Currently using: a $170 Samsung chromebook, which suits me just fine. That’s both because I heartily support the concept – a 2-lb machine that can survive droptests [if not the XO’s dropkick-test!] – and because I tend to destroy drives.
Recently, the faults tend to be electrical. And now that I have switched to SSD, these are harder to recover from. It seems to be controller death — and I still have a recent frotzed drive so I can find out, if I find someone who can debug Sandforce controllers. At any rate, I’ve lost two SSD’s on my Macbook Air in the past 2 months. All of which makes me even happier about one aspect of Chromebooks that some deplore: the lack of permanent on-disk apps or data.
Meanwhile, I wonder if it’s becoming easier to find customizable tempest cases for laptops – shielding works both ways!
Comments Off on Electricity and the body electric [robot counterpoint]
A recent discussion thread on the Wikimedia mailing list led to a somewhat emotional exchange about the pros and cons of publishing budgets and annual plans.
I have worked in organizations that avoided writing annual plans, and did so only when required by a partner. Why? Because it was easier to “get work done” without wasting time producing a summary of our work to show to outsiders. Time invested in summarizing the work of the past year was unnecessary overhead; and time invested in projecting the work of the coming year was unnecessarily binding our hands — what if we wanted to make a sudden change? These orgs also tend to make it very difficult to get a copy of their Form 990s.
I have also worked with organizations that publish everything – their current burn rate, income, future goals, what money will be spent on until it all ran out (and exactly when it will run out!). Most stick to a yearly report and analysis, though some are more flexible.
Wikimedia is firmly in the latter group. We publish our 990s as soon as they are approved; we make our fundraising totals visible in real-time; we produce thoughtful annual plans, and complement them with wonderfully thorough monthly summaries of all of our activities, following monthly metrics meetings which anyone in the world can dial into.
And we develop both our strategic plans and our individual project plans in public — anyone can comment on and make suggestions to each individual project we have ever run. For the most part, this is a warm collaboration: people leave comments, feedback, and suggestions; point to bugs and feature requests filed; and generally track the progress of their favorite projects. Sometimes people share concerns when they don’t like how a project is affecting their editing or reading. And sometimes they are critical of projects they don’t think should be there in the first place.
Across our movement, we have steadily moved towards more and more transparency in our operations and planning. Starting this past year, most Wikimedia chapters publish their annual plans before they are approved. The largest chapters have those plans vetted by an international community body, which oversees distribution of a shared pool of funds. During this process their plans, like most things involving our Projects, are publicly displayed on the Meta-wiki – along with discussion and review of them.
However the Foundation itself remains reluctant to share its plans and particularly drafts of its budget in this fashion. There is perhaps a fear that a public community discussion will lead to (unspecified) bad results, or will be distracting for WMF staff, who will feel compelled to respond to every comment. This does not seem directly tied to any past barrage of comments on plans or budgets – each year this is the source of a fairly small number of comments overall, and most of them are not negative.
The one area in which there is an explicit “call for public comments” followed by a thorough public discussion is in the area of software feature rollouts.
Jorn Barger, Mortimer Adler, Patrick Gunkel, Vince, Misha Herscu, Popova: what do all of these people have in common? Exploring the densely textured space of possible ideas, the mereology of existence, learning to see implied, hidden, missing, combinatorial spaces.
Ideonomy is a dream of structuring that work; embraced by Gunkel in lists and charts and drawings, never fully realized. Modern tools and languages bring us closer to being able to explore such spaces computationally and comprehensively, to come up with questions about idea space and experiments that can resolve their answer: not in the naively space-filling method of the Library of Babel, but in the equivalent populated entirely by meaningful and informative works.
Comments Off on A new endeavour: making ideonomy into a science
Realization: In academia, science has grad students to do anything known to be important but not yet automated / solved. A common goal once these are identified is automating / solving. Wikipedians have done the first part, but don’t clearly have an analogy to the second as a goal.
Software design (for wikis) has focused on making reading better or more accessible, or helping making manual work less arduous; but not primarily on identifying automatic classes of work and solving them / knocking them off. That’s been limited to bot developers and ad-hoc tools built on the toolserver. (If any tools in the wikiverse do this, they are often by Magnus, and regularly get rate limited by the limitations of default toolserver allocations when they get popular.)
Thought: I suspect that is really the primary work moving the project forward. We need to recognize that and start framing and articulating goals, tools, and infrastructure accordingly.
<update from the AI era: yes, with bells on!>
Comments Off on Solving problem-sets: Improving our focus on new knowledge at scale
Fermi was famous for asking difficult general questions of interest that could be approximated through a series of good guesses, such as: “How many molecules of rubber are rubbed off of car tires in the country each year?” “How many stars are visible at night to the naked eye?” Order-of-magnitude approximations are generally appropriate answers.
This is equally interesting outside of physics, though it relies on knowing different trivia and rules of thumb. “How many schools are there in the country?” “How many people climb these mountains each year?”
I was up early this morning for a walk, and at work at 6. It was gray out, yet across the river bright lights were on in a cube of windows in the middle of the John Hancock building. It has large glass windows, so its lights make an impression; there were lights in dozens of rooms in one corner of the building, looking out of the river, covering many floors. One light every 4 windows, perhaps a cube 8 offices on a side extending up 8 floors. 100 offices if just by the building’s edge, 500 if the pattern continued inside. Part of a large company, surely. The lights were bright enough to be construction, which would explain them being on round the clock [as much as anything would; I’ve never understood that practice].
So who was the company? There were about 60 floors in all, and the lights were between 35 and 45. Searching for new companies moving into the Tower turned up Bain Consulting… (more…)
Comments Off on Estimation problems and answers
Bad Behavior has blocked 150 access attempts in the last 7 days.