RE: @paulfoerster@swiss.social
Voting Republican is evil.
Not voting is being complicit with evil.
Voting 3rd party is being complicit with evil AND bad at math.
Remote
Weekend Editor
@weekend_editor@mathstodon.xyz
mastodon 4.7.2Retired physicist, after a career in machine learning & stats mostly for cancer drug discovery. Now blogging about stats in the news.
Avatar: convergence basins in the complex plane of Newton's algorithm searching for the cube roots of unity. (After a NYT column by @stevenstrogatz@mathstodon.xyz, long ago.)
Header: Quote from GK Chesterton, London Daily News, 1905-Aug-16 on epistemic humility and the ability to say "I am wrong" as the foundation of idealism.
315 Followers
113 Following
50 Posts
Joined November 21, 2022
BlueSky:
@weekend-editor.bsky.social
Email:
SomeWeekendReadingEditor@gmail.com
Member:
Locally Flat Earth Society
Replying to
@paul_ipv6
I also do not want to see your discord, I will not look at your facebook page, I will not install whatsapp, and I will never, ever, under any circumstances, talk to your chatbot.
I want to *read* a manual. I want you to *commit* to what's in that manual.
I see the innumeracy of the rulers has not gotten any better in the last couple of years.
Replying to
@jackdaniel@functional.cafe
Yes, this is fast.
But most people won't see it, because they are in the grip of the myth that "Lisp is slow".
You get the most mileage by comparing against things people think are fast, at least for certain tasks, like C or Fortran.
For a venerable example of the art of making that comparison, Steele had a number of biting comparisons between MacLISP and Fortran back in the day. Here are a couple that had quite an impact on changing my views, back in ancient times:
GL Steele, "Fast Arithmetic in MacLISP", MIT AI Memo 421, Sept 1977.
https://dspace.mit.edu/entities/publication/81ec8fec-0598-49e9-9973-0abaf85fa6d7
GL Steele, "Debunking the expensive procedure call myth", MIT AI Memo 443, Oct 1977.
https://dspace.mit.edu/entities/publication/77b3eea1-50be-4e3f-b578-264484ee0f1a
Replying to
@jackdaniel@functional.cafe
Gotta respect the classics!
That book is interesting in that it not only explores how to create whole new object systems with the metaobject protocol, but also explains how to *implement* things like CLOS nicely.
(Also, I'm fairly sure I was Gregor's physics TA back in the late 1970s or so. :-)
(Also also, respect for Borges in the background in that shot. Magical realism sometimes creeps into our software thinking. :-)
Replying to
@johncarlosbaez@mathstodon.xyz @highergeometer@mathstodon.xyz @eighteyes@hachyderm.io
... or, if you're writing in your own notebook about your own research: "How much detail is my later self going to want to know? It's all fresh in my mind now, but in a couple of years it will not be."
I tend to over-document stuff because that fool who I'm going to become in a couple of years will have forgotten everything important, and will remember only trivialities. I have very regrettable empirical data to confirm this. :-(
Replying to
@johncarlosbaez
When I was starting out doing statistics in oncology bioinfromatics, I had some gene expression data for cancer cells from which I was supposed to find gene patterns that indicated the onset of cell death (apoptosis).
Some of the genes that came up were the SNAP and SNARE families. Back in those days, they were annotated as neuro genes, making membrane vesicles for neurotransmitters and such. The first response of my biologist client was a rather rude one: "Get that neuro crap out of my data!"
Well.
It turns out they're also involved in "membrane blebbing", in which cell surface membranes make vesicles to eject cytoplasm, as a part of both cell division AND cell death by apoptosis. So "that neuro crap" was a set of genes involved in reshaping membranes, the result of which depends on if the cell is squirting neurotransmitters, dividing, or dying. Dying was the relevant context in that dataset.
Context matters!
I was polite, but it was with great satisfaction that I could eventually point this out to the rude biologist. :-)
https://www.sciencedirect.com/topics/biochemistry-genetics-and-molecular-biology/blebbing
Replying to
@FluentInFinance@mastodon.social
The ones who made the most money in the American gold rush of the 1850s were the ones selling wheelbarrows, picks and shovels.
But... gold rushes come to an end when people finally do the math.
Meta may make some money selling wheelbarrows... but the AI bubble will pop nonetheless.
Replying to
@FluentInFinance@mastodon.social
Not too long ago, we'd have said cheaper electricity enabled more profitable crypto mining.
Crypto was a grift; pretty sure AI is not an improvement.
Today's blog post:
A seasonal song for this moment in American politics.
99 days to the primary,
99 days remain.
Mark 1 down, canvass your town,
98 days to the primary!
My cats also have opinions on this subject. They are correct.
Replying to
@grimalkina
And now, alas, Trump has fired the entire governing board of the NSF:
https://www.science.org/content/article/trump-fires-nsf-s-oversight-board
Replying to
@ColinTheMathmo @vpunt
Oh, I have *so* many battle scars from this!
Imagine my frustration when Excel took gene expression data and converted gene names like DEC10 to dates, then munged the dates DIFFERENTLY, depending on whether the data was last saved by American or European colleagues! (True story, actually seen in the wild.)
Some genes have actually been renamed because this error in Excel is so damn persistent.
In 2021, about 30% of papers contain mangled gene names in supplementary data:
https://www.nature.com/articles/d41586-021-02211-4
Same in 2016, multiple sources:
https://link.springer.com/article/10.1186/s13059-016-1044-7
https://www.science.org/content/article/one-five-genetics-papers-contains-errors-thanks-microsoft-excel
This has been going on for decades, since AT LEAST 2004, which is when it bit me:
https://link.springer.com/article/10.1186/1471-2105-5-80
You can't get Microsoft to fix it, since it's not a "business problem".
You can't get biologists to stop using Excel, becaues it's a "bioinformatics problem".
Ultimately, you CAN get the international standards body that assigns HUGO symbols to human genes to CHANGE 27 GENE NAMES to bulletproof them against Excel mishegoss:
https://www.theverge.com/2020/8/6/21355674/human-genes-rename-microsoft-excel-misreading-dates
At that point, Microsoft was sufficiently embarrassed to put in a switch in Excel to turn off data conversion (apparently all of it, not just genes & dates). I was retired at that point, but I will bet (small amounts of) actual money that ALMOST NO BIOLOGIST KNOWS HOW TO TURN THAT SWITCH OFF, so the problem would have persisted had the genes not been renamed.
Replying to
@skeletor@mas.to
The word "repent" comes from the Latin "repaenetere", which literally means "to rethink", decide what went wrong, and do better next time.
The Greek equivalent "metanoia", literally something like "thinking about thinking", usually translated as "changing one's mind", also implies trying again.
The Hebrew root behind all of those, "teshuvah", means to "return" to your roots as an intrinsically good person, and try again.
Or, as the playwright Samuel Beckett put it: "Try again. Fail better next time." It's a process of continuous improvement, not sudden success.
Replying to
@geonz@mathstodon.xyz @workchronicles@mastodon.social
Yeah, I spent an entire career saying "I must be so stupid" in the last panel. No success can compensate, apparently.
Replying to
@johncarlosbaez@mathstodon.xyz
Needs Roger Bacon's famous magical brass head.
Any prankster-sculptors available? Pretty sure Banksy would do it.
https://en.wikipedia.org/wiki/Brazen_head
Replying to
@paul_ipv6
The experience FOR THEM is better in the app, because they can data harvest like mad.
In a browser, I can control their use of javascript, block ads, and have little snitch strangle their outbound network traffic.
Though, if they make me do that even in a web browser, chances are I want nothing to do with them in the future.
I wish we could all proclaim that, as loud as possible, so the vendors would maybe get it.
Replying to
@anarodrigues@front-end.social
Also, it's pretty sure that in the past layoffs targeted women (a) statistically significantly more often and (b) with bigger strength of effect (i.e., laid off a larger % of women).
https://www.someweekendreading.blog/2022-tech-layoffs-bias/
Replying to
@AmenZwa
You need to meet the screenwriter John Rogers @johnrogers.bsky.social
who is also a fountain pen nerd. I think he's more active on BlueSky, though.
Replying to
@rogersm@mastodon.social @jackdaniel@functional.cafe
The paperback seems to be in stock at Amazon?
https://www.amazon.com/Art-Metaobject-Protocol-Gregor-Kiczales/dp/0262610744
Replying to
@johncarlosbaez@mathstodon.xyz
From a couple years ago, on the unaffordability for the rest of us of keeping the rich around. GINI coefficients, class mobility vs wealth inequality, and so on.
https://www.someweekendreading.blog/unaffordable-rich/
Replying to
@skeletor@mas.to
While Mother Theresa was far from perfect, she's famous for (probably) saying:
“It’s not necessary to be a saint to do good. You need willing hands, not clean ones.” — Mother Theresa
(From my quotes page:
https://www.someweekendreading.blog/quotes/#:~:text=%E2%80%9CIt%E2%80%99s%20not%20necessary%20to%20be%20a%20saint%20to%20do%20good%2E%20You%20need%20willing%20hands%2C%20not%20clean%20ones%2E%E2%80%9D%20%E2%80%94%20Mother%20Theresa
)
Replying to
@camertron@ruby.social @joeldrapper@ruby.social
"History will judge us either stupid or complicit, and we know we are not stupid." -- Jack Hardy, "Ask Questions"
https://www.youtube.com/watch?v=zX7sARa5Bcc
Jack Hardy - Ask Questions
Replying to
@danwwilson@rstats.me
It gets worse.
Excel infamously will convert certain gene names (under the HUGO standard) to dates, like DEC10. For decades, you could not turn it off; I hear rumors that now there is a switch somewhere to turn it off, but nobody remembers where it is.
So: I once had a dataset that biologists kept loading into Excel, despite my warnings. It came out DIFFERENTLY in the US and Germany. After much frustration, it turned out to be a gene name being converted to a different date format in each country!
You will debug MANY, MANY things before you check that.
Replying to
@nigelpstanley@mastodonapp.uk
This is... absolutely gorgeous.
4 ways with stained glass + 4 ways that lead... elsewhere.
I want to apply to the caretakers for permission to lie on the floor for an hour, and weep, with... well, probably joy, and sadness.
Replying to
@AmenZwa
Knew all of these but MCAS. Did you mean the Massachusetts Comprehensive Assessment System (high school graduation test), or Mast Cell Activation Syndrome, or... ?
Replying to
@AmenZwa
Oooh! The 737Max software thing.
I totally didn't see that one. MCAS hereabouts almost always refers to the high school test, unless you're medical in which case it's the other.
I guess I don't hang around with aero/astro folk enough.
Replying to
@siderea
Instead of a RAID, I'm getting (next week?) a Synology NAS with 2 bays, 4Tb drives in each bay. The backups will alternate drives for replication.
At least, that's what my Mac systems guy recommended, and he's been pretty good in the past.
Replying to
@kentpitman@climatejustice.social @screwlisp@gamerplus.org @larsbrinkhoff@mastodon.sdf.org @eswenson@c.im
I remember being at a talk you gave on your semi-automated Symbolics MACSYMA conversion.
I asked you what it did about comments, since it processed S-expressions from which comments were removed.
You shrugged and said, "Fortunately..." and then everybody laughed because we all knew there were so few comments that they could be handled one-by-one. :-)
Replying to
@abstractcode
The traditional version of this argument involves tumbrels and guillotines.
Replying to
@Garwboy
NotebookLM made a 'podcast' about my blog. Two happy-talk 'hosts' ask each other questions, get most details wrong, and fail to answer the questions.
A friend described it as "the verbal equivalent of Muzak", which apparently means AI is unavoidable but best ignored.
I wrote about it here:
https://www.someweekendreading.blog/someweekendreading-ai-podcast/
Replying to
@ftp_alun @ColinTheMathmo @vpunt
It's because Excel is trying to be "helpful" by intervening to CHANGE YOUR DATA WITHOUT YOUR CONSENT OR KNOWLEDGE, and there was for approximately 20 years NO WAY TO TURN IT OFF. The result is nonsensical gene names in approximately 30% of all papers about genes, because you CANNOT pry Excel out of people's hands even after they see the damage it does.
I think a data analysis tool should respect data WAY more than that! Excel fits the term "data analysis tool" loosely at best, but it's the tool to hand and what people will use if they don't know better.
Ask @louisepryor sometime about her days as an actuary, patiently explaining to bankers why they can't write massively complex spreadsheets and expect anything other than disaster. Financial analysis is exactly the intended application of spreadsheets, after all. (Oh, and source control too. Much of the world's financial system apparently used to run without source control. *Exactly* which version was running? Nobody was quite sure...)
Replying to
@nosrednayduj
Last year I ordered something which initially got as close as Boston. But instead of being shipped from there to my home in deepest, darkest suburbia...
It got sent to Puerto Rico. Apparently the folks there were as confused as me, since they bounced it around among various post offices in PR for a week.
Finally somebody clued in on the ZIP code, and sent it along its proper way.
I have literally no idea how that happened.
RE: https://mastodon.social/@arstechnica/117049561381886953
"Claiming authority over limit set by Congress": basically, rule of law matters so little to Republicans that Trump can do whatever and dare the rest of us to stop him if we can.
Replying to
@christianp
My permission to operate came in a couple weeks after the install, but the state approval so I could get reimbursed for solar energy credits took about 6 months.
Of course, all the energy I generated in that time does not count, since it was pre-approval.
There are *enormous* numbers of blockades in the way of getting a working solar installation!
Jack Hardy, "May Day".
Replying to
@ColinTheMathmo
Earliest cite I could find was Tal Waterhouse on Twitter, 2018-May-03.
(Caution: link to X/Twitter, probably best avoided until you really need it.)
https://x.com/iiTalW/status/991905585174986753
Replying to
@ColinTheMathmo @tao @dimpase
There have been some really interesting psychological studies on people who use LLMs heavily (the Big Five personality test and so on).
Whether they self-select or whether the LLM *makes* them that way is still open.
Here's a walk-through of one of those studies:
https://www.someweekendreading.blog/who-uses-ai/
Replying to
@christianp
The OEIS says, "if two such numbers exist, the smaller is chosen."
It looks like they aren't following that rule, and hence the sequence is occasionally double-valued.