Saturday, February 02, 2019

Against superhuman AI

I am a strong-AI pessimist. I think by 2100 we’ll be in range of sentient AIs that vastly exceed human cognitive abilities (“skynet”). Superhuman-AI has long been my favorite answer to the Fermi Paradox (see also); an inevitable product of all technological civilizations that ends interest in touring the galaxy.

I periodically read essays claiming superhuman-AI is silly, but the justifications are typically nonsensical or theological (soul-equivalents needed).

So I tried to come up with some valid reasons to be reassured. Here’s my list:

  1. We’ve hit the physical limits of our processing architecture. The “Moore-era” is over — no more doubling every 12-18 months. Now we slowly add cores and tweak hardware. The new MacBook Air isn’t much faster than my 2015 Air. So the raw power driver isn’t there.
  2. Our processing architecture is energy inefficient. Human brains vastly exceed our computing capabilities and they run on a meager supply of glucose and oxygen. Our energy-output curve is wrong.
  3. Autonomous vehicles are stuck. They aren’t even as good as the average human driver, and the average human driver is obviously incompetent. They can’t handle bicycles, pedestrians, weather, or map variations. They could be 20 years away, they could be 100 years away. They aren’t 5 years away. Our algorithms are limited.
  4. Quantum computers aren’t that exciting. They are wonderful physics platforms, but quantum supremacy may be quite narrow.
  5. Remember when organic neural networks were going to be fused into silicon platforms? Obviously that went nowhere since we no longer hear about it. (I checked, it appears Thomas DeMarse is still with us. Apparently.)

My list doesn’t make superhuman-AI impossible of course, it just means we might be a bit further away, closer to 300 years than 80 years. Long enough that my children might escape.

Sunday, December 30, 2018

Why the crisis of 2016 will continue for decades to come

I haven’t written recently about why Crisis 2016, sometimes called Crisis-T, happened. For that matter, why Brexit. My last takes were in 2016 …

  • In defense of Donald Trump - July 2016. In which I identified the cause of the crisis, but assumed we’d dodge the bullet and HRC would tend to the crisis of the white working class.
  • Trumpism: a transition function to the world of mass disability - Aug 2016. “How does a culture transition from memes of independence and southern Christian-capitalist marketarianism to a world where government deeply biases the economy towards low-education employment?"
  • After Trump: reflections on mass disability in a sleepless night - Nov 11, 2016. "Extreme cultural transformation. Demographics. China. The AI era and mass disability. I haven’t even mentioned that pre-AI technologies wiped out traditional media and enabled the growth of Facebook-fueled mass deception alt-media … We should not be surprised that the wheels have come off the train.”
  • Crisis-T: What’s special about rural? - Nov 16, 2016: "The globalization and automation that disabled 40% of working age Americans isn’t unique to rural areas, but those areas have been ailing for a long time. They’ve been impacted by automation ever since the railroad killed the Erie canal, and the harvester eliminated most farm workers. Once we thought the Internet would provide a lifeline to rural communities, but instead it made Dakka as close as Escanaba.”

How has my thinking changed two years later? Now I’d add a couple of tweaks, especially the way quirks of America’s constitution amplified the crisis. Today’s breakdown:

  • 65% the collapse of the white non-college “working class” — as best measured by fentanyl deaths and non-college household income over the past 40 years. Driven by globalization and IT both separately and synergistically including remonopolization (megacorp). This is going to get worse.
  • 15% the way peculiarities of the American constitution empower rural states and rural regions that are most impacted by the collapse of the white working class due to demographics and out-migration of the educated. This is why the crisis is worse here than in Canada. This will continue.
  • 15% the long fall of patriarchy. This will continue for a time, but eventually it hits the ground. Another 20 years for the US?
  • 5% Rupert Murdoch. Seriously. In the US Fox and the WSJ, but also his media in Australia and the UK. When historians make their list of villains of the 21st century he’ll be on there. He’s broken and dying now, but he’s still scary enough that his name is rarely mentioned by anyone of consequence.
  • 1% Facebook, social media, Putin and the like. This will get better.

That 1% for Facebook et all is pretty small — but the election of 2016 was on the knife’s edge. That 1% was historically important.

Rupert Murdoch will finally die, though his malignant empire will grind on for a time. Patriarchy can’t fall forever, eventually that process is done. We now understand the risks of Facebook and its like and those will be managed. So there’s hope.

But the crisis of the white non-college will continue and our constitution will continue to amplify that bloc’s political power in rural areas. Even if civilization wins in 2020 the crisis of 2016 will continue. It will test human societies for decades to come.

Thursday, December 27, 2018

Life goal #2 - The CrossFit Kipping Bar Muscle-Up

In 2018 I met one of my two CrossFit Life Goals (tm) - 10 consecutive dubs (hit 42 in a wild fluke the other day). Only took me five times as long as anyone else I know.

The other goal is the Kipping Bar Muscle-Up. So that’s on the list for 2019. I’m also planning an IMBA “Epic” mountain bike trek — the Maah Daah Hey, but that’s mostly about showing up and moving my feet. The Bar Muscle-Up may be impossible, so it’s more interesting. (Watching Paoli video I should be able to do it now [1], but that’s now what I’m feeling!)

I’m putting my training notes and references on this page.

Examples

Movement notes

Some of the best tips came from a post in the CrossFit Physicians Facebook group.

  • Hands a bit wider than shoulder, but narrower than bench
  • From beneath the bar jump up and back to a big arch then fall forward into extension position — gets good start there
  • Drive hips to bar from top of “hollow” position, while “push down on bar with straight arms”. There should be a bend at hips so can “pop” the hips and pull, row bar into chest/navel (pop/pull is the trick - need timing). The pull should be very hard and fast
  • Grip. Begin with the end in mind. You want to think about the position the hands and wrists need to be in when you're on top of the bar and doing the dip portion of the BMU. A lot of people struggle because they grab the bar from directly underneath and don't really wrap their hands over the bar. Instead of doing that, wrap your hands as far over the bar as possible. It's basically a 'false grip' maneuver that you've probably seen on the rings, just on the bar. The idea is to have to turn your hands/wrists over as little as possible. The-two word actionable cue for this is "meaty grip".
  • In the kip swing, think about getting your body as loooooong as possible in as you move in front of the bar into the extreme of the arch position. Maximize the full range of motion of your shoulders, keep knees as straight as possible, and point your toes. Your body is a sling shot. Two word cue: "get long". 
  • For the pull up & over the bar. Pick a spot on the floor in front of you. Your goal is to stare at that spot all through the first part of the kip and the initiation of the transition into the hollow position & first part of the pull. At some point, as you pull your hips up to the bar, you're going to lose that spot from your vision (it's unavoidable - the head tilts backwards), but your goal is to see that spot again as soon as possible. The instant you don't see the spot anymore, your sole focus is to find it again. This will help you with the speed necessary to execute the transition. Actionable cue: "find the spot". 

Training programs

Some of these are for the (ring) muscle-up, the bar muscle-up is considered to be harder

Training exercises

  • Strict pull-up
  • Kipping chest-to-bar with elbows behind the back
  • Lat Pull-downs
  • Use gym machine with pulley’s ropes to emulate the curious straight arm downward push-pull (see still below).
  • Back extension and arching
  • Shoulder range of motion, esp. internal rotation
  • Band-assist Muscle-Up with gradually diminishing bands
  • Box jump muscle-up with gradually smaller box
  • The glide kip drill - stand on 1-2 bench, bar a bit above eyes, arms and back in line with hip flex to big hollow, hop up, glide out with feet just above bench and extend to small hollow, then reverse on return.
  • Paradiso progression
    • 3 sets of 5 high back kip swings
    • Hip to bar pull up (20 repeat 1 rep) - kip swing with hip snap
    • Transition (atop box, etc) - 20 to 30 reps of jumping box muc
  • Hip-to-bar progression with a slight arm-pull, hip drive from the hollow (I can’t get my hips to the bar yet), note in this still from Paoli video his elbows are bent, but he’s mostly pushing the bar down towards his hips and lower abdomen. Feet are below hips. Trapezius muscle here. I have to figure out how to build something like this.
    Screen Shot 2018 12 27 at 4 27 15 PM
    and note he’s actually hitting bar around navel at this point (not hips), feet are still in front as he transitions.
    Screen Shot 2018 12 27 at 4 32 13 PM

[1] Well, not now exactly. My left biceps is strained, so I have to rehab that first.

Update 2/3/2020

I posted the first version of this on Dec 27, 2018 and I succeeded on Feb 3, 2020.

It took me over 13 months. At one point, probably June 2018, I bet my 17yo daughter I’d get it done by my 60th birthday on Aug 1 2018. If I’d succeeded she owed my $10. If I failed, I would pay her college. 

Emily and I are paying her college. 

In August 2018 I was feeling kind of bleh and this looked impossible. To my surprise I started to recover in October and have been getting stronger over the past 5-6 months. Maybe the protein and creatine shakes helped. I also started doing more CrossFit for weird (but good) family reasons — 4-6 a week instead of 3-4 a week.

I didn’t end up following any of the programs above. I just worked on pull-ups whenever I could and I practiced with bands. I could do BMUs with a “green and blue” (green is BIG), then the sequence went like this (typically 3 at a time, the first is usually hardest):

- 1/9/2020: single black band 3 consecutive
- 1/20/20: red and blue then thin red, orange, blue
- 1/22: blue+thin red+orange
- 1/29: blue and 1 orange
- 2/3/20: During a “max MU in 4 minutes” WOD I started with blue, red, and orange bands and in about 4 sets of 3 I dropped bands until I did 3 with two thin orange bands and then one with no bands. When I actually succeeded it didn’t seem that hard. Timing is key.

The most useful advice I was given was "wrap your hands as far over the bar as possible. It's basically a 'false grip' maneuver that you've probably seen on the rings, just on the bar.” I didn’t really understand it at first, but it means having wrists above the bar when you start. I had note understood how hard I had to grip the bar and lever myself up.

I hope I do more, but in terms of my goal I only had to do one.

Saturday, December 15, 2018

Old doctor pet peeve - disease descriptions were better in Osler's day.

I didn’t know Sir William personally, but I think he’d be appalled by the descriptions of patient-disorders in our medical textbooks.

We typically present a collection of findings, sometimes organized by ‘history’ or ‘physical’ or ‘lab’, but the sequencing and relationships are all lost.

We should have textbooks that describe a disease or disorder with 3-8 case histories that span a reasonable spectrum of presentations. So not x% have red eyes and y% have chest ache, but a case called “common” that might be “8 yo both with cough, then red eyes for day and a swollen lymph node, then a day later some red palms …” and another case called “often”, and 2-3 called “unusual1”, “unusual2”, and so on.

Ok, it’s not just my Oldness. I’ve been grumpy about this for roughly 25 years, even when I was only old in spirit. It annoys me a great deal.

Now I’ve said it.

Saturday, December 01, 2018

Quicken for Mac -- why vendors are going to screw-up subscription pricing for software services

We’ve been using Quicken for Mac for the past year. I’m satisfied with the software, but I no longer trust their pricing and renewal.

We paid $60 for Quicken for Mac 2015 on 7/31/2015.  On 1/8/2017 we paid $48.41, presumably for 1 year of subscription service. On 12/31/2017 we paid $32.35; we probably switched from a “deluxe” plan to a basic plan.

Today I received an email requesting renewal:

Your Quicken membership will expire on 12/31/2018. In order to continue enjoying all of the benefits of Quicken, including connected services such as bank downloads, stock price updates, account sync, and free phone support, please click here to renew your plan.

The link goes, however, to Quicken for Windows where we are shown as “deluxe” plan for $50.

Ok, so that’s presumably a mistake — albeit a bad mistake. When I go to https://www.quicken.com/mac/compare I can see the Mac plans - Starter at $35 (so probably a 15% price hike from last year) and “most popular” Deluxe for $50. I can’t compare to last year but it looks like most of the features added in the past year or so require the “deluxe” option now.

Which leads me to reconsider my previously relatively positive attitude towards software subscriptions.

I’ve been generally in favor of subscription pricing for software. I think Microsoft has done a great job with Office 365. It does, however, come with temptations for vendors. Subscription pricing makes it too easy to hide price increases and game features. It promotes “information asymmetry”.

I think Quicken has fallen for that trap.

I don’t trust them now.

We are evaluating options.

 

 

Saturday, November 03, 2018

Amazon reviews now unreliable - negative reviews filtered (Anker example)

Amazon reviews have long been helpful to me, and were once a big part of Amazon’s value proposition.

That is no longer true. Amazon is filtering out negative reviews.

I learned this after attempting to review Anker bluetooth earbuds I bought for Emily’s birthday. The power switch was defective. That wasn’t a complete surprise, I have a similar pair and I often have to push 2 or 3 times. Anker should have spent another 10 cents on that part.

Amazon made the return easy, but when I tried to write a review I got this notice:

Screen Shot 2018 11 02 at 10 12 04 PM

“Sorry we are unable to accept reviews for this product …”

I then switched to Emily’s account. There I was at able to start a review, and even able to give in a two star overall rating. When I clicked 1 star for material quality however the “unable to accept reviews” notification appeared:

Screen Shot 2018 11 03 at 12 18 40 PM

This is, of course, worse than if Amazon removed all product reviews. They are promoting systemic bias in their closed world. The Fox model is catching on elsewhere, Apple is doing something similar with the Mac App store.

On certainty

A week ago, in a Facebook group for CrossFit physicians, I read a post blaming poor parenting (i.e., poor mothering) for childhood obesity.

There and elsewhere I read confident statements on nutrition. I read confident statements about market movements (usually retrospective) and white men.

I always imagine these confident people are young, but I believe that is not always true.

There are likely things. There is entropy and death (but also the inexplicably low entropy of the early universe, and it is possible for an Old person to be stronger than they were in middle age).

Most things though, they are … complicated. They are some of this and some of that.  If you cannot imagine a worse outcome than an obese child you have not opened your eyes enough, or lived long enough for the world to pin your eyelids back and burn reality into your retina.

Look for the Old of any age. Some have something that might be wisdom. Ask them about certainty.

Sunday, October 21, 2018

Nasty flaw in Minnesota mail ballot process

There’s a nasty flaw in Minnesota’s vote by mail process.

When you apply for a mail ballot you are asked to provider either a SSN last 4 or License number or State ID number.

When you complete the ballot you are asked to provider either a SSN last 4 or License number or State ID number.

The two numbers have to match or the ballot will be rejected.

Hope you remember if you used your MN License number or your SSN Last 4 on the ballot application.

I tried testing for the identifier I used by querying my absentee ballot status, but it found the same status regardless of which identifier I used.

Saturday, October 20, 2018

Attack of the Clones - New disease mechanism identified

First came CHIP - Clonal hematopoiesis of indeterminate potential (Jan 2018)

… a bizarre accumulation of mutated stem cells in bone marrow increases a person’s risk of dying within a decade, usually from a heart attack or stroke, by 40 or 50 percent. They named the condition with medical jargon: clonal hematopoiesis of indeterminate potential…

… Up to 20 percent of people in their 60s have it, and perhaps 50 percent of those in their 80s …

… large numbers of study participants had blood cells with mutations linked to leukemia — but they did not have the cancer. Instead, they had just one or two of the cluster of mutations…

… [mutations], especially those linked to leukemia, seem to give stem cells a new ability to accumulate in the marrow. The result is a sort of survival of the fittest, or fastest growing, stem cells in the marrow…

… researchers described a 115-year-old woman. Nearly her entire supply of white blood cells was generated by mutated stem cells in her bone marrow.

At the first she had developed just two mutated stem cells. But over time their progeny came to dominate her bone marrow. She lived about as long as a human can, nonetheless, and died of a tumor.

… Mutated blood cells began proliferating in the mice, and they developed rapidly growing plaques that were burning with inflammation.

“For decades people have worked on inflammation as a cause of atherosclerosis,” Dr. Ebert said. “But it was not clear what initiated the inflammation.”

Now there is a possible explanation — and, Dr. Ebert said, it raises the possibility that CHIP may be involved in other inflammatory diseases, like arthritis.

That was mindboggling. An entirely new mechanism of disease! It’s easy to speculate on relationships to unexplained disorders like osteoarthritis.

This week the clones are everywhere …

Researchers Explore a Cancer Paradox Oct 2018

… a large portion of the cells in healthy people carry far more mutations than expected, including some mutations thought to be the prime drivers of cancer…

… rogue cells spread out across the esophagus, forming colonies of mutant cells, known as clones. Although these clones aren’t cancer, they do exhibit one of cancer’s hallmarks: rapid growth.

These mutant clones colonize more than half of your esophagus by middle age” …

… By examining the mutations, the researchers were able to rule out external causes for them, like tobacco smoke or alcohol. Instead, the mutations seem to have arisen through ordinary aging. As the cells divided over and over again, their DNA sometimes was damaged. In other words, the rise of these mutations may just be an intrinsic part of getting older…

It’s been a long time since we’ve had an entirely new class of pathophysiology. We may be entering a new and exciting era of medical research with near term clinical implications. Nobel prizes have been awarded for less.

We still need a way to explore and resurface old blog posts

Six years ago I wrote about browsing the blog blacklist and the need to resurface content from old blog postings. Today, even in the supposed twilight of blogs [1], I was again reminded how much we need a tool for excavation of old posts. I can think of at least one way do it (standard metadata for blog history, random selection of past posts based on internal and external inbound links) but there are probably several.

Maybe something for a future blog renaissance to tackle. Or if Feedbin is looking for a new feature set …

- fn -

[1] On the one hand I accept that RSS and blogs are vanishing, on the other my Feedbin stream is a rich and engrossing as ever, covering hundreds of sources.

Why is the hamstring connected to the paraspinal muscles?

The hamstrings (biceps fermoris, semitendonosis, semimembranosus, some include adductor magnus) flex the knee and extend the hip. The erector spinae and latissimus dorsi flex, extend and rotate the spine [1].

These muscles don’t directly connect with one another. They are innervated by different nerve roots. As far as we know [2] they only connect in the brain.

So it’s curious to observe the connection between the minor back strains I get [3] and my hamstrings. I normally have a good hamstring stretch for my age, but even a minor erector spinae strain will immediately tighten the ipsilateral (same side) hamstring. Improving the back strain is likewise intimately related to stretching the hamstring.

I presume it’s some kind of injury reflex, but I can’t figure out why it’s adaptive.

[1] Kudos to my all-time favorite medical app, Visible Body’s Human Anatomy Atlas.app, for helping me visualize these areas. It takes a while to learn this powerful app, but it’s worth the time.
[2] Decades after we thought we understood anatomy we keep learning new things. So who knows :-)?
[3] There’s a personal history here. I have been mildly surprised how little this history interests other people — including people with disabling back pain. The back strains I get now are more annoying than painful; they are usually related to heavy weights and intense exercise.

Saturday, October 13, 2018

Bostromism rediscovered

While recently rereading Banks’ marvelous book The Algebraist I decide I ought to write a post about “The Truth”, which is basically Bostrom’s simulation hypothesis made theology.

An initial topic search lead me to a 2013 pinboard post, where I learned I’d started to write that post 5 years ago - in 2013. I didn’t finish then, because at that time a topic search uncovered a post from 5 years before that. In 2008, 10 years ago now.

My 2008 post includes an excerpt from The Algebraist — it’s worth a read. Banks begins with the consensus response to the simulation hypothesis — “a difference that made no difference wasn’t a difference to be much bothered about, and one might as well get on with (what appeared to be) life.”  I’m not sure the hypothesis is totally irrelevant though; it is one answer to the Fermi Paradox.

Now I’m looking forward to doing this again in 2023.

PS. deGrasse in 2016.

PPS. I’ve been lately thinking about the other odd aspects of Oldness. Such as realizing I’m supposed to be the sober and silent sage in meetings whose primary duty is to do introductions and conclude with expressions of appreciation. Most unfairly, for someone who never missed a nap during meetings and classes, I’m not allowed to fall asleep any more.

Thursday, September 20, 2018

CrossFit 59 and so on

I used to write a birthday blog post each year about doing CrossFit. I missed this year; probably because its kind of a regular thing now. I had several personal best lifts this year, but none in the past couple of months. Still not bad for almost 60. Stronger than I was at 55, much stronger than I was at 40. I finally did more than 10 consecutive dubs (14!) … and then I dropped back to doing a few at a time. I still haven’t done a bar muscle-up, but I haven’t given up. I think I’m more impacted by heat than I used to be. My cardio isn’t improving much but it’s probably no worse when the temperature is good.

There is one new thing. My 16yo daughter started doing CF with me this summer. She’s progressing quickly; she’ll be doing sets of 50 dubs within a year. She’s good at CrossFit and it makes her happy. She hangs with very healthy, very fun, women from ages 20s to (maybe?) 50s. Pretty great role models.

I have no currently active wear and tear injuries, despite my familial arthritis. Shoulders, knees, wrists, back, hands, ankles — all behaving for a change. I don’t know why. Bodies are weird. I can do CF four or five times a week rather than just three times. Maybe I’m just better at scaling.

PS. For my birthday this year I got a mountain bike tire and some mountain bike shorts.

Saturday, August 18, 2018

Random thoughts on replacing Twitter

  1. Twitter will become a blend of home shopping network, daytime TV, and tabloid news. That might be quite profitable.
  2. There is money in sane social communication, but there isn’t big money. In particular there isn’t publicly traded corporation money.
  3. With current software and hardware stacks a base social network doesn’t have to be very expensive. I suspect without video storage and without advertising it could be done for $20/user a year and perhaps less.
  4. A consortium of newspapers, foundations, and universities with some grant money is enough to develop and support a standards based solution. Remember USENET* was basically supported by universities. USENET was also an open standard.
  5. USENET didn’t have to support a billion users though. A sane social communication network will require either an ad model similar to 1980s newspapers or user fees. I like the idea of free read access for all, contribution requires subscription ($20/year), people can donate subscriptions (free pool).

And that’s all the time I have to think about this for now …

* Still around, by the way. Searchable too.

Friday, June 29, 2018

Dubs

Ten. Consecutive. Dubs.

It’s not too much to ask, is it?

Oh, and one muscle-up. Or bar muscle-up. I’m not picky.

Dubs aren’t that hard. Most CrossFit newcomers figure it out in a few months, maybe a year. I don’t know anyone who has failed as persistently as I have.

A few months ago I got 8. It was a bit of a fluke. I’ve gotten 5 or more several times. I looked like a meth-addled feeding frenzy but, like I said, not picky.

Then I started to get worse. Finally I couldn’t get any at all.

I think I’ve figured out what happened. My technique got better. Instead of swinging my forearms I learned to do singles with my wrists — like you’re supposed to. Turns out my wrists can’t move the rope fast enough. I don’t have anywhere near the coordination. This isn’t only being Old, I’ve had horrible coordination all my life. Old isn’t helpful though.

Once I thought of this I went back to the arm swinging feeding frenzy and I can do a few again. So I’ll practice that for a while. Preferably somewhere nobody can see me. Which means not my gym, where sympathetic looks of sorrow and helpless frustration are hard to bear.

Meanwhile, I can work on those lat pulldowns.

8/6/2018: I0. Also, in that set, some 7s, 8s and a 9. #lifegoal

8/12/2018: 14

12/19/2018: 42 consecutive.