Category: Uncategorized

Post #1830A: Leaf extras-for-stat-nerds

 

The statistically savvy among you might have noticed something odd about the graph at the end of the last post.

When I let unconstrained math (via Excel) determine the best straight-line fit to these data points, it appears to tell me that these cars lost about 4% of battery capacity for every 10K miles.

But …

But what that doesn’t do is guarantee that with zero miles, I predict that you’ll have 100% of original capacity.

And that matters, because it looks like a lot of these cars must have lost a lot of range early on, that is, at low mileage.  And that loss of battery range doesn’t get factored into the 4% per 10K range loss estimate.

So this is a rare instance of a straight-line fit for which you are justified in “setting the intercept” manually, rather than letting an unconstrained least-squares fit to the data do it.  By definition, the line has to pass through zero miles matching 100% of original range.

Look what happens when I do that:

On average, pinning the linear trend to pass through 0 miles = 100% of original range, range loss is more like 8% per 10K miles. 

The whole cluster of dots is quite “low” on that graph, so to speak.  Those cars on average followed a path of losing a lot of range early on.  And having the loss taper off to a mere 4% per year.

My guess is that exponential decay is the line you’d like to fit, for something wasting away.  But I can’t seem to get Excel to give a trend passing through the required point, with the required shape.  So I freehanded what I think the Conventional Wisdom says about the time-path of battery loss.

No matter how you cut it, the actual observed battery loss over this range amounts to much more than 4%/10K miles.  The mid-point of the fitted line sits around a 40% loss over 5K miles.  That’s the 8%/year calculated when the regression line was pinned at 100% capacity at zero miles.

Why the high range losses?  Is this just a market for lemons, or are the dashboard estimates biased?

George Akerlof, economist, once wrote a piece whose title began “The Market for ‘Lemons’: ”  The paper is pretty deep, but the takeaway is pretty simple.  In this case, it boils down to: People sell their cars when those cars are lemons, not when they are peaches.

That’s one obvious explanation of why these losses appear far larger-than-expected.  (Where, lurking in the background is the idea that EVs lose, oh, like two percent a year, maybe, based on looking at a few graphs in the past.)  The point is, maybe most of these cars appear in the used car listings precisely because they had above-average battery capacity losses.  And that’s why their prior owners sold them.  And I’m seeing them.

The other possibility is that the mileage estimated from the dashboard readout is substantially biased downwards.  (I know it has a high variance, as it depends on recent driving style.  Leaf aficionados refer to it as the “guess gauge”.  That should just add noise, not bias, I think.)

There is one element of bias in what I did in my calculation, in that the range bars are each about 8 percent thick.  For example, when I saw four range bars, my calculation assumed the car was (4/12 =) 33% charged.  But if four bars is really just 3.5 bars, on average, then that charge level is actually just a shade over 28%, on average.  I would then triple that error in arriving at 100% of charge, and so end up understating total range by a factor of about 15%.

Adjusting for that  would require multiplying my range estimate by (1/.85 =) about 18%.  The upshot is that what looked like a 60-mile range based on the dashboard could plausibly be a 70-mile range in reality.

That said, I’m not sure this materially changes the situation.  No matter how I slice it, the range of the car I’m interested is far lower than the Recurrent.com estimate that I had been looking at.

Finally, I can’t fully discount that the losses observed in this low-end used car sample are typical of a random sample of Leafs.  But if it were, I doubt there’d be many Leaf fans out there.

On net, I think the explanation is that I’m looking at a market for lemons.  By looking only at the low-cost end of the market, and only looking at what people are trying to sell, I’m probably looking at a fairly biased sample of all Leafs.

Unfortunately, that’s the sample of cars I’m buying from, if I continue down this path.  Perhaps time for a re-think.

Post #1816: Bow rehairing finale, fitting the tip wedge.

 

In brief:  I finished rehairing a cheap fiberglass bow by making and installing a version of the tip wedge from this YouTube series by Gilles Nehr.

It’s crude.  I used low-grade poplar wood from Home Depot, and shaped it using sandpaper.

But it works.  My cheap bow is now (inexpertly) rehaired, as shown below.

The one thing I can say for sure is that this goes a lot faster once you have a little experience with it.  Making and fitting the tip wedge took less than an hour.

Continue reading Post #1816: Bow rehairing finale, fitting the tip wedge.

Post #1815: Wedges. Tediously carving tiny wooden wedges.

 

This is the first time I’ve ever rehaired a violin bow.  In this case, it’s a cheap fiberglass bow.  Nothing much will be lost if I screw it up.

This is orders-of-magnitude harder than I thought it would be.  I very nearly gave up when I finally figured out everything the task entails.

Good sense would dictate abandoning this  project.  Pride goads me onward.  I refuse to call it quits and admit my mistake.

We’ve fought lengthy wars based on more-or-less the same rationale.  So don’t judge me.

In any case, this entire post is about carving and fitting tiny little wooden wedges.  The critical wedges that keep the horsehair in place, on the bow.

Crafting precisely-shaped pieces of wood the size of  popcorn kernels, using full-sized hand tools.

It’s every bit as hard as you might imagine. This is the step at which I balked originally.

Consult the checklist:

  1. Do those little pieces of wood go flying across the room when you cut them with a chisel?
  2. Are those flying proto-wedges then virtually indistinguishable from waste wood chips that litter your work area?
  3. Will you sand a couple of strokes too much, thus rendering the object uselessly loose?
  4. Will you sand your fingertips to the point of pain, as you try to shape the wood on sandpaper?
  5. Will these tiny fragments of wood occasionally shatter, as you try to cross-cut them with a chisel?
  6. Will you have to re-cut each one several times, in an attempt to get the fit just right?

I assume you get the drift.  Making and fitting the wedges, for the first time, is the worst kind of fiddly work.

An infinite number of monkeys, with an infinite number of chisels and an infinite amount of time, will eventually carve an exact replica of the Statue of Liberty.

And yet, that is not a recommended production technique if you run a gift shop.

I’m sure there are artisans who can knock these out in a trice.

Me?  Start to finish, it took me about three hours to make and fit the wedge that goes into the heart of the frog. 

The first trick is to know your limits.  It’s just not possible to cut a piece of wood, to the required tolerances, using hand tools, in one go.  Not for me, anyway.

In my case, I first split inch-long shards of wood of a segment of quarter-inch basswood poplar board, until I got one that was a snug fit for the width of the hole into which the wedge must fit.  Then cut an over-long piece off that, and went to work with sandpaper, on that tiny fragment of wood.

Work it until it fits, or you’ve lost it on the floor, or you’ve made it too small and have to start over. Whichever comes first.

With enough monkeys, chisels, and time, you will eventually get one that fits.  I think my winner was maybe my fourth attempt.

Wordless workshop.

 

A few further observations.

Some people start at the tip of the bow.  Seemed like most started at the frog.  Frog wedge looks much easier to do than the tip wedge.  So that’s where I started.

You can buy ready-made “wedge blanks”, for just a few dollars.  That is, you can buy the chips of wood, roughly the right size and shape.  But my take on it is that the bulk of labor is in achieving the exact fit. You’ll put in that labor whether you buy a ready-made blank or make the rough chip yourself.

You cut the wedge so that the grain of the wood runs parallel to the length of the bow stick.  This is to prevent the wedge from falling out in drier weather.   Wood shrinks as humidity falls, but (essentially) only across the grain.  Cut this way, the width of the wedge may vary, but the critical length dimension should remain very nearly constant, regardless of humidity.  I kept losing track of which which direction was which, on the wedge, so I eventually colored the top face of the board with sharpie, so I would know which end was up.

Only one end of the hair comes bound, like that.  The other is free because you have to cut the hank to size, to fit your bow.

You have to shove the bound end of the hair, into the cavity, so that the hair makes a right angle turn and the bound end of the hair sits flat on the bottom of the cavity.  To make the end of the hank of hair more flexible, wet it, give it half a minute, then towel it dry.  Use a sliver of wood or other dull object to press the end of the hank into place.  I assure you that a screwdriver will cut horsehair under these conditions.

The frog slide — the bottom plate of the frog — has to slide over the hair, as it sits on top of the wedge. This means that the top of the wedge must be more-or-less flush with the top of the hole in which it sits.  I lubed the frog slide with powdered graphite (sold as lock lubricant).  That was a mess.  I now see that savvier people get the same effect just by running a pencil down the slot into which the frog slide sits (reference).

But, ultimately, the wedge is what holds the horsehair in place.  So, at the end of the day, you have to answer the following question:

Why are they called wedges?

This little wooden plug, inside the frog.  Why is that called a wedge? It’s not obvious.  Why is this a “wedge” and not a “plug”, say.  Sure looks like a little wooden plug.

The reason is that you don’t insert the wedge vertically into its respective hole.  Like a plug or a cork.  Wrong-o.  These are not plugs or corks.

In fact, if you cut it just right, you can’t insert it vertically.  It should be just a hair too long for that.  You insert the edge of the wedge furthest from the horsehair first, leaving the wedge tipped up slightly, high end resting on the hair.  Then you press it flat, and in.

That motion — like closing a trap-door, as you push the wedge into place  — locks the hair firmly against the wood of the bow or frog.

That’s what makes it a wedge.  And that’s why the length dimension is critical, but the width, not so much.  The distance between where the wedge meets the hair, and the back of the cavity, has to be just right.

That’s the theory.

In practice, expect some trial-and-error.  The wedge is too big, so it won’t fit.  Or too small, so it pops right back out.  Or it pops out and gets lost in the rug.  Or it fits, but you can’t squeeze the flat plate on the frog over top of it.  And repeat.

Conclusion:  That was the easy one.

Good news:  Just three hours later, and one wedge is done and fitted. 

Bad news:  That’s the easy one.

Engineering-wise, the modern violin bow frog clearly has a lot of safety and redundancy built in.  Multiple systems keep the horsehair in place.  The hank of horsehair is both bound to itself (so individual hairs can’t pull out), and mechanically attached to the frog (so the entire hank can’t pull out, once the bottom plate is slid back into place.)  And there’s yet another wedge — with glue, yet — to be driven under the ferrule of the frog.  The pressure of the internal frog wedge merely aids those existing systems.

In short, this looks like overkill, for a modern bow frog.  I’m guessing it’s done this way because this is the way it has always been done. Perhaps for older styles of frog, or those lacking the precision machining of a modern frog, or with horsehair that did not come pre-bound, the internal wedge fit matters more.  But for this frog, from an engineering standpoint, it sure looks like any hard object that fills that hole would have kept the hank in place just fine.

The bow tip, by contrast, has zero redundancy.  All that stands between the violinist, and performance disaster, is a single, tiny, carefully-fitted wooden wedge.

Which, by itself, is a reason to start at the frog, if this is your first bow.

Finally, in hindsight, I completely understand why the manufacturer glued the tip wedge in place, on this cheap bow.  I may end up doing the same.  We’ll see how it goes.

Post #1814: The unanticipated skill requirements for rehairing a violin bow.

 

Maybe I should just stop watching YouTube.

On the one hand, I learn a lot.

On the other hand, I often learn just enough to get myself into trouble.

Such is the case with re-hairing a couple of violin bows.

The technology of the fiddle bow hasn’t changed in a couple of centuries.

What could possibly go wrong? Continue reading Post #1814: The unanticipated skill requirements for rehairing a violin bow.

Post #1813: Won’t be hiring ChatGPT as my AI staff writer, either.

 

AI: It’ll always give you an answer.  It might give you a very convincing answer.  And it might sometimes even give you the right answer.  But the last part is far from guaranteed.

Inconceivable!

In my ongoing quest to automate this blog, I threw my test questions at ChatGPT.  These questions are:

  1. Explain why ceiling fans are more efficient than box fans.
  2. Explain why box fans are more efficient than ceiling fans.

I already know the answer.  Ceiling fans are more efficient, and I already worked out why, in a post that I have currently withdrawn.  To move the same volume of air per minute, a small fan has to move the air much faster, and that eats up energy, compared to moving a larger volume of air, more slowly.  Ceiling fans are more efficient at moving a given volume of air simply because they are larger, so they can move that required volume of air at a much slower speed.

To save you the reading, ChatGPT produces nicer text, and more of it.  But at root, the answers had the same problems as Bard.ai.

  • Just like Bard.ai, ChatGPT will do its best to give you the answer you’re looking for.  Even if you’re wrong, as in the second question above.
  • Just like Bard.ai, ChatGPT conflates the definition of efficiency (CFM/watt) with the reason for efficiency.
  • Just like Bard.ai, I get a lot of extraneous stuff that I now realize probably came from various ads and marketing materials.  Aerodynamic blade shapes and whatnot.

Even when I tried to point ChatGPT in the exact direction I wanted — explain the physics of why ceiling fans are more efficient in terms of CFM/watt — I got back nothing useful.

Worse, when I gave it the “wrong ” question, and specifically worked it in terms of CFM/watt, damned if ChatGPT didn’t spit back the wrong answer, forcefully and clearly worded in terms of CFM/watt.

Worse yet, when I gave it a neutrally-worded question, it did in fact pick ceiling fans over box fans.  But the explanations were just so much nice-sounding nonsense.  It never did get down to the basic physics of it.

But even worse, when I gave it a different, simpler wording, of that neutrally-worded question (“Which is more efficient …”), it gave me the wrong answer.  And backed that up with nonsense reasoning.  (e.g., Box fans are more efficient in terms of CFM/watt because they are compact and portable.)  That’s the last fan question entry below.  So not only did it seem to give me either answer (ceiling is more efficient than box, box is more efficient than ceiling), but which answer I got appeared to depend on irrelevant and arbitrary changes in the wording of the question.

Finally, I gave it a couple of questions for which a) I have a well-researched and definitive answer, and b) for which I am a national expert.  Does the hospice benefit reduce Medicare spending?

The plain answer to that is no, but you have a whole lot of people who wish the answer were “yes”, and a whole lot of bad research that seems to give them that answer.  (Along with considerable research supporting the answer of “no”.) So when I asked ChatGPT, twice, what I got back was a highly selective reading of the scholarly literature, and the usual pat answers about why hospice saves Medicare money.

(It doesn’t.  Among other things, a third of Medicare hospice outlays are for those who are outside of their last year of life.  Other than my own study, typical research methodologies exclude these costs.)

My take on it at this point is that if you want bland filler, that does in fact say some stuff, about some defined topic, these writing AIs are fine.

If you want the contents to be factually correct?  Not so much.  The answer you get depends heavily on the question you asked.   Seemingly neutral changes to a neutral question got ChatGPT to flip its answer.  And for a separate question where I have expert knowledge, ChatGPT presented a selective reading of the literature that more-or-less mimicked the for-profit industry’s party line.

To me that says that you really can’t ask these writing AIs for reasoned analysis.  Or right and wrong.  At best, what you’re really asking them is one of two things:

  • Give me some arguments that support my point of view.
  • What does the majority of text, that has been input to this AI, say about a particular topic.

You may think you’re asking it to provide you with some true information.  But at root, what you’re actually doing is a lot closer to one of the two bullet points above.

This, by the way, dovetails with a friend’s report of AIs that had been trained to be racist, by groups of individuals who chatted with the AI using racist language.  The AI was just a reflection of what it had been fed.

If you are looking for any sort of deep explanation, you’re out of luck.  Even if you know what that deep explanation is.  In terms of explaining why ceiling fans are more efficient, none of these went much beyond the tautology of telling me that they are more efficient because they move more CFM/watt.  Which is, in this context, the definition of efficiency.

I wonder if this explains what I see on blog sites when I’m researching a topic in gardening.  On any given topic, a lot of sites will reproduce some bland version of the standard talking points.  Many of them don’t seem to have first-hand knowledge of the topic.  And for sure, almost none do what I do, and actually test things to see if they are true or not.  Maybe a lot of what I’ve been seeing is based on AI-generated text.  A lot of that text has the same bland, one-hand-or-other-hand style that seems to be the default style of ChatGPT.

In short, decent writing?  Yes.  Good sense?  No.

But, in hindsight, these are billed as writing AIs.  They write.  That’s what they do.  Plausibly, you could use them to generate multiple answers to a question, and you’d pick the one you liked.  But in no sense that I see can you trust them  them to answer a question, in terms of sorting fact from fiction.

The might get it right.  They might not.  They might waffle.  For sure, if you ask them the same question, enough ways, you might get back an answer that you like.

If the AI manages to hit the nail on the head, it provides you with cheap, well-composed text.  But it’s going to be up to you to judge whether the AI nailed it, or flubbed.

It’ll always give you an answer.  It might give you a very convincing answer.  And it might sometimes even give you the right answer.  But the last part is far from guaranteed. 

It boils down to this.  In trying to find an artificial intelligence to write my blog, I keep using “AI”, but I do not think “AI” means what you think it means.  For sure, AI writers don’t do what I had hoped they would do.

Here’s how it went.

Explain why ceiling fans are more efficient than box fans.

ChatGPT response, emphasis mine

Ceiling fans are generally considered more efficient than box fans due to several factors:
  1. Air circulation: Ceiling fans are mounted on the ceiling, allowing them to circulate air throughout the entire room. They create a downward airflow that creates a wind chill effect, making you feel cooler. This circulation helps maintain a consistent temperature in the room and reduces the reliance on air conditioning or other cooling systems. On the other hand, box fans are typically placed on the floor or a surface, and their airflow is limited to the immediate area in front of them.
  2. Coverage area: Ceiling fans have larger blades and are positioned overhead, allowing them to cover a larger area compared to box fans. They can effectively cool an entire room by distributing air more evenly. In contrast, box fans are smaller and have a limited range of coverage. They are better suited for localized cooling or for smaller spaces.
  3. Energy consumption: Ceiling fans are generally more energy-efficient compared to box fans. They consume less power to provide the same level of cooling due to their design and the way they circulate air. Ceiling fans typically use around 15-90 watts of electricity, while box fans can consume 50-200 watts depending on their size and settings. This lower energy consumption translates into cost savings and reduced environmental impact.
  4. Integration with HVAC systems: Ceiling fans can complement the existing HVAC (heating, ventilation, and air conditioning) system in a room. By using a ceiling fan in conjunction with an air conditioner, you can set the thermostat at a slightly higher temperature while still feeling comfortable due to the cooling effect of the fan. This can lead to significant energy savings by reducing the load on the air conditioning system. Box fans do not integrate with HVAC systems in the same way.
  5. Convenience and space-saving: Ceiling fans are fixed in place and do not require additional floor or surface space, making them more convenient and space-saving compared to box fans. They are also operated by a switch or remote control, allowing for easy adjustment of speed and direction. Box fans, on the other hand, need to be placed in a suitable location, and their portability can be a disadvantage in terms of convenience and aesthetics.
It's important to note that the efficiency and effectiveness of fans can vary based on the specific models, settings, and conditions. However, in general, ceiling fans tend to provide more efficient and widespread cooling compared to box fans.

Continue reading Post #1813: Won’t be hiring ChatGPT as my AI staff writer, either.

Post #1810: Top 25 AIs for fill-in-the-blank? When did this happen?

Let me say that I don’t ever use a grammar checker.  My wife refuses to use spell-check.  Says she, “I think that an educated person should be able to write clear English.”

Yet, on a lark, I decided I’d have an AI write my next blog post.

Still working on carrying through on that.  It ain’t as easy as I thought.

Continue reading Post #1810: Top 25 AIs for fill-in-the-blank? When did this happen?

Post 1808: Some thoughts on AI, part 1.

 

AIs of various sorts seem to be able to do a lot of jobs that traditionally required the use of a human mind.  They are particularly good with text generation.  They can also easily do “literature reviews”, to offer a summary of current understanding or thinking regarding a topic.  Some say they are good at writing computer code, but I’d like to see evidence of that before I’d believe it. Continue reading Post 1808: Some thoughts on AI, part 1.