Category: Engineering

Texas flood lives could have been saved with a smartphone app that tailored alerts to exact GPS location?

by philg

A year ago, in the context of hurricane evacuation: Why not a simple web site or phone app to determine whether one must evacuate?

In the latest Texas flooding tragedy, it seems that emergency alerts were sent out to mobile phones. As with hurricane warnings, however, it fell to the individual to determine whether the phone’s current location merited going back to sleep or running for higher ground at top speed.

Why can’t it be a computer’s job to intersect the alert with the phone’s location and, ignoring any sleep silence modes, recommend a definitive course of action? (A year ago, I asked “Why is it a human’s job to do something that can be done much more reliably by a computer?”)

The Wall Street Journal says that flash flood warnings went out, but they weren’t specific:

The National Weather Service said that on Thursday morning it briefed emergency management. By 1:18 p.m. it issued a watch that said locally heavy rainfall could cause flash flooding across eight counties, including Kerr, where Camp Mystic, the girls summer camp that was heavily struck by the flooding, is located. Kerr County is dotted with riverside summer camps as well as recreational-vehicle parks and hotels for vacationers. Overnight it would become the most severely impacted county.

At that time, forecasters expected a maximum of 3 to 7 inches of rain.

The first flash-flood warning—which means flooding is imminent or already happening—came at 1:14 a.m. Friday from the National Weather Service office in nearby San Antonio hours ahead of the Guadalupe’s rise.

Two hours later, the office issued a catastrophic warning, or a flash-flood emergency, for the region.

The warning covered “eight counties” and, in fact, the vast majority of people in those eight counties didn’t have anything to fear. Kerr County alone is 1,100 square miles. The typical resident of the U.S. is not a hydrologist. Why isn’t it a computer’s job to figure out whether a phone is currently located in a river’s floodplain and, if so, provide specific directions regarding how to reach higher ground?

Here’s a FEMA flood map showing what I think is part of the camp at upper right. It shows that there is always an “area of minimal flood hazard” close to any place that has been calculated to be at risk of flooding.

Here’s a photo from the New York Times that shows how close a lot of camp buildings were to areas of perfect safety and how critical a person’s GPS location would be to the safe/unsafe decision.

Below is an example warning from the National Weather Service’s X account, which seems to require a lot of interpretation, e.g., knowledge of where one’s location is relative to the Guadalupe River. A typical young American doesn’t know north from south (I remember calling a mobile carrier store on Route 9 in the Boston suburbs and asking the clerk if he was on the north or south side of this road, which run east-west. He was completely stumped by the question and, even after seeking assistance from some fellow workers, couldn’t answer it.) Also, the warnings were issued only in English and Spanish. In a country with an asylum-based immigration system there is no reason to expect someone living in the U.S. to speak either English or Spanish. A phone app, however, can work in any language supported by a smartphone (77 for Android; roughly 40 for iPhone).

The official government attitude seems to be to keep doing whatever failed in the past. In a world where almost everyone has a smartphone (taxpayer funded for those who don’t work), the go-to idea is an early 1900s-style siren system (PBS):

Another idea from the same PBS article is that more humans can be a substitute for a computer system:

Full post, including comments

Why isn’t there a mesh of water leak sensor tape on roof deck plywood?

by philg

The “roof is failing” sensor in a house is typically a homeowner noticing a stain on a ceiling.

Wouldn’t it make more sense to put down sensor tape on the plywood roof deck before the peel-and-stick material, shingles, tiles, or whatever are applied? If there is a leak in the roofing system or flashing and water gets down to the wood layer there can be a notification of exactly where the leak is happening.

Even if mass-produced by our brothers, sisters, and binary-resisters in Asia this wouldn’t be cheap, but I still think it would make economic sense given the cost of a roof ($15,000-$150,000) and the cost of repairing water damage in a society where the average skill level falls each year.

It’s an obvious idea so why hasn’t it been done?

Related:

Full post, including comments

Climate Change Reading List: Johnstown Flood

by philg

Earth’s population is heading toward 10 billion and beyond (nobody knows if we’re already there). This level of density requires ever more complex engineering with ever higher stakes in the event of engineering failure. Let me therefore recommend as timely a 1989 book by David McCullough, author of Path Between the Seas (fantastic book about the Panama Canal): The Johnstown Flood.

I can’t quote precisely from the book because I listened to it on Audible. The short story is that Johnstown was part of a canal route from Philadelphia to Pittsburgh. The canal needed a reliable water supply for its locks and, therefore, the state built an artificial lake 450′ above the then-small town (at 1,161′ above sea level). The railroads rendered the canal and lake unnecessary so the state sold the lake and associated earth embankment dam to some rich douches from Pittsburgh who wanted to sail and fish in the mountains during the summer. The douches repaired the dam incompetently, removed outflow pipes so that they couldn’t control the water level, and then let the dam fill up much higher than it ever had. The result, of course, was that the dam failed on May 31, 1889 during some heavy rain and wiped out what had become a town of 10,000+ inhabitants. More than 2,200 people were killed.

The risk was foreseen by even some casually competent engineers who looked at the reconstructed dam, but everyone became complacent.

We’re much smarter today and, therefore, this kind of thing can’t happen to us? It actually happened again in Johnstown itself. There was a significant ordinary flood in 1936. FDR sent in the Army Corps of Engineers and promised that the town wouldn’t be flooded again… which it wasn’t until 1977.

Another interesting aspect of the book is how effective private relief efforts were. The flood did about $17 million of damage in the dollars of the day and people, not yet subject to income tax, voluntarily contributed about $4 million in cash relief. In addition, trainloads of volunteers and supplies were provided to clear debris and rebuild the town. Private companies, especially the railroad, made major contributions. The state government didn’t do much other than help maintain order. The federal government did nothing at all in the way of relief. Today, by contrast, we give a high percentage of what we earn to the Feds and state and then try to get some of that money back when there is a problem. It certainly deprives us of the satisfaction of being charitable and of the pride from volunteering. Who among us can do something significant for storm or flood victims compared to FEMA?

More;: read The Johnstown Flood.

Full post, including comments

Why don’t iPhone and Android incorporate AI for answering phone calls?

by philg

The United States managed to destroy the utility of Antonio Meucci‘s invention, the electromagnetic telephone. We did this via a combination of (1) free calling, and (2) failure to authenticate callers. This has enabled half of India and their robots to work as phone scammers.

Why not simply turn off conventional phone calls completely? Health care, banking, and some other essential-for-most-people services still rely on this now-useless technology.

But why can’t the phone itself screen all calls and make phone scamming unprofitable? An AI resident on the phone could put through calls from recognized numbers and silently answer the rest of the calls to see if a legitimate human is on the other end. It should be able to quickly learn to recognize folks offering home renovation, solar panels, final expense insurance, Medicare benefits, etc., from some combination of long wait before caller speaks, Indian accent, saying “the reason of my call” instead of “the reason for my call”, and the use of previous scripted phrases. The AI could be programmed to lead on scambots and the human scammers behind them (“transfer to my senior supervisor”) so that 5-10 minutes of their time is wasted without any personal information being divulged. Answer

It seems as though there are some third party apps vaguely trying to do this, but since answering a phone call is a core function of an iPhone or Android phone, shouldn’t the capability be built directly into the operating system?

“Who’s Making All Those Scam Calls?” (New York Times, January 27, 2021):

I flew to India at the end of 2019 hoping to visit some of the call centers that L. had identified as homes for scams. Although he had detected many tech-support scams originating from Delhi, Hyderabad and other Indian cities, L. was convinced that Kolkata — based on the volume of activity he was noticing there — had emerged as a capital of such frauds.

Late in the afternoon the day after I met with Nath, I drove to Garden Reach, a predominantly Muslim and largely poor section in southwest Kolkata on the banks of the Hooghly River. Home to a 137-year-old shipyard, the area includes some of the city’s noted crime hot spots and has a reputation for crime and violence. Based on my experience reporting from Garden Reach in the 1990s, I thought it was probably not wise to venture there alone late at night, even though that was most likely the best time to find scammers at work. I was looking for Shahbaz.

Full post, including comments

Grafitti Drone countermeasures?

by philg

On a recent trip to San Francisco, a local friend took us to Andy Goldsworthy’s Spire in the Presidio:

We had just come from a parking lot where quite a few cars were virtuously marked:

Our friend said “What we need is a drone to paint the Spire sculpture in the Palestinian flag colors.”

Let’s suppose that residents of the U.S. with a lot of community spirit did build some drones that could paint the sides of building with huge messages such as “From the River to the Sea, Palestine will be Free”, “#Resist”, and “Trump is a Nazi.” It is much easier to build a spray-painting drone than a scrubbing drone, I think. How could cities and building owners defend against virtuous painting drone owners/operators?

(Though moderately rich by average American standards and blessed with a garage at home, our friend who lives in SF drives a 22-year-old car for fear that anything nicer will attract thieves.)

Full post, including comments

Power restoration after Hurricane Milton

by philg

Landfall: October 9, 2024 at 8:30 pm. Supposedly about 4 million customers lost power (source: Ron DeSantis press conference).

Mid-afternoon the next day:

Early evening:

The morning of the second day:

Apparently there was power at the big airport in Tampa because they resumed operations about 36 hours after the hurricane made landfall:

(Orlando had reopened a few hours earlier, so they too had power despite being in the middle of the Band of Destruction (TM).)

Afternoon of second day:

About 48 hours after the hurricane hit, the total customers out has declined from 3.4 million to 2 million:

The bad news is that restoration for some Floridians won’t be until 8 days after the hurricane made landfall. Here’s FPL’s estimate:

I’m not sure if people in neighborhoods with underground lines (like ours!) will get power sooner. Currently, 10 percent of FPL’s customers are out versus 17 percent for the state.

I can’t figure out why the customer numbers are so high. I thought that the transmission lines were designed to handle hurricane-force winds (and they were further beefed up after 2019; see Tough questions from reporters for Ron DeSantis). Maybe there are a lot of neighborhoods with above-ground powerlines for local distribution?

Strong independent female linewomen continued to work through the night, apparently…

2.5 days after landfall, it looks like Naples and Fort Myers are on their way back to normal while half of Tampa is dark. More than half of the Floridians who originally lost power now have it back (thanks once again to the efforts of linewomen who identify as female):

around lunch time…

Three days (72 hours) after landfall:

The pace of restoration seems to slowed down in the dark:

3.5 days after landfall:

Not a great situation in Tampa, with more than one third of customers without power. On the other hand, the total is down below 1 million compared to 4 million at the start.

Four days (96 hours) after landfall and about 500,000 customers are still out. More than 235,000 of them are Tampa Electric customers, which has only 840,000 total customers.

Florida Power and Light now says that they’ll have nearly everyone restored, even in directly hit Sarasota, by Tuesday night. (Also that they’ve thus far restored 90 percent of their affected customers, 1.8 million people who’d lost power at one point.) Speaking of FPL, if you were to watch their X feed you’d learn that electricity restoration is definitely not something that white males do:

And as of Tuesday at noon, FPL indeed had all but 38,000 customers back online. Tampa Electric (TECO) continued to be an outlier with 100,000 customers still dark.

Full post, including comments

Follow-up to American Factory, Taiwan edition

by philg

Happy Labor Day to those who celebrate by working!

Back in 2020, I covered Netflix: American Factory, a documentary of what happens when Chinese glass-making experts try to train Americans to be useful and also what happens to foreign investors when Democrat politicians circle the investment. In case you missed it, the New York Times ran an interesting follow-up to this movie: “What Works in Taiwan Doesn’t Always in Arizona, a Chipmaking Giant Learns” (August 8, 2024).

TSMC modeled its facility in Phoenix on one at home. But bringing the company’s complex manufacturing process to America has been a bigger challenge than it expected.

“We keep reminding ourselves that just because we are doing quite well in Taiwan doesn’t mean that we can actually bring the Taiwan practice here,” said Richard Liu, the director of employee communications and relations at the site.

In recent interviews, 12 TSMC employees, including executives, said culture clashes between Taiwanese managers and American workers had led to frustration on both sides. TSMC is known for its rigorous working conditions. It’s not uncommon for people to be called into work for emergencies in the middle of the night. In Phoenix, some American employees quit after disagreements over expectations boiled over, according to the employees, some of whom asked not to be named because they were not authorized to speak publicly.

While it was under construction, the company sent American engineers to Tainan for training and to shadow their Taiwanese counterparts, observing TSMC’s all-hands-on-deck way of working up close.

Jefferson Patz, an engineer fresh off a master’s degree from the University of California, San Diego, went to Tainan in 2021 for 18 months of training shortly after he joined the company.

“Oh, my gosh, people work hard,” Mr. Patz said. He recalled that this initial impression had given him a strong sense of what it took to succeed in the industry.

After returning to Arizona, Mr. Patz said, employees were expected to pitch in with work outside their job descriptions because construction of the facility was behind schedule.

This approach did not sit well with everyone. Workers were required to do whatever was needed to finish the most pressing job, he said. Some of the American workers also found it difficult to spend a long stretch of time in Taiwan.

TSMC should be able to make this work simply by paying $2 million/year to each worker in order to get smart conscientious people from among the U.S. population of 335 million (or maybe 350 million if we count the undocumented more accurately), but that could be a painful hit to profits! Let’s check out the labor pool in Taiwan vs. Arizona. World Population Review:

The same source gives an average IQ for Arizona of 98, substantially lower than Taiwan’s average of 106 (I question the use of 5 digits of precision, but maybe someone with a higher IQ than mine prepared the above table). American average IQ is falling, so the spread between AZ and Taiwan will only get worse.

How about conscientiousness, the willingness to show up to work every day and try to do every step of a procedure correctly? That’s heritable and the Americans with the highest fertility are those who barely work (source; the high fertility of those earning $300,000+/year can be ignored on a population-wide basis because there aren’t a lot of those parents).

Maybe it won’t matter for profits how ill-suited the average American worker is to working in a state-of-the-art fab because TSMC will be so stuffed with U.S. tax dollars that they can pay to get the workers they need.

Full post, including comments

Why you’re likely safer on a Panamanian- or Liberian-flagged ship than an American ship

by philg

A maritime safety lecture from Into the Storm: Two Ships, a Deadly Hurricane, and an Epic Battle for Survival by Tristram Korten…

SS El Faro’s hull, a towering wall of blue-painted steel, loomed over the wharf at the Port of Jacksonville’s Blount Island terminal as gantry cranes loaded her decks with cargo containers. She was a steamship (designated by the “SS” before her name), using two large boilers to power a single-propeller shaft. And she was old, built by the Sun Shipbuilding and Dry Dock Company in Chester, Pennsylvania, just south of Philadelphia, which rolled her into the Delaware River for service on January 1, 1975. That put her in the minority of big ships [in 2015]—less than 9 percent of the world’s merchant fleet is over twenty years old.

This doesn’t sound good for structural integrity:

Twenty years later, in the mid-1990s, she was hauled into a shipyard in Alabama, where her mid-body was lengthened to increase cargo capacity.

The 790-foot El Faro didn’t make it past her 40th birthday, sadly, but it turns out that she would almost certainly have been scrapped many years prior to the dramatic events of 2015 if not for U.S. laws to restrain maritime trade.

Sun Shipbuilding has since closed, a casualty of America’s decline in manufacturing, leaving a dwindling number of shipyards able to construct big cargo ships in the United States, which also means a dwindling number of shipyards capable of fulfilling the requirements of the Jones Act, a 1920 law requiring that any cargo transported from one U.S. port to another must travel on ships that are American built, American crewed, and American owned. (Puerto Rico, being a U.S. territory, counts as a U.S. port.) The law was designed to protect America’s supply routes during times of war. Today its primary effect is to protect the jobs of American sailors, preventing companies from hiring much cheaper crews from Third World countries. But there is a cost—and it is steep. To build a Jones Act ship costs $120 million to $140 million. To build the same ship in South Korea, which is a developed nation, would cost about $32 million, according to Court Smith, an industry analyst with Shipping Intelligence and Analytics. It’s even cheaper to build one in India or China. South Korea builds roughly two hundred commercial ships a year, according to Smith. America puts out maybe four. As a result, shipping companies pushed the life spans of their expensive American-made ships to the absolute limit. The average age of the U.S.-flagged cargo fleet is thirty-three years, compared to thirteen years for the global fleet, according to UN statistics, and most shipping experts say the average age a cargo ship is retired worldwide is around twenty years. El Faro was a product of this dynamic. Due to its age, it was allowed to remain outdated in certain areas. For example, a regulation requiring new ships to carry enclosed lifeboats was waived for older ones, for which compliance would require a costly retrofit. Grandfathered in, El Faro continued to carry two old-fashioned open-top lifeboats. Likewise, the ship’s emergency position-indicating radio beacon, or EPIRB, did not have to be encoded with GPS, which would give the ship’s position in a time of distress.

In addition to putting American sailors lives’ at risk, the Jones Act dramatically drives up costs for businesses and individuals in Puerto Rico, Alaska, and Hawaii. In addition to the obvious costs described above, there is the non-obvious cost that the closed market facilitates collusion:

For two decades, Sea Star had been one of several shipping companies that sailed supplies to Puerto Rico on a regular schedule. All that competition meant slender margins and low profits. But companies stayed with the route to Puerto Rico—travel between a U.S. state and a U.S. territory—because they had invested so much to comply with the Jones Act. There were no dodgy flags of convenience for El Faro; she flew the Stars and Stripes. That was a high barrier to entry for would-be competitors—they would need an American-built ship, crewed by U.S. sailors. Then, in 2002, one of those firms, Navieras, went bankrupt. Suddenly the companies still afloat—including Sea Star, Crowley, Trailer Bridge, and Horizon—started seeing increased business and profits for the first time in decades. Rather than risk losing their newfound earnings to any potential newcomers, the companies bought Navieras’s ships, and executives from at least three of the companies, Sea Star, Crowley, and Horizon, conspired to fix their prices. They created secret email accounts to communicate, and set up spreadsheets that kept track of their rates, which increased by as much as 30 percent—so that, as Forbes magazine wrote, “they could assure that nobody was cheating, while they were cheating.” They weren’t clever enough to fool the FBI, however, which got involved after learning of a meeting between executives from the competing firms.

Even a newly built South Korean ship would have had some trouble handling what the American captain and crew did with El Faro, but the new ship might not have sunk and, if it had, the modern lifeboats would have given the crew a chance to survive.

Full post, including comments

Why doesn’t anyone want to buy Intel’s Gaudi AI processors, supposedly cheaper than Nvidia’s H100?

by philg

Intel claims to have a faster and more cost-effective AI system than Nvidia’s H100. It is called “Gaudi”. First, does the name make sense? Antoni Gaudí was famous for doing idiosyncratic creative organic designs. The whole point of Gaudí was that he was the only designer of Gaudí-like buildings. Why would you ever name something that will be mass-produced after this individual outlier? Maybe the name comes from the Israelis from whom Intel acquired the product line (an acquisition that should have been an incredible slam-dunk considering that it was done just before coronapanic set in and a few years before the LLM revolution)?

Intel claims that their Gaudi 3-based systems are faster and more efficient per dollar and per watt than Nvidia’s H100. Yet the sales are insignificant (nextplatform):

Intel said last October that it has a $2 billion pipeline for Gaudi accelerator sales, and added in April this year that it expected to do $500 million in sales of Gaudi accelerators in 2024. That’s nothing compared to the $4 billion in GPU sales AMD is expecting this year (which we think is a low-ball number and $5 billion is more likely) or to the $100 billion or more that Nvidia could take down in datacenter compute – just datacenter GPUs, no networking, no DPUs – this year.

Nvidia’s tools are great, no doubt, but if Intel is truly delivering 2x the performance per dollar, shouldn’t that yield a market share of more than 0.5 percent?

Here’s an article from April 2024 (IEEE Spectrum)… “Intel’s Gaudi 3 Goes After Nvidia The company predicts victory over H100 in LLMs”:

One more point of comparison is that Gaudi 3 is made using TSMC’s N5 (sometimes called 5-nanometer) process technology. Intel has basically been a process node behind Nvidia for generations of Gaudi, so it’s been stuck comparing its latest chip to one that was at least one rung higher on the Moore’s Law ladder. With Gaudi 3, that part of the race is narrowing slightly. The new chip uses the same process as H100 and H200.

If the Gaudi chips work as claimed, how is Intel getting beaten so badly in the marketplace? I feel as though I turned around for five minutes and a whole forest of oak trees had been toppled by a wind that nobody remarked on. Intel is now the General Motors circa 2009 of the chip world? Or is the better comparison to a zombie movie where someone returns from a two-week vacation to find that his/her/zir/their home town has been taken over? Speaking of zombies, what happens if zombies take over Taiwan? Humanity will have to make do with existing devices because nobody else can make acceptable chips?

Related:

Full post, including comments

Why aren’t there front load washing machines with the same depth as top load washing machines?

by philg

A lot of houses have closets and hallways architected to fit a standard top loading washing machine, which seems to have been roughly 27x27x44″ high. Here’s a 3.9 cu. ft. Whirlpool:

By cheating just a little, e.g., stretching the depth to 27 and 7/8″, Whirlpool can deliver a machine with a 5.2 cu. ft. drum.

It would seem obvious to build a front load washing machine with the same 27×27″ footprint, but nobody seems to do that. One of Whirlpool’s smaller front loaders is advertised as “closet-depth” and, in fact, is about 31.5″ deep. Their bigger front loaders are over 33 inches deep with the door closed. If you scale down to a “compact” front loader, as seen in Europe, the footprint is 24×24″ and capacity drops to just 1.9 cu. ft.

What’s the engineering challenge to making a front loading washer that exactly fits the footprint of a legacy washing machine and, thus, fits into an older American house as it was originally designed?

(Our house is an example of one in which 27-inch depth is the limit. The laundry room connects the family room/kitchen to the garage. A machine deeper than 27 inches will stick out beyond the door frame (top of the figure below) and obstruct access into the garage:

In fact, the only way to have 27-inch deep machines not poke into the hallway is to dig into the 4-inch drywall behind the machines, e.g., to make water and gas connections. Everything must be perfectly positioned for the machines to sit flush on the baseboard.

In other fun appliance news, an architect who redesigned our Harvard Square apartment’s kitchen notched out cabinets to precisely fit a particular LG fridge that we bought back in 2013. The fridge has French doors, which introduces another point of failure beyond a conventional side-by-side fridge or bottom-freezer fridge. The “mullion door” or “flapper door” in the middle of the French doors had a failed spring. I thought about buying a replacement, but was concerned that the notches wouldn’t work for the new fridge and also I couldn’t find any current fridges that had stainless steel sides as the old one did. Thus, it was time to think about repairing the minor problem with the 11-year-old fridge.

I was renting it out on AirBnB, couldn’t get up to Cambridge to fix it myself, and didn’t want to hassle my friend from MIT who is a mechanical genius but has his own 130-year-old 3-story wooden house to maintain. I called LG service and they offered a fixed $399 flat-rate repair fee. I gave them the model and serial number in advance and told them exactly what the problem was and what part was needed. The technician came out to the apartment, diagnosed the problem as the flapper door, and then said that no replacement part was available (the LG Parts web site showed a compatible replacement and eBay had the exact part number available from an appliance store that apparently had a lot of old stock). While monkeying with the fridge, he managed to short out the control board so the fridge went from “tough to close” to “completely dead.” The flapper door actually has an electrical connection to the control system in order to run a heater that prevents condensation from forming on the door. I then asked a series of people who answered LG’s 800 number with thick Indian accents whether I could perhaps get a refund of the $399 repair fee since they themselves acknowledged that they hadn’t repaired anything. They never simply refused to refund the money, but always said that it would be considered by some other group and that someone would get back to me. Of course, nobody ever did get back to me and LG never did issue any refund.

Full post, including comments