MO Electoral Cognitive Dissonance

Two stories in the news on September 18, 2024 in Missouri point out the continued cognitive dissonance of Republican voters in the state. One story involves recent polls which show Missourians leaning overwhelmingly to Republican candidates for President and US Senator, even when those Republican candidates are Donald Trump and Josh Hawley. The race for Governor is also dominated by the Republican candidate in polling. Yet the same polls are showing that statewide ballot initiatives to re-instate abortion rights and raise the minimum wage to $15/hour are likely to pass.

https://fox4kc.com/news/republican-candidates-lead-missouri-races-on-november-ballot-poll

Doesn't that sound rather, um…. Counter-productive?

If Missouri voters WANT things like codified protections around abortion and prenatal care and decent wages, wouldn't it make more sense to vote for more Democratic candidates in state and local races? Missouri is absolutely gerrymandered which accounts for the over-weighted tilt towards Republicans but as statewide elections for Governor, Attorney General and Secretary of State illustrate, the state is still heavily Republican. If the citizens feel the need to consistently revert to ballot initiatives to override the legislature and governor they keep electing, it would appear the minds of many in that electorate fail to grasp how the process is working and how it could work.

If Missouri voters needed any reminder of how capricious and plain mean Republicans in the state can be, news outlets also covered a story regarding the state's failure to pass a budget that provided legal required services for mentally disabled citizens under the Americans with Disabilities Act (ADA). The Department of Justice sent the state a formal notice on June 18, 2024 that it was out of compliance with federal law.

https://www.justice.gov/opa/pr/justice-department-finds-state-missouri-unnecessarily-institutionalizes-adults-mental-health

The government's position is that the state is reverting to merely institutionalizing mental disabled adults in need of specific types of care rather than funding community based programs that allow the patient to spend more time in non-institutional settings closer to home and family. Presumably, avoiding institutionalization is preferable for the patient and possibly preferable to the government by avoiding the higher cost of delivering such services in a more expensive institutional setting.

So how did the Missouri Republicans react? The outgoing Republican Governor included $79.4 million dollars in his formal budget proposal for 2025 but Republicans in the House and Senate stripped that language before passing the budget. Knowing the DOJ already formally notified the state of its ADA compliance failure, the governor warned his own party that dropping the funding was purely political slight of hand to allow people to claim to be frugal in spending the people's money when the funds WOULD have to be provided in a separate supplemental budget process for needs that arise after a formal budget is set.

Keep in mind, the state is actually running a small surplus with a budget around $50 billion dollars. The surplus doesn't stem from some huge windfall tax that is raking in billions in a bubble economy. Missouri's income tax rate tops out at 4.95 percent. Despite that absurdly low tax rate, the state is still running a surplus because Republicans are that tight on spending for everything from social services to infrastructure rehabilitation. The entire state is crumbling apart but hey, our taxes are low.

The required $79.4 million would be 0.16% of the total budget. As the governor himself noted, the money WILL eventually be budgeted and provided. But in the short term, it IS being withheld and services to approximately 1952 previously identified individuals. That is pushing care responsibilities onto family members that is disrupting their work schedules, causing needless stress and not providing the actual care required under federal law.

And Missouri Republican politicians don't care. They get to carry on the charade of being stingy with tax dollars and use that in their re-election bids. Apparently, none of them have family members or know anyone affected by this charade. Or maybe they do and they don't care? Or maybe campaign contributions from nursing home companies who profit from over-use of nursing homes are somehow worth more than individual donor dollars?

WTH

AI Power Consumption

The channel WelchLabs on YouTube has great content regarding the concepts behind AI. It posted a video on 9/13/2024 discussing how the latest releases of these LLM engines are continuing to demonstrate an outer bound of "efficiency" of AI models when their performance is graphed as a function of the compute power burned to train the model. In that video here

https://www.youtube.com/watch?v=5eqRuVp65eY&t=1112s

at 10:39 in the video, a graph is displayed showing the performance change from GPT-3 to GPT-4 . Next to the two graphs is a legend that reflects an astounding statistic. The GPT-3 model training burned 3,640 PetaFLOP-days of compute resource. The GPT-4 training took an astonishing 200,000 petaflop-days of compute power.

To put that in perspective...

A FLOP is a "FLOating Point operation", a unit of horsepower consumed by a typical operation performed within a CPU. It's not a calculation of a complex matrix dot-product of two arrays with 1024 dimensions but it's not a simple ADD operation to a local register either. A PetaFLOP is one thousand trillion floating point operations.

   1 petaflop = 1,000,000,000,000,000

An Intel i9 processor that might be found in a high-end desktop machine used for gaming has 24 cores, 8 running at up to 6 GHz and 16 running at up to 4 GHz. Intel rates that CPU at 1,228 gigaflops or

   1,228,000,000,000 operations/second

It would take 1,000,000 / 1,228 or 814 similar servers to provide 1 petaflop of computing power.

That means training of GPT-3 that took 3640 petaflop-days would require 3640 x 814 or 2,962,960 individual desktop computers to match that computing power. For GPT-4 that took 200,000 petaflop-days. That would require 200,000 x 814 or 162,800,000 desktop PC equivalents.

Obviously, this is a bit of an exaggeration. The computers used for this training were equipped with the latest GPU (Graphical Processing Unit) blades that are optimized for matrix mathematical operations. A top of the line consumer grade GPU made by NVIDIA branded the RTX 4090 can perform 3.6 teraflops so the training loads would be

   GPT-3 training = 3,640 petaflops = 3,640,000 teraflops / 3.6 = 1,011,111 RTX 4090 GPUs

   GPT-4 training = 200,000 petaflops = 200,000,000 teraflops / 3.6 = 55,555,555 RTX 4090 GPUs

NVIDIA's top data center oriented GPU labeled the A100 is rated at 312 teraflops so the same training loads would equate to these counts of A100 processors (which cost about $23,000 each):

   GPT-3 training = 3,640 petaflops = 3,640,000 teraflops / 312 = 11,667 A100 GPUs

   GPT-4 training = 200,000 petaflops = 200,000,000 teraflops / 312 = 641,025 A100 GPUs

Note that the original statistics were not in petaflops but petaflop-days. A petaflop-day is the output of compute providing one petaflop of processing operating for an entire day. Here's the power consumption of each of those hardware examples above:

• Intel i9 14900K processor = 360 watts
• NVIDIA RTX 4090 consumer grade GPU card = 450 watts
• NVIDIA A100 data center grade GPU card = 400 watts

The kilowatt power consumption for each for a full day of operation would be:

• Intel i9 14900K processor = 0.360 watts x 24 = 8.64 kilowatt hours
• NVIDIA RTX 4090 consumer grade GPU card = 0.450 watts x 24 = 10.8 kilowatt hours
• NVIDIA A100 data center grade GPU card = 0.400 watts x 24 = 9.6 kilowatt hours

Mapping those power consumption levels to the processing required for GPT-4 training that totaled 200,000 petaflops is jaw-dropping:

• Chat GPT-4 training on desktop equivalents = 8.64 kW x 162,800,000 = 1,406,592,000 kilowatt hours
• Chat GPT-4 training on RTX 4090 GPUs = 10.8 kW x 55,555,555 = 599,999,994 kilowatt hours
• Chat GPT-4 training on A100 GPUs = 9.6 kW x 641,025 = 6,153,840 kilowatt hours

For comparison, the largest single power plant complex in the United States is Grand Coulee Dam which is rated at a capacity of 7,097 megawatts or 7,097,000 kilowatts. If I haven't scrambled a units conversion somewhere, the power consumed by the GPT-4 training could have consumed all of the power from Grand Coulee Dam for 52 minutes.

And that's just the training compute. I haven't seen a good summary anywhere that explains how the final output of a training model is then scaled out for interactive use. That processing is equally dependent on extremely large matrix mathematics operations so I would presume the computing investment to run the final model is equivalent to that required to train it, especially as millions start using it on a daily basis.

It seems obvious at this point that an accurate cost / benefit analysis of AI has not been attempted. The assumption seems to have been that this compute is just laying around doing SOMETHING, let's have it do THIS and see if anything interesting results. It's only when you see announcements of one hundred billion dollars in new equipment being planned that it becomes apparent there are environmental impacts to AI that need to be considered as public policy, not solely as private investment decisions that treat power and water as freely available resources.

WTH

Anything You Say Can and Will...

Since the concept of a "digital assistant," an app listening continually for pre-configured trigger words to initiate helpful actions, was first launched with Siri in 2010 then spread to other smartphone devices and televisions, even unsophisticated consumers seemed to recognize an inherent privacy danger with this technology:

If you're always listening to EVERYTHING to match on the trigger word, what are you DOING with the information gleaned from what I said BEFORE saying the trigger word?

That privacy issue is coming up again along with even larger economic and legal issues as capabilities first referenced by Cox Media in December 2023 are getting new attention in the media in September of 2024. In December of 2023, information Cox Media provided to investors referenced a capability for advertisers and commerce firms to leverage information about "pre-trigger speech" captured from digital assistants to more explicitly target ads and -- AND -- influence shopping interactions after a user follows an ad to a commerce site to buy. The message from Cox Media was not merely referencing a hypothetical, it touted it in the present tense, essentially saying we can do this for you now.

This capability and concerns about its potential for abuse have likely come back in September 2024 after multiple companies have testified that their commerce sites have leveraged data mined from customer behavior reflected in cookies and other data sources to tailor shopping recommendations. That's a nice way of saying the sites are leveraging per-customer data to filter the products, providers AND PRICES listed to a user shopping on the site. And that's a nice way of saying that some of these companies have been implementing per-customer price discrimination in their systems, something that is illegal and a tremendous abuse of market power.

How did we get here? And where is this headed?

Evolution of Voice Recognition

In the 1990s and 2000s, the relatively limited CPU processing capabilities, available memory and audio quality of speech resulted in recognition algorithms that were "speaker-dependent." To recognize user X's voice saying a limited number of utterances, the system had to record user X saying each utterance multiple times (often 10-20 times) to average those samples together mathematically to use in recognition. After that training, if user Y said those same words, the recognition would not be as good as for user X who trained the system.

When smartphones introduced digital assistants, the recognition could be performed ON the device using audio signals captured locally in much higher fidelity than a classic telephone signal limited to 4kHz bandwidth to fit into a 64,000 bits/sec digital stream in phone networks. That boost in processor power aboard the phone also allowed use of "speaker-independent" speech models captured from (presumably) thousands or millions of human voices, making the recognition quality much higher. But while those initial speaker-independent models may have involved thousands or millions of voices, their "dictionary" of trained words was still relatively small. And therein lies the ethical and legal root of this story.

Illicit Data Collection

As already stated, at the introduction of these digital assistants in 2010, even non-technical users immediately raised the concern:

If you're always listening to EVERYTHING to match on the trigger word, what are you DOING with the information gleaned from what I said BEFORE saying the trigger word?

From the start, device manufacturers essentially waved their hands and said the device does nothing with "pre-trigger" utterances. Any digital data reflecting parsed speech PRIOR to a trigger word and subsequent "request" is simply sent to the bitbucket, never to be seen again.

That was the claim. However, anyone with a background in software engineering suspected this would not remain the case for very long, if it was ever true to begin with. Why? Because all of that digitized speech uttered PRIOR to a trigger phrase is a gold mine of data to use in expanding the "dictionary" of speaker-independent training models that would boost accuracy rates of utterances AFTER the trigger word.

All the device and application makers had to do was enable collection of ALL of the recognition data, both BEFORE and AFTER the trigger word. And clearly they did. It isn't clear if each device and application maker actually reflected that clearly and accurately in their terms of service agreement with the customer. If not, this business decision alone could result in multi-billion dollar class action lawsuits over privacy violations. These lawsuits would be perfectly justified.

But the thought of "just collect the pre-trigger utterance data as well" clearly triggered another thought. Since we have that pre-utterance data parsed and mapped to the originating speaker by their device and IP address, what if this utterance data was combined with THAT data for ad targeting? Keep in mind Cox Media is a subsidiary of Cox Communications. Cox Communications sells cell phone service and high-speed internet service. They have viewership data from set-top boxes, viewership data from any streaming video app they provide to their customers and some data from their customer's internet usage. Cox Media sells on-screen and online advertising to businesses on its parent's traditional cable TV network and acts as a broker for buying online ad impressions. This same business model is used by other communication carriers, whether telco or cable.

In the case of Cox Media, their December 2023 presentation to investors included this smoking gun:

Don't just know what they're searching for, know what they're talking about.

That's CLEARLY a reference to sharing "pre-trigger" data with ADVERTISERS. And that is not likely an action that Cox Communications divulged in its terms of service agreement with subscribers of its cell phone service.

The core point here is that data collection methods for improving voice recognition capabilities clearly strayed beyond the consent collected from paying customers. At least the universe of humans who turned on the assistant feature.. Or…? Were device makers sampling speech regardless of whether the customer enabled the assistant function or not? These capabilities are hard to turn off and, in some cases, impossible to un-install even if "disabled." So are they really off? That's a different technical topic that won't be addressed here.

Illicit Price Discrimination

All of those data collection capabilities on the part of device makers and service providers interacted with another capability that many online commerce companies have already been implementing for years which is just coming to light: the use of per-user profile data to drive shopping experiences. Again, that's a nice way of putting it. This practice essentially amounts to highly customizable price discrimination. Instead of a monopoly setting one price in the market that matches the producers marginal cost but reduces overall supply and creating inefficiency that way, this approach lets the seller set a unique price per customer, reflecting everything the seller can derive about the customer's individual "demand curve" for the product involved. How does this work? In multiple ways, all of which would likely be undetectable to you as a consumer.

Scenario #1 - Big Ticket Purchase Imagine you've been considering purchasing a new car for the last year. You've considered sedans, you've considered trucks, you've considered SUVs. Over the past year, you've been thinking more and more about one vehicle, a Toyota 4Runner. That term has come up in conversations with your spouse, it's been appearing in your Google searches. You've watched a couple dozen videos on YouTube about 2024 versus 2025 4Runners. You've been surfing to Toyota's web site at an increasing frequency as the 2025 model year nears.

Now, in light of all this data collection going on, ponder all of the personal data tied to your smart phone, tablet, smart TV and PC browser related to your actions over the past year.

• DNS data shows repeated surfing to www.toyota.com
• YouTube viewership data shows videos correlated to 4Runner
• pre-trigger utterances show an increase in the appearance of "4Runner" in speech

Now the new 4Runner comes out and Toyota sends you an email with a link announcing new inventory will be arriving shortly, click here to see the brochure or schedule a test drive. If you click on that test drive link, all of that prior data collection can be linked with your click that registered you for a test drive and chat with a salesman.

Now imagine showing up to that dealership. That salesman could have your surf and search history summarized and know you aren't just a casual shopper, you probably already have your mind made up. How good a deal are you going to get when the seller knows you are 50% less likely to buy any other vehicle?

Scenario #2 - Routine Purchases -- Imagine you are a regular Amazon shopper and user of Alexa on various devices. Imagine you've been having conversations about buying a new big-screen TV. Imagine Alexa has been capturing your pre-trigger utterances for two years, analyzing them against your Amazon shopping behavior and has determined that you tend to be rather, ahem, "decisive" about purchases. Once you start talking about buying something, you generally buy within a day or so, rather than debating features and prices for two weeks.

That behavioral pattern could be used by Amazon to "customize" your shopping experience to ensure the TV models most profitable for Amazon appear on the first page, despite being significantly more expensive than competing products. Amazon knows you won't scroll beyond four pages of results and adjusts the appearance of products accordingly. In contrast, a different user found to be more cautious would still see a customized shopping view but even more price-sensitive options would still be prioritized for Amazon's financial benefit.

These are not hypotheticals Just today, I received an email from an online musical instrument retailer at 12:54pm touting new Gibson Les Paul Studio models. I opened that email around 2:20pm, clicked on the link to bounce to their web site and looked at the detail page of one specific color. At 2:27pm, I got another email from the vendor with the subject of "Take another look?" with content that showed the exact color I viewed online. That's how fast data collected by cookies can be turned into new outbound communication and fed back into online portals for customized experiences.

Where are These Capabilities Headed?

The first answer to that question should be federal criminal court. If the summary above is remotely close to what these companies are actually doing, the service providers involved would be subject to massive lawsuits for violating terms and conditions and customer privacy laws. The commerce customization capabilities implemented by online merchants would amount to illegal price fixing and discrimination. If taken to extreme, they could be used to silently, invisibly red-line sales to specific areas or members of specific ethnicities.

For example, if a company doesn't want to sell to a specific group, their "commerce customization" could include logic to fake inventory levels of zero when the shopper maps to designated metadata. Unless that shopper suspected something odd and asked a different person in a different home to surf to the same site at the same time for the same item and find available stock, they might never know they were being discriminated against.

The fact that there are three smartphone makers, two dominant PC operating system makers and only five or six dominant service providers in the entire United States makes these data mining economies of scale nearly irresistible to those in control of the data. The real question is whether the value provided to consumers -- in this case seemingly driven by the "convenience" of a digital assistant -- outweighs the economic harm resulting from abuse of these monopolistic powers.

As stated multiple times in the recent past, American consumers have lapsed into thinking that our cool toys will someone stop working or stop getting more cool if ANY limits are imposed on the giant corporations so generously bestowing upon us the fruits of their genius. This is what many thought prior to the divestiture of AT&T in 1984.

Forty years later, innovation and economic efficiency are again getting strangled by abusive monopolies. Sure, long distance calls are free instead of eighteen cents per minute but now the phone costs $1300, you're convinced you need a new phone every three years, cell service is $19.99/month AND your modern day "Daughter Bell" is listening to everything you say and selling your profile to every merchant you deal with.

WTH

Unique Challenges in a Gun-Crazed America

It's been seven-plus hours since Secret Service agents fired at a man and later captured him after he was spotted with an automatic weapon outside the grounds at Trump's golf course at Mar-a-lago within one hole of where Trump was playing. More details will emerge in the coming days but it's not too early to begin anticipating how this event will be turned into another front in the 2024 election battle and the larger political war and Rorschach test that American civics have become.

What is known is that the man involved in the incident had not yet actually FIRED his weapon but was clearly staking out a position that would be well within range of Trump on a normal round of golf. And it's worth re-mentioning the man was NOT actually ON the grounds of the course. He was outside a security fence in shrubbery able to point through a fence at the holes in front of him. The man fled the scene, left his weapon behind and was later captured and identified as a 58 year-old named Ryan Wesley Routh who appears to be originally from Hawaii and spent some time in North Carolina. Different outlets are pointing out he contributed a total of $140 dollars to various Democratic candidates since 2019 but also tweeted comments to Republican candidates encouraging them to stay in the race against Trump. He has eight arrests in North Carolina going back to roughly 2000, mostly for "minor offenses." (It isn't clear to me how minor offenses can consistently result in arrests.) He also claims to have been active in recruiting people to fight in Ukraine against Russia and actually fighting in Ukraine, though there are no details on what that "service" was. Once source indicates he claimed to lead an organization called the International Volunteer Center (IVC) which stated it organized efforts to delivery humanitarian aid in Ukraine.

The golf course at Mar-a-lago is located "five minutes" from the actual Mar-a-lago club and contains 27 holes. The fact that it is "five minutes" from the main club means any time Trump is on the course, it is a completely separate security zone that must be somehow guarded by the Secret Service, while still protecting any other Trump family members still remaining at the club itself.

Given that Trump golfs frequently and a zone should be ideally secured hours before he enters the zone, frequent golf outings essentially double the Secret Service work load. If the Secret Service was going to guard the ENTIRE zone, they would have to secure the ENTIRE perimeter which would be astronomically expensive and impractical... So they don't. Instead, the normal modus operandi seems to involve securing a smaller moving "bubble" that might be plus or minus one hole around the current hole he is playing, plus presumably any other holes that might be adjacent to the current hole.

In this golf course event, the would-be shooter was exactly that. He didn't even fire a shot. He was identified and shot AT and fled the scene without directing fire at Trump and was later apprehended. Given the nature of the threats facing any protectee, that's exactly what should happen.

That's a very difficult logistical task. But remember... Trump owns several golf courses and routinely plays at all of them, often in the same week. He held an outdoor press conference at his golf club in Rancho Palos Verdes near Los Angeles on Saturday, September 14 that required agent coverage to secure. That appeared to take place just outside the clubhouse next to a practice green.

Prior to the Trump era, the Secret Service was never funded and staffed to cover the logistics and costs of protecting a President or ex-President with four or five principal residences in five different states, each with dozens of acres of ground theoretically requiring full perimeter protection. Nixon had San Clemente. Carter had his house in Plains. Reagan had Rancho de Cielo. Bush had Kennebunkport. Clinton had... (I can't even find a reference to Clinton's non-WH residence during his term). Dubya had his ranch in Texas. Obama had a home in Chicago.

Think about it this way. If each Trump residence was going to be given the traditional perimeter monitoring and protection afforded to the properties owned by any prior President or ex-President, Trump's lifestyle essentially raised the budget requirement by a factor of 5x. Permanently. Beginning in 2017. Has the Secret Service budget increased by 5x since 2017? None of us will ever know but I'm pretty sure the answer is no.

Given the nature of a golf course amid a metropolitan area, here are a few questions worth asking.

1. Is there no way to chain together a series of holes on a 27-hole course to yield an 18-hole game that AVOIDS play on holes on the perimeter of the course abutting external public areas?
2. Was Trump advised of this obvious inherent danger by Secret Service staff?
3. Did he ignore a recommendation NOT to play outdoors at the course because of these risks?
4. Will the FBI find any "smoking gun" communications in the man's online communications, texts to acquaintances, etc. reflecting a clear threat that went unprocessed?

Remember, there seem to be very few things the Secret Service can dictate to a protectee. Whatever a protectee wants to do and however they want to do it, the Secret Service is obligated to protect as best they can.

Let's stick with the concept that this isn't anyone's fault. It's just an event that can happen when mental health isn't appropriately funded, the public has easy access to a wide variety of weapons and public figures routinely, predictably appear in outdoor settings which are impossible to protect from long-range, high powered guns. If that's the only takeaway allowed, then I know what I'd be doing if I was President or running for President. I'd be listening very closely to my Secret Service detail and following their recommendations to the letter.

I don't want ANY elected official or person running for elected office subjected to these threats. I also don't want any other public figure or private citizen exposed to the same threat of random violence at a distance brought on by a mentally ill person or completely rational criminal. I also don't want anyone in a position of power or influence to take actions that unnecessarily raise the risk of such events and magnify the fear already being focused on people of any ideological persuasion. This madness needs to stop.

WTH

Georgia Judge Drops Two Charges

The judge hearing Trump's state criminal case in Georgia issued a ruling in the case dropping two charges against Trump and a different charge against others in the case.

https://www.washingtonpost.com/national-security/2024/09/12/trump-georgia-case-charges-dismissed/

The charges involved false documents -- the actual filing, the preparation and conspiracy to file. As the Post story stated,

McAfee ruled that because filing false documents in federal court is subject to sanctions under federal law, state law cannot be used as it was in one of the counts he dismissed Thursday.

The idea that an action that is a crime under both federal law and state law can ONLY be prosecuted at the federal level is NOT the intent of the supremacy clause in the US Constitution. The purpose of the supremacy clause is to dictate that when any state law conflicts with a federal law, the terms of the federal law supersede those of any state law. The supremacy clause was not intended to settle issues of jurisdiction over prosecution of an action in violation of both state and federal law. The documents involved in the charges dismissed by McAfee could be viewed as "dual scope" documents. When you prepare certification of a slate of electors for a particular state, you are first preparing documents clearly subject to STATE law since a state's Legislature must somehow sign off on those documents. Once the legislative approval is provided, in essence that same document now gets submitted to the Electoral College under the jurisdiction of federal law.

McAfee's decision is being interpreted by some as saying Georgia's existing state law that references filing false documents to either state or federal authorities is unconstitutional because of his interpretation of the supremacy clause.

Trump's lawyers cited an 1890 case that reached the Supreme Court in which the court ruled that a state could not prosecute a perjury charge stemming from testimony in a disputed federal election. The problem with that example is the perjury involved occurred in a federal trial related to that contested election. So the alleged crime (perjury) occurred in a federal court room so it could be argued it was a federal responsibility to protect the integrity of its proceedings and that was not within the purview of a state to file that charge. In this Georgia case, before there could be a federal crime of submitting false documents to the Electoral College, there were crimes that occurred within the State of Georgia that conspired to falsify information and falsely submit that information to state officials BEFORE submitting to a federal process. It is totally within the State of Georgia's authority to protect the integrity of its state processes by prosecuting those acts, even if they served as inputs to federal crimes as well.

It seems McAfee might be as ignorant of basic law as Aileen Cannon. It isn't clear to me if his decision to dismiss charges can be appealed, but I am assuming it cannot. At the same time, the judge separately affirmed the felony racketeering charge against Trump and team, describing it as "facially sound and constitutionally sufficient."

WTH

Solving the Housing Crisis

Can first-time buyer tax incentives solve American's housing crisis? Let's cut to the chase.

No.

A tax credit cannot cure the problems being experienced in the US housing market. A tax credit provided to buyers alters the financial calculation on the demand side of the equation. The affordability problem being targeted is actually due to the supply of housing. This is much easier to see by creating a relatively simple graph of some of the key variables driving supply and demand.

Visualizing the Problem

The housing problem in the US stems from the simplest factors in any economic analysis -- supply and demand. The supply of housing is not keeping up with the demand by new households. When supply Is behind demand, basic economic theory indicates prices go up. But what is the magnitude of this imbalance? Plotting population growth and housing starts over a period of time is an intuitive way to show the relationship between the two.

If supply was responding adequately to changes in population, a graph of housing starts should have a positive 1:1 correlation with population. But does it? In the real world, the correlation won't be perfect because actual housing purchases are influenced by business cycle issues. However, one would still expect the overall correlation to be fairly direct over a longer period of time. But is it?

Using federal government statistics on population and housing starts from 1990 to 2023, the following graph was produced to reflect the variation in housing starts over that period.

The purple line shows how US population has grown from 248 million in 1990 to 341 million in 2023 against the right y-axis scale.

The black line reflects actual yearly housing starts in 1,000s units.

The gray line depicts a smoothed estimate of housing starts required to keep up with actual population. This "average" was generated by first averaging the ratio of starts/population from 1990 to 1999, a period that began in recession and ended with the first Internet bubble but in general reflected some of the cyclical nature of the housing market in a 1990s context. That average value for starts/population was then applied to the actual population figure for each year from 1990 through 2023 to graph that line.

The yellow line is each year's shortfall between the "smoothed" starts number that "should" have been built based on historical averages and actual starts.

The red line is the running accumulation of all prior year shortfalls (excess) of actual starts compared to the smoothed benchmark. Because it is labeled a "shortfall", it graphs as a negative number when there is an actual excess.

So what are the key points conveyed by this graph?

1. While US population is still growing, the RATE of growth is slowing steadily year after year.
2. Housing starts peaked in 2005 at 2,068,000. While everyone thinks of the financial collapse occurring in 2008, this reflects that actual housing starts began falling precipitously in 2006 and that drop actually triggered recognition that mortgage markets would collapse.
3. Beginning in 2007, actual housing starts fell BELOW the smoothed gray trend line and have never surpassed it since. If you accept the intent of that gray smoothed line, this means actual new housing starts have never caught up to erase the prior deficits.
4. The red line shows that all of the housing inventory of homes started during the housing bubble between 2002 and 2009 was absorbed by demand and since 2009, the deficit of new housing units has grown every year.

The real takeaway from the graph is that based on the assumptions explained above, the actual magnitude of the housing gap in the US right now is 7,759,000 units. That can trigger thoughts in a few different directions. First, in the entire series of data, the biggest "overage" between actual starts and "needed" starts was about 500,000 units in 2005 that occurred at the peak of the housing bubble. If 500,000 is some sort of upper bound on the ability to create housing in excess of normal market demand, that means eliminating that cumulative gap of 7,759,000 units would take nearly sixteen years to catch up.

Second, a legislative plan that aims at providing a $25,000 in down-payment assistance and a $10,000 tax credit for first-time homebuyers may stimulate demand for HOUSING but may do NOTHING to provide incentives for builders to build NEW housing in lower price ranges. Think about it. The benefit is provided to a person who buys a home for the first time. The requirement is NOT constrained to the home being new which means the incentive doesn't drive behavior regarding actual construction that addresses the net shortage. Such incentives could simply encourage a shuffle between renters and owners, drive up prices in the short term and leave the total supply of housing unchanged.

Thirdly, and possibly most imporantly, any incentive targeting first-time buyers is altering incentives for buyers who are already on the fringe of affordability. That fringe cuts both ways. They might think they only need one or two more years of scrimping and saving to afford a home and now a tax credit would bypass that wait. They might also NOT be one or two years away. The credit might tempt them into buying something they still cannot afford through an unexpected repair, a temporary loss of income, etc. and they may wind up in a cash flow situation that forces them to sell to escape the financial commmitment. They could take a significant percentage loss on the house and pay a significant commission to an agent to assist with unloading the house. Buying a house at this economic stage of life is one of the riskiest decisions one can make.

This graph still doesn't convey all of the complexities of the housing market and all of the forces at work in placing homes out of reach of average citizens. Consider these factors:

Household Size -- Technically, housing demand isn't driven directly by population. It's driven by the number of "households" which is a function of total population divided by average household size. Household size is a function of individual decisions to marry or cohabitate and to have children. For a given fixed total population, halving average household size from 4 to 2 would logically double demand for housing units. The graph above shows that overall population growth is slowing which would imply needing lower growth in housing units with all other factors held constant. But at the same time birth rates are falling, the share of adults living alone has been growing, which increases demand for housing, possibly negating the impact of slower total population growth.

Household Income -- Splitting up the cost of a mortgage across two incomes is a common benefit sought by anyone who marries or cohabitates. Obviously, if fewer adults are getting married or cohabitating and instead are choosing to live alone, they are presumably facing a mortgage payment alone as well, on a single income. By definition, that will remove a large swath of buyers from the market.

Construction Labor -- The plot of population in the graph above has implications for the market for construction labor required to build new housing. In the US, yearly births peaked in 2007 at 4.32 million and are around 3.67 million as of 2022. The "Alpha" generation born between 2013 and 2027, that bloc of workers is estimated to be about 56.48 million. When they begin reaching working age in 2031, that generation will be 13.1 million smaller than the generation preceding it, which was already smaller than its predecessor. And trades are not attracting new workers. (Edtiorial anecdote: A family member recently remodeled a 1971 home and the firm that did the work was a brother duo who were both nearing 70 years old.) Twenty-somethings may be unhappy with the prospect of sitting behind a keyboard on Zoom meetings all day but relatively few are volunteering to learn how to frame a house or install and repair plumbing.

Zoning Constraints -- Local zoning restrictions are based on assumptions and prejudices (social and technical) dating from the 1950s. From 2024 looking back, this collection of zoning patterns could be termed America's second original sin -- the decision to subsidize standalone home construction via highway construction, tax deductions for mortgage interest, building height limits, street setback rules and restrictions on MDUs (multi-dwelling units). As dozens of YouTube content creators have documented, the result is a world in which modern, efficient housing solutions that could meet demand for single-tenant units and offer walkable streets and mass-transit friendly access to office centers for work are actually ILLEGAL in the vast majority of the US. This environment reflects a complete failure of imagination on the part of politicians, planners and the public. Americans cannot fathom the concept of not owning a car, of taking a ten minute bus ride to work, of having a corner bar three blocks away they can walk to. Call it Synchronicity Syndrome.

Another working day has ended
Only the rush hour hell to face
Packed like lemmings into
Shiny metal boxes
Contestants in a suicidal race

Bigger Homes and Bigger Margins -- If you are a contracting firm facing these market conditions of limited supply of locations zoned for building, limited labor supply and rising prices, what is the best way to maximize your profit? Building a larger home for the few who can actually afford a home. What does that do to new home prices? Increases them. What does it do to "comparable" valuations for nearby existing homes? Increases them as well. What does it do to average prices overall? Pulls them up? What happens to entry-level customers? They get priced out of not just the new home market but more existing homes as well. This dynamic should be very familiar to Americans. This is exactly what happened with vehicle prices during the pandemic, especially large trucks and battery powered vehicles.

What COULD Work?

The overall thesis here has been that attempts to manipulate demand for housing will do little to solve the core problem of supply. Solutions must focus on the supply side of the equation; the supply of available lots, the density of construction allowed on those lots and the supply of labor required for construction. Here are a few policies likely to have a direct impact on supply.

Tax Credits for MDU Complexes Between 4-16 Units -- To encourage higher density without triggering fears among existing citizens of the next Cabrini-Green or Pruitt-Igoe disaster, federal subsidies could be provided to cities for completed construction of new MDU complexes (apartments or condominiums) sized between 4 and 16 units. Why that range? The lower value ensures a significant increase in average density over 1.0 needed to make headway while utilizing space more efficiently. The upper limit ensures developers don't build one, large, cheap hell-hole tower to claim the credit and build an unlivable space that antagonizes local citizens. Paying the subsidy to the city ensures the city alters codes and actually permits the construction to produce actual new units.

Tax Credits for Office Rehabilitation -- Work-from-home trends have easily reduced demand for office space by 20 to 30 percent nationwide. Much of that empty space is "trapped" because buildings are seldom emptied completely. A firm may still retain 100% of its lease but only have 60% of workers present on any given day or some tenants may leave with others staying. Office space constructed in the last twenty to forty years was heavily optimized for office use and often lacks elements that would simplify re-purposing the space as individual housing units. If WFH continues indefinitely, this stranded space represents hundreds of billions of dollars in trapped real-estate that will crater in value and cripple municipal tax revenues unless a solution can be found. Offering tax incentives for firms that can devise economically sound ways to re-purpose these buildings is very much in the nation's interest.

Tuition / Training Reimbursement for Trades -- The US is already operating with 4 percent unemployment. Any able-bodied person with trade skills interested in working is already employed and unable to take on more work. Individuals wanting to get trade work completed for a new home or remodeling project routinely see three to six-month delays in scheduling. Actual construction intervals are now seven to twelve months. Twenty to thirty years ago, construction might have taken six months from contract to close. Slips in schedules due to labor shortages for a sub-contractor have exponential impacts on the overall schedule so any programs that increase the supply of trade workers can reduce construction times and increase units delivered.

Immigration Reform -- The US already offers special work visas for in-demand high tech workers. If businesses, communities, educators and politicians are unable to devise incentives to encourage young people to enter and stay in the trades, the most immediate alternative to increase the labor supply is to further leverage immigration. I say "further leverage" because America is already highly dependent on this labor pool. Even in the landlocked Midwest, I have not seen a roofing crew or a fiber optic cable construction crew that wasn't heavily dominated with immigrant workers.

Innovation in Construction Techniques and Materials -- In a world where homeowners and general contractors use CAD tools to design decks and kitchen remodeling projects, it seems odd that "stick frame" construction is still used in most US new homes, especially when most new homes are NOT custom designs but instead picked from a set of maybe eight available plans per builder who may build twenty nearly identical homes in the same subdivision. An easy way to improve quality, improve efficiency of materials usage and reduce labor content involves modularizing construction. Builders already do this for roof trusses, why not walls? If Americans don't want to work outdoors AND think trade work is beneath them AND oppose immigration to let someone else do the labor, the only alternative left is to eliminate more of the labor. Should the federal government create tax incentives for firms that manufacture equipment for CAM (Computer Aided Manufacturing) of modular homes or use CAM to build modules for retail builders? The tools themselves already exist. The problem is more related to providing enough incentives to get builders to adopt the methodology in their designs.

WTH

Lessons from the Harris / Trump Debate

In a sane, rational world, the only viable conclusion about the Harris / Trump debate on September 10 was that Harris wiped the floor with Trump. However, we are not dealing with a totally rational electorate so that simple answer is mostly immaterial. The debate was like a tennis match in which the victor won in straight sets. That's an easy way to summarize the match on TV but voters should be more concerned about all of the individual points played because they can be HIGHLY instructive about the strength of each player's complete game. And the Presidency is not a game in which we want a player with multiple fatal weaknesses in their game.

In no particular order...

Trump and Math -- Trump not only doesn't understand math, he doesn't understand NUMBERS. For four years, Trump has backed his claim that he won the 2020 election because in that election, he won more votes than any prior Republican candidate. Ummm, Donald… America is a big country. America is a big, growing country. Every year, the population gets larger. If you won 49 percent of the vote of a growing voter population, your VOTE COUNT will go up. But you don't WIN by winning more votes than you did four years ago or more than any other Republican. You have to win more votes than your OPPONENT. This is the logic that comes out of the mouth of an innumerate narcissist who cannot accept losing. Sadly, it is logic that seems to work with his equally innumerate MAGA base.

Trump and Ukraine -- Trump refused to answer the simple yes/no question from David Muir. Do you want Ukraine to win the war or not? He argued he simply wanted the war to end as soon as possible and that he would end the war in a matter of days. He then pivoted to blaming Europe for tempting Putin in the first place via his normal "NATO dues" rant. Of course, Trump is overlooking the fact that Europe has in fact contributed MIGHTILY to the war effort in Ukraine, most notably by accepting something like 3 million refugees from Ukraine who fled the country and required housing, schools and extra resources in hospitals, etc. all over Europe. Germany alone has taken in 1.2 million refugees. God knows a Trump Administration would never welcome Ukrainian refugees to the US.

Exiting Afghanistan -- Trump made completely illogical points about the causes and effects of the US military exit from Afghanistan. For the most part, his own incoherence was sufficient to lose this point on his own but Harris failed to hammer home the point as effectively as possible and deserved. First, Trump negotiated the Status of Forces agreement and final exit date that drove American withdrawal and his own military advisors argued at the time it would be fraught with risk for American interests and stability in Afghanistan. Biden shifted the actual data around by a month but the core result was dictated not even by the fact that Trump set the date. The core result was dictated by the fact the US lost the war in Afghanistan. We altered our goal at least four times over twenty years and failed to achieve a SINGLE goal. When you lose a war abroad after twenty years, you aren't going to spend the money and inflate your risk to troops by trying to "evacuate" ten to twenty year old used-up gear. You destroy it and get your people out as quickly as possible.

Combating Iranian Influence -- Trump argued that he re-instated sanctions against Iran but he also terminated the 2015 agreement with Iran that had suspended development of uranium refinement capabilities. He gave Iranian hard-liners a rationale for further hyping anti-American sentiment rather than dealing with its own vast economic problems. Trump's exit from the deal wasn't driven by any certifiable expert proving that Iran was not complying with the deal. He terminated the deal because it was made by Obama and he didn't want any deal of a predecessor standing when he was supposed to be the great deal maker.

Trump Was Right on One Thing -- In the midst of a much larger topic Trump is wrong about, Trump did make one correct point. In the segment addressing abortion, Trump stated it won't be up to Harris to vote for a law re-instating a national right to abortion across all states. He is correct because such a law cannot be passed until Democrats win 60% of the Senate to thwart filibusters which will normally kill this bill. But even that's not enough. If passed by the House and Senate and signed by Harris, the law would be immediately appealed to the USSC which has already shown they would kill it as unconstitutional. That means no national law re-instating abortion rights will remain law until the USSC balance is altered through retirements or packing. Harris should have used this point of Trump to turn to the camera and say "This is why this election is about more than me. Every American who cares about this issue needs to vote for Democrats in the House and Senate to begin building that majority needed for legislation and altering the USSC.

Trump Needs Schoolhouse Rock -- At one point, Trump said that Harris should just walk out of the debate, return to Washington DC and the White House, wake Joe Biden up and sign a bill to fix the immigration problem. Just sign a bill. Make it happen. It isn't clear Trump understands how the law-making process functions. The changes he would impose have already been attempted by both his first administration and the Biden administration as executive orders and they have been rejected by courts. The fact that Trump doesn't understand how a bill becomes a law is somewhat incredulous given his appetite for television. I would have thought through sheer probability he would have stumbled across I'm Just a Bill by Schoolhouse Rock in reruns amidst Saturday morning cartoons.

Manipulating a Narcissist -- Harris' performance in the debate was a masterclass in the manipulation of psychologically flawed personalities. Her preparation clearly cemented a model in her head for optimizing each turn at bat. How much time to I spend refuting one of his bullets? How much time do I devote to MY points? Based on the conversation so far, what is my next trigger word to dangle in front of him to distract him and further take him off message? The exact mix varied but after five minutes, you could practically watch Harris do the calculus in her head. Each time, she could end with a magic trigger word such as McCain and he would immediately latch onto that like a shark and not let go. This reminded the public exactly how petty Trump has always been but it also demonstrated how simply he can be manipulated, either by strokes to his ego or references to his enemies, a list which grows by the day. He is unfit for haggling with a souvenir vendor in Tijuana, much less negotiating in a room with a foreign adversary with national security at stake.

Object Permanence -- Trump routine states contradictory views to different audiences as if those audiences will NEVER talk to one another or hear what happened in the other event. In his answer on abortion, he attempted to argue two different points for two different audiences, as if each was deaf and blind to the "other" message. They're all watching you live on TV. You can't argue that you AREN'T in favor of a national ban because it isn't necessary cuz the issue has been pushed to the states when the audience in favor of a national ban is working wherever possible to push exactly that national ban. More importantly, you can't argue overturning Roe was a win for people who WANT abortion rights because it pushes the issue back to state control when the Supreme Court hinted in its Dobbs decision that additional reproductive rights are in jeopardy if the right case comes along (hint…hint…hint…). Trump cannot comprehend that making these two arguments SIMULTANEOUSLY in front of a non-segregated office wins him NO votes, only distrust. Even most pets master the concept of object permanence and that an object doesn't disappear when placed behind another object.

WTH

Missouri's Grand Slam of Election Interference

It's happened.

Missouri Republicans have hit a grand slam in terms of election interference by their efforts to block Missouri citizens from voting on a state constitutional amendment that would have re-instated abortion rights after the Republican controlled legislature and Governor enacted a strict abortion ban immediately after the Dobbs ruling of 2022.

Republican officials and partisans took all of the following measures to thwart efforts to place the initiative on the November 2024 ballot.

• Missouri's Republican Secretary of State Jay Ashcroft attempted to provide a grossly misleading summary of the initiative that would appear on material shown by canvassers attempting to collect signatures. This language was thrown out by a judge in September 2023.
• Ashcroft later attempted to also provide a grossly misleading summary of the initiative required by a "plain language law" in Missouri that would appear next to sample ballots at polling stations to "explain" the law's intent which AGAIN grossly misled voters about the law's intent. This language was thrown out by a judge in September 5, 2024.
• Anti-abortion partisans immediately appealed that September 5 ruling, claiming the proposed amendment covered too many topics, a legal rationale previously argued which was rejected, September 6. The anti-abortion plaintiff resurrected prior legal challenge to the amendment, arguing it addressed too many topics. This motion led a judge to reject the amendment (while staying his rejection based on a State Supreme Court decision slated for September 10) because the judge claimed canvassers did not properly explain that the amendment would completely un-do Missouri's abortion ban.
• On September 9, before the state Supreme Court had a chance to rule on the bogus ruling from September 6, Ashcroft stepped in again and rejected his prior certification of the initiative, citing the 9/6 ruling as support, despite defending his prior certification IN COURT on September 6 prior to the judge's ruling.

In a nutshell, these fringe conservative judges, politicians and partisan groups are arguing the following two things must happen SIMULTANEOUSLY:

• A constitutional amendment initiative must explicitly identify EVERY change it would make to existing law and to existing processes and systems so would-be voters are properly informed...
• ...YET the amendment initiative language can only address one topic.

In a situation where the LEGISLATURE has enacted a sweeping bill that outlaws abortion and imposed penalties in related areas of reproductive care and rights of citizens to information about reproductive care, this interpretation prevents a single amendment from un-doing that law.

If the initiative simply identifies the prior bill to be reversed, the public hasn't been told EXPLICITLY all the things it will undo, the public is not properly informed so the public doesn't get to vote on the issue.

If the initiative explicitly identifies all of the impacts, which extend beyond the direct right to abortion but related protections for patient confidentiality, retaining proper regulation of reproductive care providers, etc. so all impacts are reflected in ONE bill so the public knows EXACTLY what it is voting on, then that violates the one -amendment / one-topic rule.

Sadly, this isn't just fringe politics as usual. Jay Ashcroft, the man driving much of the interference in this initiative's fate, had been running for Governor to replace term-limited Mike Parsons. That would have explained his interference in 2023 as a means of bolstering his bid for Governor. However, Ashcroft lost his bid for Governor by a significant margin in the August 2024 primary. Ashcroft's term as Secretary of State ends in 2024 as well so his choice to run for Governor means he will not be Secretary of State past 2024. His continued interference in this issue is thus not helping him in any immediate political campaign with the Republican base in Missouri.

So what explains Ashcroft's conduct?

Ashcroft may still be looking ahead to some future run for office and may be assuming this will bolster his credentials with the conservative fringe but frankly, despite his "Missouri-famous" last name, he isn't much of a brand in the state. He might just be this adamant about abortion. Or maybe he is greasing the skids to land himself a cushy job at a high paying conservative think tank. However, regardless of his personal beliefs about abortion or attempts to snag a soft landing in the private sector, his conduct in this matter amounts to a gross violation of his oath of office to "faithfully execute" the law.

What's the impact on Missouri?

Unless the Missouri Supreme Court agrees to still hear the appeal filed on September 6, overturns the lower court decision AND rejects Ashcroft's decertification, Missouri voters will NOT have the opportunity to vote on un-doing the draconian ban enacted after Dobbs. Because of the issues raised in this debacle and the timing of the various efforts, any effort to get the measure back on a ballot faces numerous decisions:

• How should the single issue be worded and possibly subdivided into individual amendments to satisfy bogus concerns about single-amendment / single-topic compliance?
• How can any revised amendment be reviewed to ensure it WILL pass muster per the bogus interpretation being enforced at some layers of the Missouri court system?
• If multiple initiatives are required, then for the desired "un-do" effect to be provided, ALL of the initiatives have to pass the signature collection test and ALL have to be passed by the voters, otherwise some critical protection might still wind up missing because the voters didn't understand how they all tied together.

If the Missouri Supreme Court does NOT hear the appeal on September 10 and make a ruling, abortion rights advocates won't know if the strict one-amendment / one-topic interpretation will again crop up nor be given any hint how much separation will be required. This adds uncertainty to allow similar interference in the future, further delaying the point at which citizens will have the choice of overruling their Legislature and Governor and re-instating abortion rights in the state. This measure was targeted for the November 2024 ballot in a Presidential election year to ensure the largest pool of voters had a chance to weigh in on this issue. Interfering with that is the ultimate intent here.

And that's really the essence of this saga. This strategy is being adopted all over the country by fringe conservatives trying to protect laws enacted with much lower thresholds of support by tampering with any available processes available to The People to raise the threshold required to un-do those laws. Democracy at its saddest.

WTH

More Missouri Election Tampering

Efforts by Missouri citizens to place a constitutional amendment on the November 2024 ballot to un-do the near total abortion ban enacted immediately after the Dobbs ruling of 2022 have been thwarted again.

Earlier this week, on September 5, a Missouri judge had to intervene for a second time to correct grossly misleading "explainer" language written by Secretary of State Jay Ashcroft to mislead voters about the intent of the amendment by claiming it would allow abortions through birth and would leave all reproductive care providers unregulated and accountable for their actions.

That September 5 ruling AGAINST the anti-abortion partisans was immediately appealed and that appeal was heard by Judge Christopher Limbaugh on Friday, September 6. The anti-abortion plaintiff resurrected a prior legal challenge to the amendment, arguing it addressed too many topics.

The judge hearing the motion didn't toss the measure off the ballot because it violated the one topic / one amendment law. Instead, he went in a new direction, ruling that the organization that worked to get the initiative on the ballot failed to meet legal requirements because...

...wait for it...

...the organization did not adequately inform people being solicited for signatures that the amendment would un-do Missouri's abortion ban. Missouri's cutoff for finalizing the November ballot is September 10. Limbaugh stayed his ruling that would drop the initiative from the ballot until September 10 to give backers a chance to appeal to the Missouri Supreme Court.

Of course, this curveball literally DAYS before the cutoff means that ANY change ruled necessary after the state Supreme Court rules will not have time to be reflected in the final ballot so ANTI-abortion forces wanting to continue thwarting the will of a majority of citizens in the state will have succeeded in pushing off a correction for at least another year, maybe two, maybe four.

I was one of the 380,000 Missourians who signed the petition to get this on the ballot. I was approached in probably four different parking lots outside big box stores over multiple months by workers soliciting my signature. EVERY one of those solicitations started off with words to the effect of "Would you like to sign a petition to re-instate abortion rights in Missouri?" There was NO DOUBT about the intent of the campaign or the intent of the language of the ballot initiative.

This abortion rights effort in Missouri is perhaps the poster child of conservative extremist schizophrenia that has taken over much of the country. In this case, anti-abortion forces first attempted to tamper with consideration of this initiative by fraudulently wording a summary description that had virtually zero correlation with the actual language of the proposed amendment.

Anti-abortion partisans LOST that first round when a Missouri judge tossed out flawed language written by the Republican Secretary of State Jay Ashcroft in September of 2023.

The petition was circulated statewide and collected 380,000 signatures. Missouri's population is about 6.2 milliion, 75% of which are of voting age so the petition gained support from 8.1% of adults. Anti-abortion partisans trying to preserve Missouri's draconian anti-abortion law enacted in 2022 filed suit claiming the proposed language covered more than one topic, something prohibited under Missouri statute for constitutional ballot initiatives. (Missouri legislators fear Missouri voters are too dumb to read multiple clauses in a single sentence. Missouri voters fear that Missouri legislators are too dumb to read multiple clauses in a sentence.)

The too-many-topics argument was dropped at some point only to have Secretary of State Jay Ashcroft AGAIN tamper with the process by AGAIN writing a completely false "plain language summary" required by state law for for display in sample ballots at polling locations. Like the first time, Ashcroft's intent was to scare voters into thinking it permitted abortions up to birth (it does not) or completely removed all regulation from reproductive care providers (it does not). That triggered a suit from abortion rights backers which was heard by a judge who tossed out Ashcroft's language entirely and specified alternate language to use for the summary.

The rejection of Ashcroft's SECOND attempt to mislead voters triggered the counter-motion here that led ANOTHER judge to latch back onto the theory that the amendment addresses too many topics and those poor ignert Missourians couldn't have possibly understood the impact of the petition they were signing in the parking lot at Wal-Mart or the spring boat show. This must now be decided by the state Supreme Court on the last day that any changes can be made to November ballot content. That means the Supreme Court can defer the appearance of the initiative by simply requiring a slight wording change which would never be approved in a single day.

Keep in mind, Judge Limbaugh's ruling that puts this amendment in jeopardy assumes enactment would trigger the changes claimed in the false language drafted by the Secretary of State and claimed by the anti-abortion plaintiff. Those claims are verifiably FALSE.

The judge is intentionally and falsely claiming the amendment requires creation of a new regulatory structure around reproductive care and abortion and fails to specify any of that detail. As a result, it was impossible for signers of the petition to properly understand the impact of the amendment they were supporting and thus the initiative should be dropped. Note that one of Ashcroft's fraudulent descriptions of this amendment was that it eliminated ALL regulation of reproductive care providers. So which is it? What day of the week is it?

The actual amendment does not call for a NEW regulatory structure, it simply calls for reverting to the pre-Dobbs status quo in which

• women had a right to abortions up to viability and where needed to protect the life and health of the mother
• women and care providers had rights to discuss all care options
• doctors could not be prosecuted for performing abortions
• all reproductive care would be subjected to existing medical regulations
• other parties could not discriminate in hiring, etc. based on reproductive care decisions

Regardless of the decision made by the Missouri State Supreme Court, this charade illustrates how far Republicans will go to preserve their fringe policies against the wishes of a strong majority of voters, even in a state dominated by their own party. There's probably no better example of how extremism feeds on itself, continuing to insist on ever-higher levels of philosophic purity and conformity to positions that sixty to seventy percent of the population find abhorrent.

And for those curious, Christopher Limbaugh is the son of a federal judge and cousin of Rush Limbaugh. It shouldn't have anything to do with it, but the cynical hypocrisy in this case was too much coincidence to ignore and leave unverified.

WTH

Trump Counsel Says the Inside Words Outside

Former federal prosecutor and current legal analyst Glenn Kirschner posted an analyis of the discussons and outcomes from Tanya Chutkan's courtroom from September 5. Per his analysis, which seems sound, Trump's lawyer John Lauro said some of the "inside words" out loud in court. These were inside words that most people likely already processed anyway but hearing them in open court was astounding.

After the US Supreme Court issued its ruling on Presidential immunity, it was immediately noted that justice Thomas (I'm leaving the term uncapitalized since for him it really isn't descriptive regarding his work) added a separate opinion for which no other members of the Court added their concurrence. That opinion went beyond the already appalling concepts laid out by Justice Roberts regarding immunity and further stated that the very use of a special counsel for prosecuting a President had dubious legal provinance. Again, no other justice concurred with that opinion, for good reason. The use of special counsels and the right of the Department of Justice to use them and the legality of their funding has been argued MANY times and the propriety of special counsels has been held up in EVERY case across multiple jurisdictions. There IS no issue with special counsels.

Nonetheless, in court on September 5, 2024, John Lauro again raised the special counsel argument. Chutkan shut down the argument immeidately.

Chutkan: You have dicta in a concurrence written by Justice Thomas and you have an opinion filed by another district judge in another circuit which frankly this Court doesn't find particularly persuasive.

Lauro: Justice Thomas in effect directed us to raise this issue and suggested that we do it immediately in light of his view in that concurrence.

(insert sounds of pins dropping here...)

Chutkan: He directed you to do that?

Lauro: Well, I shouldn't say he absolutely said, you know, "Do it." But when you read that opinion, it's absolutely clear that that'ssomething that we have to now to preserve this issue.

Now the Supreme Court faces an entirely new existential issue.

At a minimum, it appears a defendant has attorneys who believe they are being spoon-fed advice from a sitting member of the Supreme Court regarding a case the Supreme Court has CHOSEN to engage in despite the absence of any prior legal precedent requiring them to do so. That same case will absolutely trigger subsequent appeals which WILL return as cases to this Supreme Court. At a minimum, there is no possible excuse for Thomas to be allowed to continue hearing any appeals related to this case.

It is also more certifiable fact than conjecture that a sitting Supreme Court justice is actively cooperating with a party with vital issues before the Supreme Court in what now qualifies as the single biggest ethical and legal conflict that has ever arisen in Supreme Court history. A conflict involving a former President accused of orchestrating and executing a conspiracy to manipulate Electoral College results to retain power and in parallel orchestrating and directing a riot to interfere with Electoral College certification.

If Chief Justice Roberts had any ethical backbone left, he should be spending this weekend looking in a mirror asking himself how he let his court rot this much under his watch and contemplating what he is willing to do to correct it. However, it isn't clear Roberts has any spine so he may sleep like a baby.

WTH

Justice Delayed Is Justice Denied

Judge Merchan's decision to defer sentencing in Trump's New York State criminal conviction is highly problematic, for two reasons. His decision stated:

The public's confidence in the integrity of our judicial system demands a sentencing hearing that is entirely focused on the verdict of the jury and the weighing of aggravating and mitigating factors free from distraction or distortion. The members of this jury served diligently on this case, and their verdict must be respected and addressed in a manner that is not diluted by the enormity of the upcoming presidential election. Likewise, if one is necessary, the Defendant has the right to a sentencing hearing that respects and protects his constitutional rights.

The jury already fulfilled its obligation well outside any reasonable "window" that should be granted around a public figure engaged in a political campaign being subjected to criminal justice for actions years before the current campaign. The judge is responsible for extending their decision about guilt or innocence by factoring in potential penalties and making that final decision, which is the judge's decision alone to make. This particular instance was complicated by a Supreme Court meddling in issues about federal law and evidence about communications of a sitting President about "Presidential acts" which in theory have NOTHING to do with the core crimes adjudicated in this New York State case that occured BEFORE the defendant became President.

In short, it would appear the jury did the hard work. The remaining work may be awkward, it may be unpleasant, but it would appear to have less opinion involved in the final outcome that could be distorted by any perceived political pressure or concerns about an election. As long as external politics don't affect the JUDGE'S sentencing decision, it isn't the JUDGE'S job to worry if his sentencing decision affects external politics. This smacks of James Comey all over again... Someone trying to outguess multiple layers of caroms arising from his job and altering the timing of his actions to somehow "optimize" the "fairness" of a larger universe of actors and behaviors he can neither control or predict.

More importantly, Merchan's rationale for defering the sentencing decision seems to directly conflict with the approach publicly stated by US District Court Judge Tanya Chutkin in a hearing on September 5, 2024 in which she rejected a request from Trump's counsel to postpone formal presentation of arguments regarding dismissal until after the election. The attorney for Trump objected to any schedule for conducting that hearing earlier than the election saying the prosecution could expose new information at a "sensitive time" for the defendant. Chutkan rejected the argument, stating such concerns were "not relevant" and "I am definitely not getting drawn into an election dispute." In essence, she was stating this entire criminal process was STARTED months ago and, once initiated, her obligation was to ensure it proceeded as quickly as possible while preserving all due rights of the defendent.

Having a federal and state court issue seemingly conflicting opinions on back to back days regarding the scheduling of legal actions they must perform involving the criminal prosecution and sentencing of a citizen of the United States simply because that citizen is a public figure not only highlights the fact the defendent IS getting special treatment in violation of the courts' obligation to dispense equal justice, those conflicting opinions serve as additional excuses for the defendent to file more motions and objections about disparate treatment and further delay justice by doing so.

WTH

Election Interference In Missouri

Like groups in several other states whose legislatures immediately passed strict abortion bans immediately after the Supreme Court decision in 2022 overturning Roe v. Wade, forces within the State of Missouri have been working to place a constitutional amendment on the November 2024 ballot that would re-instate access to abortion rights within the state and ensure doctors are not targeted by politicians with vindictive penalties providing reproductive care. Those efforts are being thwarted by Missouri’s Secretary of State, Jay Ashcroft, using a process unique to his official appointed duties. For a second time, a Missouri Court has ruled Ashcroft is abusing his role and essentially interfering with the rights of citizens to place issues on the ballot.

Missouri has a “fair ballot language” law requiring any constitutional amendment put to a vote by the people to include a summary approved by the Secretary of State that summarizes the proposal and provides an unbiased estimate of the recurring cost of the proposal to taxpayers. This requirement essentially pits the Missouri Secretary of State as a final layer of protection of Missouri’s ignert voters against the hordes of special interests who might attempt to get an intentionally ambiguous amendment in front of a naive public who would vote to pass it without understanding its financial consequences.

Of course, that requirement does nothing to protect those same ignert rubes from being misled by a partisan Secretary of State in a position to apply additional political spin around any initiative being put to a vote. Like many other states which enacted these abortion bans in 2022, Missouri’s legislature is overwhelmingly Republican and poses no threat to reversing that abortion ban. But the CITIZENS of Missouri? Like many other bright red states, they may have voted for Republicans overwhelmingly for state offices but they are not in favor of abortion bans either, hence the effort to put a constitutional amendment on the ballot to override the legislature.

This is infuriating to Republican officials so they are using the one tool left in their arsenal to thwart The People from getting what they want via the electoral process. Jay Ashcroft has now used his “power of the pen” TWICE to distort language put in front of voters as a means of peeling away potential votes that would reject the state’s abortion ban. His first attempt involved rewriting the language used to solicit signatures from voters to place the amendment on the ballot. That first effort triggered a lawsuit in 2023 which led a judge to reject Ashcroft’s language on September 25, 2023, stating

The court finds that certain phrases included in the secretary’s summary are problematic in that they are argumentative or do not fairly describe the purposes or probable effect of the initiative.

For the final amendment approved for the November 2024 ballot, Aschroft crafted the following description, which not only appears on his official web site for the office of Secretary of State but also appears at each polling location next to a sample ballot for voters to read before filling in their ballot:

A ‘yes’ vote will enshrine the right to abortion at any time of pregnancy in the Missouri Constitution. Additionally, it will prohibit any regulation of abortion, including regulations designed to protect women undergoing abortions and prohibit any civil or criminal recourse against anyone who performs an abortion and hurts or kills the pregnant women.

The actual amendment

• does NOT allow abortion at ANY time during pregnancy
• does NOT protect doctors from malpractice suits or professional citations for improper care

Here is the language of the actual amendment voters will consider on November 6, 2024:

Do you want to amend the Missouri Constitution to:

• establish a right to make decisions about reproductive health care, including abortion and contraceptives, with any governmental interference of that right presumed invalid;
• remove Missouri’s ban on abortion;
• allow regulation of reproductive health care to improve or maintain the health of the patient;
• require the government not to discriminate, in government programs, funding, and other activities, against persons providing or obtaining reproductive health care; and
• allow abortion to be restricted or banned after Fetal Viability except to protect the life or health of the woman?

State governmental entities estimate no costs or savings, but unknown impact. Local governmental entities estimate costs of at least $51,000 annually in reduced tax revenues. Opponents estimate a potentially significant loss to state revenue.

Note that language in the official ballot EXPLICITLY states that the measure would allow restrictions or bans on abortion after fetal viability except to protect the life or health of the mother. Note that language in the official ballot EXPLICITLY calls for appropriate regulation of providers offering reproductive care. Note that Ashcroft’s summary EXPLICITLY mis-represents these points.

Here is the revised “fair ballot language” summary Judge Cotton Walker required the Secretary of State to post on its website and at the polling locations:

A ‘yes’ vote establishes a constitutional right to make decisions about reproductive health care, including abortion and contraceptives, with any governmental interference of that right presumed invalid; removes Missouri’s ban on abortion; allows regulation of reproductive health care to improve of maintain the health of the patient; requires the government not to discriminate, in government programs, funding, and other activities, against persons providing or obtaining reproductive health care; and allows abortion to be restricted or banned after fetal viability except to protect the life or health of the woman.

Of course, the games do not end in the Secretary of State’s office. Another group of anti-abortion activists filed a suit to block Amendment 3 from appearing on the ballot because it violates Missouri law by including more than one subject and does not explicitly identify which existing Missouri laws would be altered or repealed if passed. That suit will be reviewed in a bench trial slated for September 6, 2024.

So much for that "faithfully execute" language in the oath of office. Ashcroft's conduct and abuse of power in the handling of this issue is beyond rationalization.

WTH

Are AI Vendors Cheating?

A YouTube Short video appeared in my recommendations from the WelchLabs channel that raised an interesting insight regarding current claims of AI capabilities. After reading more on the issue, that finding serves as a useful prompt for explaining some other limitations engineers and business leaders are not fully comprehending about the use of AI technology.

Here is a link to the WelchLabs short with his take on the root issue:

https://www.youtube.com/watch?v=kRVrwjiflrU

A brief summary of the issue being discussed is provided here for context, followed by an analysis of its economic, legal and ethical implications.

How to Train your LLM Dragon

AI systems using Large Language Models (LLMs) operate by ingesting terabytes or petabytes of text data representing anything you can render in text – novels, news stories, computer source code, questions about computer source code, answers about questions about computer source code, etc. – then feeding that text through a multi-stage process. That process

1. splits the entire input set into two subsets:
   • a training set – typically about 90% of the original input
   • a validation set – typically about 10% of the original input
2. chops the training set into arrays of smaller tokens or letters
3. uses powerful specialized chips to perform billions of matrix algebra calculations that derive probabilities of the next token to follow a prior set of tokens based on the training data
4. tests the validity of the probabilities derived from the training data by using strings from the validation set then determining how accurately the LLM predicted the next token or phrase in the validation set
5. takes the results of the test and iteratively feeds that back into the model to adjust its probabilities

If a team building an LLM based AI follows these procedures and obtains a data set that is uniformly distributed in the nature of content and properly splits it 90% /10% into training and validation sets, they can confidently state that a large language model that completes a random string taken from the validation set with “acceptable output” 97.98% of the time is performing better than an LLM only yielding acceptable output 96.1% of the time.

Engineers designing these models rely upon this scoring concept to evaluate their progress in improving the underlying algorithms. Corporate executives managing companies selling these LLMs rely upon these scores to promote their product over the competition.

This competition for engineering bragging rights and revenue makes performance of these tests crucial to all in the industry. Any "leak" of validation data into the training data, either by innocent administrative faux pas or active attempt to manipulate results, has the potential to falsely raise performance scores of the LLM. Experts who design independent benchmarks for evaluating LLMs have devised an approach referred to as a "canary" for detecting such issues. The "canary" is a long random string that would normally NEVER occur in "nature" (normal human literature, source code, etc.). Embedding that string ONLY in text to be used as validation data guarantees that if the LLM ever spits out that canary string, it had to have encountered in in its training.

The easiest way to produce canary strings is to use an encryption hash algorithm such as SHA-256 to generate a fixed-length hash of some subset of other validation data then include that in the validation data. For example, the SHA-256 hash of this string

This is a random string of text supposedly never
occurring in all of the millennia of recorded history
that would be placed in the validation data when
training a Large Language Model.

is

a54cfcf5c13e22d19d52c59320cf7b337723b5e8f221a42dd905f85d6036e7bb

The benefit of a hash algorithm is that ANY change to the input being hashed results in a different output hash value that is not remotely close to the prior hash value. For example, if the above string is altered by adding one additional word, the output hash value looks totally different:

This is a long, random string of text supposedly
never occurring in all of the millennia of recorded
history that would be placed in the validation data 
when training a Large Language Model.

the output hash value looks totally different:

4da1dbe466ab22c43ae1ce712608354183e1fa03ca5e92a353ca0afdde36cab8

If a “canary” is dropped in a 10 terabyte data set in the last terabyte and that 10-terabyte dataset is split 90% / 10% into training and validation sets with the canary string in the validation set, there should be ZERO probability a LLM trained solely on the 9 terabytes of training data WITHOUT that long random canary could ever spit out that long random canary string as an answer to any prompt. If it does, it’s a certainty that the validation data was fed into the training data and the AI’s performance numbers are thus artificially higher than its core algorithms are capable of providing.

Are AI Vendors Cheating?

Cheating is a loaded word. Cheating implies selfish motivations for ignoring a rule as a means for improving one’s outcome over those of others in a competition. What CAN be said about the detection of canary strings in LLM output is that the engineers and companies designing and operating these LLM models have what could be termed a provenance problem which is both a technical problem and a legal / ethical problem that remains unsolved.

At a technical level, LLM engineers are aware these canary strings exist in public test data sets. However, the tools developed to ingest the terabytes of raw data needed for training and validation lack the controls required to explicitly include and exclude specific sources. That has profound practical consequences. If the engineers are simply turning a mindless “crawl script” loose on millions of web sites as a means of accumulating data with no thought as to the validity of the underlying data, the LLM content can be rigged by purposely exposing manipulated data.

Why? And more importantly, how?

The LLM training process doesn’t KNOW anything. For example, it doesn’t scientifically KNOW that

e = mc^2

It can only spit that out in relation to a discussion on physics because it’s seen that sequence of characters in millions of references in physics papers, journals and textbooks dating back to about 1901 along with some dude named Albert Einstein. If someone wanted to hack an LLM to change its answer for “what is the relationship between energy and mass”, they could theoretically turn up millions of web sites faking other scientific papers that include

e = mc^3

and create fake links between those papers with fake publication dates, etc. No matter how much logic was added to the LLM training to try to “pre-scrub” different sources based on original discovery date, URL, reputation, etc., some of those fake e = mc^3 references would make it into training data and eventually alter the probabilities used to answer the question… INCORRECTLY.

From a legal and ethical standpoint, detection of canary strings in LLM output confirms that AI vendors have yet to design their systems to provide a mechanism for data sources to opt out of having data ingested into an AI system. Publication of content on a publicly accessible website is NOT legal approval for that content to be ingested into another system, abstracted then used for financial gain or other purposes. The fact that major vendors in this space have been essentially crawling the web for four years to develop and train AI systems without the consent of content owners reflects a stunning leapfrog beyond existing legal concepts and existing intellectual property protections. The method by which these systems were launched amounts to some of the largest oligopolies in the history telling their government and the citizenry Just try and stop us.

Scrubbing Large Language Models

This canary problem points out a more fundamental flaw with large language models and correction capabilities for the data within them. There IS no correction capability within a large language model. The summary above about training stated that LLMs function by parsing source data into smaller tokens and letters, then mathematically deriving the probabilities of the “next token” based on prior tokens. The training process itself involves loading the raw training data as tokens or characters into arrays with thousands of elements, doing the matrix algebra, then saving the calculated probabilities in output matrices and iterating the process with more training data. Once the training phase is completed, the final running model consists ONLY of those matrices of probabilities. None of the source data is referenced again when processing the requests of individual users.

The segmentation of modeled probabilities from the original data means that if the operators of the model are informed that their training data included text from 10,000,000 social media posts that were forwarding a meme that joked that e = mc^3, there is no way for that operator to go into their running LLM system and surgically purge those ten million references that threw off the model’s accuracy when prompted for the relationship between energy and mass. The mathematical models don’t preserve a backward “chain of provenance” between an output probability and the training data that produced it.

By analogy, this problem is equivalent to being instructed to average a collection of one million integers, compute the average, then throw away the set of one million integers and only hold the average. If someone then tells you that ten thousand of the inputs were bogus and should not have been used, all you have is the average and maybe the count of values from the original data. The only way to correct your average is to re-gather the original input set, somehow identify and exclude the ten thousand bogus values, then calculate a new average.

But in the LLM world, the discovery of canary strings in LLM output seems to have proven that LLM designers and operators either don’t care that they lack control over which data is being ingested to generate their model or (equally likely) the logic doesn’t exist to provide the needed fine-grain control during training. And even if the operators HAD the tools to more diligently filter input, removing the flawed probabilities from a running model requires re-training the model from scratch.

Given that some of these models require multiple MONTHS of processing by thousands of dedicated servers, it is apparent those developing these models are unwilling to incur those costs simply to exclude data they weren’t authorized to use or data shown to be flawed. Their attitude seems to be Sucks to be you. Sorry, we’ll try better in the next release. And since they still may lack the controls to exclude certain data, the only way the flawed probabilities resulting from intake of undesired data can be corrected is by loading MORE data from other sources to tip the calculations to a “truthier” state. That is not a logically or ethically sound foundation for any technology.

WTH

Citizens United - The Ultimate Grift

Nearly every candidate sends them. Emails with subjects like "Federal Deadline Looms" or "Send a Message With a Donation" or similar nonsense. All of these requests are a reflection of how the Citizens United case warped the operation of democracy and distorted perceptions of who really votes in elections, people or dollars? The message behind these annoying fundraising attempts could not be more clear. Or misleading. The message is that giving money to a political candidate can essentially allow you to vote multiple times. Every monthly Federal Elections Commission reporting cut-off for fundraising and expenses is a chance for a candidate to generate buzz for their campaign and a sense of momentum if the report shows they out-raised their opponent by a material amount. And since money is speech according to Citizens United, more money can buy more speech and a greater likelihood of victory.

Of course, there is a logical flaw of trying to run a Citizens United based campaign in the physical world, especially a campaign for President.

The idea that money is speech that leads to the elimination of limits on money into campaigns doesn't solve a core problem for a candidate or party who thinks they can buy an election win by buying up airtime / ad impressions. The problem is that there is a finite amount of TIME available in which those "impressions" can hit targeted voters. And the more you know about the status of a race in every gerrymandered portion of the entire national footprint, the more tempted you are to NOT spend money where you're leading so you can TRY to spend more money where you are behind but withing striking distance. And you likely WILL spend more money in those narrowly identified markets, partly because you HAVE the money by not spending it elsewhere but also because you are focusing your spend on a smaller number of TV stations, radio stations, and ZIP-targeted online ad impressions. Like any other market, when demand goes up and supply is fixed, rates go up. But a voter can only watch so much TV or YouTube per day. A candidate can spend a BILLION dollars instead of a MILLION and it won't give them 100x the lift.

In a world with a rational understanding of physics, media, politics and economics, the costs of running a national campaign should reach some limit of some sort. In a highly gerrymandered environment further skewed by electoral college distortions, a Presidential election with (say) 160 million voters comes down to only about four million voters in maybe twenty counties across seven states. Every other district is so far in the safe zone for one party or the other, there is literally zero payback for advertising there. So in the remaining "swing" districts, there has to be some upper limit on how much can possibly be spent on TV ads, radio ads, YouTube ad inserts by ZIP and direct mail ads by ZIP. So why would ANY Presidential campaign need to raise and spend five hundred million dollars?

The real problem with hundreds of millions of dollars being spent on political campaigns courtesy of Citizens United is that such spending isn't actually resulting in hundreds of millions of dollars of political awareness on the part of voters for or against any candidate. It is simply funneling money to future cronies who act as "media consultants" and take a percentage of the "ad buys" for creating some of the spots or supposedly fine-tuning the selection of media outlets to use for each ad. But look at the ads being aired. When you are flooding the zone with ****, it doesn't take a rocket surgeon and it doesn't take multiple rocket surgeons to make those decisions. Taking a 5% or 10% "cut" of these ad buys is pure political corruption in a time machine.

Of course, the other obvious pattern of corruption in this Citizens United hell we occupy is that these hundreds of millions of dollars in ad spending funneled through cronies to collect their skim eventually goes into the pockets of highly concentrated traditional media and social media firms. This bi-annual manna from heaven provides continued incentives for them to do nothing to prevent abuses of their platforms and zero incentive from a journalistic standpoint to report on this corruption in general or the particular corruption of any given candidate. Their attitude seems to be: With these financial windfalls coming in every two years, we'll just shut up and cash the checks.

WTH