Clive Robinson November 20, 2023 9:16 AM

@ Bruce, ALL,

Re : Using Generative AI for Surveillance

“My guess is that it isn’t very good yet, but that it will get better.”

It’s a very obvious use of two different ways it has of analysis,

1, What the user types in as queries etc.
2, From scraping user produced Internet and similar content.

As I’ve been warning on this blog for a while now LLM’s with or without ML assistance will be primarily used for one of two main purposes,

1, Surveillance and aquisition of PPI.
2, Arms length excuses to avoid responsability for “mantra based” activities for political or financial gain.

The fact that Microsoft are going “all in” on this tells you that their use is not going to be in anyway benign, which is why I’m surprised Sam Altman has taken a position with them.

But to give you an obvious example for use that will in the very near future get “talked up” if it has not already is,

“Finding potential persons of interest”

By various “Guard labour” entities.

Apple have demonstrated that the CSAM argument is not going to be at all well received for,

“On User Device Scanning”

Less so than the push for the sofar failed attempts for,

1, Golden key back-doors.
2, End of E2EE.

So they want to jump from basic technical methods against sustems and devices, which was always going to fail as I’ve explained before all a user has to do to beat either is,

“Move the security end point beyond communications end point”

By taking “encryption off device” and using if push comes to shove a secure “hand cipher” which can be done by anyone capable of rational and deliberative thought.

From the “Guard Labour” perspective you are mostly only a threat if what you say or write attracts public attention. To do that it has to be “publicly available”.

Even if not specifically tagged as being from an individual be it by real name or pseudonym. For it to work all it needs is “stylistic match” which LLM’s do incredibly easily. Which then gives a “time line” that can be correlated with individuals time lines.

This can then be pushed by NSL’s and similar to turn into if required “probable cause” to go for more intrusive surveillance, or other anti-personel activities.

On of which is employing a technique seen at Bletchly during WWII they called “gardening” where information and events get focused around you such that you make revealing crypto or similar activity.

None of this is difficult to think through, which means it is going to happen unless made expressly illegal at all levels of “guard labour” and similar agencies / entities.

Which means that “Data brokering” and similar surveillance capitalism has to be made illegal and very promptly. Which would kill the main “Internet income” model of the big corps in Silicon Valley etc.

And in the process kick one of the legs out from under the “the user as a product” model that is so dangerous to society.

lurker November 20, 2023 12:26 PM

“There is never going to be a way of making AI unbiased…”
– Andy Martin, Cellebrite engineer

So, we made it in our own image. No surprise there.

AL November 20, 2023 12:27 PM

This AI is going to be a barrel of laughs. I’m using Windows 11 “Copilot”, so I gave it a whirl with one of the regular posters here.

I ask it “Internet postings by XXXX #####” (name omitted)

XXXX ##### has made several posts on the internet, particularly on the Schneier on Security blog. Here are some of his recent posts: …

Same with a Reddit user. This is going to get so much better going forward, particularly once AI for the masses becomes objective, which Copilot isn’t. We’ll be asking it questions like “What are the odds that Jeffrey Epstein committed suicide versus Epstein being murdered? What are the odds that Covid-19 emerged naturally versus a lab leak?

This stuff is going to be a riot. I love it already.

AL November 20, 2023 1:25 PM

I asked the robot “How can AI be made unbiased?”
Artificial Intelligence (AI) can be made unbiased through several approaches:

  1. Careful Analysis, Design, and Testing: AI can fall prey to bias in many ways, but careful analysis, design, and testing can ensure it serves the widest population possible¹.
  2. Blind Taste Tests: More strategic use of AI systems through “blind taste tests” can help identify and remove decision biases from the underlying algorithms. This involves denying the algorithm the information suspected of biasing the outcome to ensure it makes predictions blind to that variable².
  3. Human-in-the-loop Processes: Consider how humans and machines can work together to mitigate bias, including with “human-in-the-loop” processes³.
  4. Pre-processing the Data: Maintain as much accuracy as possible while reducing any relationship between outcomes and protected characteristics, or produce representations of the data that do not contain information about sensitive attributes⁴.
  5. Feedback from Devil’s Advocates: To create an algorithm that is as fair and unbiased as possible requires humans who essentially act as devil’s advocates. They will examine the results of the AI decisions and look for what it’s doing right and wrong. This allows the designers to take the feedback from people who are looking for these biases and alter their algorithm⁵.

Remember, it’s easier to program bias out of a machine than out of a mind².
Source: Conversation with Bing, 11/20/2023

lurker November 20, 2023 2:58 PM


programming a machine or a mind?

All those suggestions from Bing involve subjective tweaks by humans. Asking an AI to fix its own bias cannot end well.

AL November 20, 2023 3:56 PM

I think we’re at a very rudimentary phase of AI. Copilot is basically a summarizing service for search. There is certainly a lot of bias in search. It does save time though.

My view on AI is, garbage in, garbage out. The inputs need to be carefully controlled. But, maybe they’ll be AI to control other AI. And then what do you have – 5+5=10, or 5×5=25? I do think it is possible given some time to get AI programing itself to reduce bias. But at this point, we’re at the top of the 1st inning with this. Copilot is eye opening. This technology has a future.

There has been talk about AI generating disinformation. But it can also detect and get rid of disinformation. Where this goes is anybody’s guess.

vas pup November 20, 2023 6:45 PM

@lurker – I am with your point absolutely.
Moreover, there danger of #3 and #5 as #AL posted when objective analysis would overridden by subjective motivational biased. E.g. override objective test results of universal test for university admission by so called holistic approach which is basically kept secret and let decision making biased and subjective.
The problem as I see we should agree that temperature now is let say 65 degrees F and AI should accept as a truth by our opinion/feeling of it could be different: you fell warm but other person cold.

@Clive – thank you for thoughts very insightful as usually.

@Bruce – thank you for the link and this part of the article in particular:
“Rory Mir, associate director of community organizing with the Electronic Frontier Foundation, said that AI could help law enforcement broaden their surveillance efforts, allowing smaller teams of cops to surveil larger groups with ease. Already, =>police agencies frequently use fake profiles to embed themselves in online communities[fake celebrity accounts in particular – vp]; this kind of surveillance has a chilling effect on online speech, Mir said. He added: “The scary thing about stuff like ChatGPT is that they can scale up that kind of operation.” AI can make it “easier for cops to run analysis quicker” on the data they collect during these undercover operations, meaning that “AI tools are [effectively] enabling” online surveillance, he added.”

My nickel: you know now when you call business or/and gov entity you hear prerecorded message that your conversation could be recorded for training and [whatever other] purpose. Bank could record your voice for verification. I am absolutely sure those recording are also used as input for AI training/analysis and used by business and gov for surveillance. As last week post suggested you voice could disclose too many information about your health, mental status which kind of protected by HIPAA but in reality as LEAs like to say “You have Constitution/Law – we have secret internal instructions/manuals which let us do secretly whatever we want” That adage came from Soviet/Stasi era but we are not here from far of it as well.

Erdem Memisyazici November 20, 2023 10:46 PM

The article is somewhat misleading I thought. Surely Chat-GPT can be used to guess sentiment but you’d need more than a single comment and even then a percentage may be correct. Sarcasm is a good one. “This is fantastic” is quantified as “positive” yet you’d have to run the article, and several more comments to see if the billions of numbered sentences are likely to follow where some sense of sarcasm is indicated. You’d need biological information like cortisol levels, heartbeat, breathing rate, pupil contraction, gait tracking etc. long-term to really have an okay shot at getting somewhere in the 90%. Good enough data by the numbers perhaps but nothing to bet it all on.

ResearcherZero November 21, 2023 4:44 AM

Data becomes more useful as it is aggregated, packaged and stored for use with other tools.

Babel X will be used to “identify potential derogatory and confirmatory information” associated with travelers, persons of interest, and “persons seeking benefits.” …results from Babel X will be stored in other CBP operated systems for 75 years.

Gravy Analytics repackages that location data and sells it to law enforcement agencies via its related company Venntel.

Locate X can draw a shape on a map known as a geofence, see all devices Babel Street has data on for that location, and then follow a specific device to see where else it has been.


The location data CBP bought access to from Venntel was “global” in scope…

“search for which devices were in a particular place at a specific time, or selecting a unique identifier for a target device and then seeing a history of where that individual phone, and potentially by extension a specific person, has been”


Babel X allows users to input one piece of information about a surveillance target, like a Social Security number, and receive large amounts of collated information back. The returned data can include “social media posts, linked IP address, employment history, and unique advertising identifiers associated with their mobile phone. The monitoring can apply to U.S. persons, including citizens and permanent residents, as well as refugees and asylum seekers.”

Then track that data against some other database…

“The scale of the data available to and routinely searched for the benefit of law enforcement under the Hemisphere Project is stunning in its scope,” Wyden’s letter to Garland says.

Call records stored by AT&T do not include recordings of any conversations. Instead, the records include a range of identifying information, such as the caller and recipient’s names, phone numbers, and the dates and times they placed calls, for six months or more at a time.


Official government presentations estimate 4 billion call detail records populate the Hemisphere database on a daily basis. That includes records dating as far back as 1987, which is much further back than the records most telcos store.

Whereas normally public and private security cameras are separate, with only their respective owners or administrators able to view any feeds, Fusus aims to bring all of that footage together in one place.

That can include doorbell cameras, drones, robots, fixed surveillance cameras, helicopters, hidden cameras, police body cameras, and cameras in schools and churches, according to multiple documents. Police departments then gain access to a map interface showing all of the camera locations. (Everywhere)

“Our school systems, our churches, our places of worship, our kindergartens, they’re all on board and becoming video-savvy and a part of the system.”

Fusus’ product not only funnels live feeds from usually siloed cameras into one central location, but also adds the ability to scan for people wearing certain clothes, carrying a particular bag, or look for a certain vehicle.


The model expands police access to personal information collected by private cameras that would otherwise require warrants and community conversation. Because these cameras are privately owned, police can enjoy their use without having to create and follow records retention and deletion policies.

You Better You Bet November 21, 2023 2:15 PM

“My guess is that it isn’t very good yet, but that it will get better.”

“Better” for what purpose? The example provided discusses police using it to ingest and summarize social media activity. So your “guess” seems to imply that getting better at surveillance-oriented sentiment analysis will somehow be a good thing.

If so, who are you and what have you done with @Bruce?

AL November 21, 2023 10:25 PM

@An Old Man
Absolutely. I’ll give it a whirl.
“Why was Sam Altman fired?

Sam Altman, the CEO of OpenAI, was fired following a “deliberative review process by the board”. The board concluded that Altman was not consistently candid in his communications with them, which hindered its ability to exercise its responsibilities⁴⁵⁶. As a result, the board no longer had confidence in his ability to continue leading OpenAI⁴⁵⁶. However, the exact reasons behind his dismissal are still unclear¹²³. Some reports suggest that his other interests, including ChatGPT, might be to blame³. After his departure, Altman joined Microsoft to run a new AI effort⁴. This is a developing story and more details may emerge over time⁴.

Source: Conversation with Bing, 11/21/2023
(Followed with references.)

It is pretty much a summarizing service after conducting a search.

vas pup November 22, 2023 5:46 PM

Backed by Intel and Comcast, Israel’s AI21 raises another $53m at $1.4b valuation

“Israel’s AI21 Labs, a natural language processing (NLP) startup, said on Tuesday that it has secured another $53 million for its Series C funding round, which values the unicorn at $1.4 billion.

=>NLP is the ability of a computer program to understand human language by speech and by text. With the recent hype over ChatGPT, a so-called large language model that uses deep learning to spit out human-like text, other startups such as AI21 Labs have been quick to come out with competing AI models.

AI21 Labs, which has a vision to bring generative AI to the masses, said that it will use the new funding to focus on providing “reliable and useful” AI
technology to companies globally, expanding the company’s fields of activity, and hiring about 100 additional employees by the end of 2024. Currently, the startup has about 260 employees, mainly in Israel, Shoham said.

!!!AI21 Labs created a software platform where developers can build text-based
applications like recommendation engines, chatbots and virtual assistants. The
company launched a text simplification tool called Wordtune, a Google Chrome
extension that helps clients improve or streamline content, and Wordtune Read, a tool that analyzes and summarizes documents.”

vas pup November 22, 2023 5:57 PM

Israeli startup comes out of stealth to tackle cyber threats in the GenAI wild west

“While one of its founders is drafted in the Israeli army to fight the Hamas terror group, Tel Aviv-based startup Lasso Security came out of stealth on Monday armed with $6 million in seed funding.

The Israeli startup has secured seed funding led by London-based Entree Capital
with the participation of Samsung Next, to enter the fray in the market for the
protection against cyber threats as more and more businesses are adopting
generative AI technologies.

Lasso confirmed it rejected investment capital from Web Summit’s venture
capital arm, after the outgoing CEO of the tech summit company accused Israel of committing “war crimes” in its response to the Hamas terror group’s devastating onslaught.

With the recent hype over OpenAI’s ChatGPT, a so-called large language model
that uses deep learning to spit out human-like text, and Google’s Bard chatbot, an array of new applications and tools in the field of AI have been sprouting at a swift pace. The fast adoption of new GenAI and LLMs technologies by businesses and enterprises also !!!makes their network systems and devices vulnerable to multiple security threats, including data exposure.

With the recent hype over OpenAI’s ChatGPT, a so-called large language model
that uses deep learning to spit out human-like text, and Google’s Bard chatbot, an array of new applications and tools in the field of AI have been sprouting at a swift pace. The fast adoption of new GenAI and LLMs technologies by businesses and enterprises also makes their network systems and devices vulnerable to multiple security threats, including data exposure.

!!! Lasso has developed a security platform to help organizations make the
transition to the GenAI technological revolution, while protecting their data,
customers, and employees.

=>The startup says that its platform vigilantly safeguards every data point,
command, and prompt generated by LLMs, allowing businesses to embed GenAI
technologies without jeopardizing both data and user privacy while protecting
against malicious threats.

Lasso says its platform is addressing numerous, sophisticated and emerging LLM
cyber security threats, such as
=>model theft, malicious code generation, prompt injection, data poisoning, and supply chain attacks.”

JonKnowsNothing November 23, 2023 5:30 PM

@vas pup, All

re: AI in emergency departments easing some of the burden on staff

Not the best idea, nor what will really happen should you end up in the ER with an AI-Attending.

There is a medical check list called the SOFA Score. That’s the check list that determines your fate. During 2019-2022 the fudging about SARS-CoV-2 unveiled this not-medically secret but secret to civilian list.

  • The SOFA scoring system is useful in predicting the clinical outcomes of critically ill patients.

Current medical training is based on an economic model of Least Cost Medicine and Evidence Based Effectiveness. These are bell curve models for everything in the western medical universe aka practice. The upper end will get well, no matter what, the middle stripe is the area of practice, the bottom are marked off for death (not necessarily immediate).

AI can certainly calculate your SOFA Score fast enough. It won’t save your life, because your life has already been divided by the bell curve for your condition.

Generically, MDs focus on the common conditions where ever they practice eg: Frostbite is not common in the Sudan and Snake Bites are not common in the Antarctic. AI may have a larger built in medical dictionary but the treatment depends on the MD’s personal experience.

The main thing AI might bring is more restrictions on the Evidence Based Medical Front. By hallucinating some output about your condition, it can be helpful to skew the medical profile for it. That medical profile is used by all MDs to determine where on the bell curve the lines get drawn.

It’s on the same level as the UK plan:

  • Disabled people must work from home to do ‘their duty’. Laura Trott UK Treasury

Essentially moving the left slice point on the bell curve to mark people “Well Enough To Work” who are officially “Not Well Enough To Work”. Moving the line does not make magically make people “all better”.

AI Medicine is all about moving the line and hallucinating medical treatments.


ht tps://en.wikipedia . org/wiki/SOFA_Score

  • The sequential organ failure assessment score (SOFA score), previously known as the sepsis-related organ failure assessment score, is used to track a person’s status during the stay in an intensive care unit (ICU) to determine the extent of a person’s organ function or rate of failure. The score is based on six different scores, one each for the respiratory, cardiovascular, hepatic, coagulation, renal and neurological systems.

All hospitals have TRIAGE plans. These are pretty well hidden from public view. TRIAGE plans include SOFA Score adjustments and logistics of how long a hospital can operate without outside support.

The state of hospitals in Gaza demonstrate how long a hospital can function without support systems and supplies. AI won’t make any difference.

Clive Robinson November 23, 2023 6:19 PM

@ JonKnowsNothing, vas pup, ALL,

Re : AI as clinical decision maker.

“AI won’t make any difference.”

Actually it will to,

1, Increasing profit
2, Reducing cost
3, Reducing life expectancy

And because it’s “arms length”

4, Increasing lawyer costs.

Which in turn will,

5, Increase medical,liability insurance costs.

I could go on but the main point people realy need to remember is health care is a big Merry go-round of money, where those actually giving healthcare like doctors and those receiving it like patients are the loosers and the others especially the white collar profiteers are the winners.

What AI will do is “tip the bias” further in the direction of the profiteers…

As an example…

Hearing aids, use the same chips and technology as those less than 50USD phone headsets… Yet in the UK a hearing aid will cost you upwards of 4000USD…

Epi-pens that cost maybe 1-2USD in parts well 1200USD for two.

Various Diabetes medications are so expensive that a person on an average income could not earn enough in a year to buy even a fraction of the medication cost for a year… Oh and for many years the AMA advice on diabetic control, load them up with carbs and wackin loads of meds. They will get fat develop all sorts of diseases and die a very early death, but it makes doctors costs less and drug companies and insurance companies profits higher…

So much for “caring and sharing”…

JonKnowsNothing November 23, 2023 11:16 PM

@Clive, All

I concur with your economic forecast of the knock on effects of AI Medicine.

To get a good detail on the changes in the USA since ~1970 there is a book on the rise of Deaths of Despair in the USA. It’s written so the average person can understand the changes. This phenomena in the USA is completely at odds with the rest of the world. Things might not be great at the NHS but in the USA this classification of deaths is skyrocketing off the charts.

The book is written by 2 Econ Profs and one has a Nobel in Econ, so the numbers are pretty solid.

If you are (M 45+ non minority) you have a serious problem. If you do not have a college degree, the problem is significantly worse, by magnitudes. This only covers deaths from suicide, alcohol and drugs. In the rest of the world the rate of deaths in this category has been declining steadily but in the USA our rates are cliff lines going up.

It’s pretty much as you describe for AI Medicine. If the USA death of despair rate is so horrendous with standard medical care, one can hardly envision how much worse it will be under the economic scenario you envision.

Deaths of Despair and the Future of Capitalism.
Princeton University Press.

Case, Anne; Deaton, Angus (2020)

ISBN 978-0691190785.


ht tps://en.wikipedia.o rg/wiki/Deaths_of_despair

  • A disease of despair is one of three classes of behavior-related medical conditions that increase in groups of people who experience despair due to a sense that their long-term social and economic prospects are bleak. The three disease types are drug overdose (including alcohol overdose), suicide, and alcoholic liver disease.

Clive Robinson November 24, 2023 12:24 AM

@ JonKnowsNothing, ALL,

Re : Book on lower middle class and working class in the US.

“To get a good detail on the changes in the USA since ~1970 there is a book on the rise of Deaths of Despair in the USA. It’s written so the average person can understand the changes.”

There is a “working paper” from the authors that can be downloaded which covers things a little bit more briefly.

You can read a very brief outline and download the paper from The “National Bureau of Economic Research”(NBER),

Importantly it explains the “how” of Corporate Capitalist Culture acting as the primary enabler and driver.

Which is also happening in the UK all be it more slowely and was started by Tony Blair and pushed by David Cameron…

Winter November 24, 2023 4:28 AM

@Clive, JonKnowsNothing, All

Re : Book on lower middle class and working class in the US.

There is a sentence in hte abstract that drew my attentions [1]:

Concurrently, all-cause mortality in the US is diverging by education—falling for the college-educated and rising for those without a degree—something not seen in other rich countries.

It puzzles me why this is so different.

One reason I can think of is the way election campaigns are waged. US campaigns are all about the economy and cultural issues that have no bearing at all on income.

In Europe, campaigns are much more explicitly about income, eg, “purchasing power”, livelihood, minimum wage, growth of wages, etc.. There are obviously many politicians who want to distract the voters with, eg, “immigration” scares. But I see income much more often taking a central stage.

Or is it just the fact that Corporate&Political America have succeeded in crushing the unions?

[1] Full text in article section Education and Mortality in Other Rich Countries:

As is to be expected, those with more education have lower mortality rates. In most of the countries, age-adjusted mortality rates fell from 1980 to 2014 for all three education levels, and in many countries they fell more for less-educated than for more-educated people. However, because the larger decreases for the less educated started from higher baseline mortality rates, the percentage declines in mortality were smaller for the less-educated group—showing a widening in relative mortality inequality while absolute mortality inequality decreased.

These mortality patterns in Europe are different from, and much more benign than, those in the United States presented in Figure 1, where mortality rates for those with and without a BA are going in different directions. Mackenbach and colleagues do not find this divergence in any European country in the twenty-first century, but there are six Eastern European countries—Hungary, Czechia, Poland, Estonia, Lithuania, and Slovenia—where, prior to 2000, mortality rates were rising for less-educated men while falling for more-educated men. (The same was true for women in two of these countries.)

Winter November 24, 2023 5:25 AM

@Clive, JonKnowsNothing, All

Re : Book on lower middle class and working class in the US.

Looking around, I found this (scroll down):
‘Deaths of despair’ on the rise in the US: Why here and not in other nations?

They also note that educational options are handled more at the federal level, college costs are lower, and employers allow more vacation time. They believe that these factors might contribute to the lower rate of despair-related deaths in these countries.

Dr. Sterling elaborated to MNT: “All these countries provide support from cradle to grave to help families make it through each stage of the life cycle, including prenatal care, maternity leave, preschool and elementary education, post-secondary education, healthcare, and vacations. The U.S. provides no such assistance, and the National Academy report does not focus on them as key to solving our crisis.”

And then the kicker:

“Every symptom of despair has been defined as a disorder of dysregulation within the individual. This incorrectly frames the problem, forcing individuals to grapple on their own by learning resilience, self-help, and so on. It also emphasizes treatment by pharmacology, providing innumerable drugs for anxiety, depression, anger, psychosis, and obesity, plus new drugs to treat addictions to the old drugs. We cannot defeat despair solely with pills — to the contrary, pills will only deepen it.”

So what is completely different between the USA and “those other” countries is that in the US, solutions are sought in changing the individual. In “those other” countries, the solution is sought in supporting those individuals in their community, eg, prenatal care, maternity leave, cheap education, vacation time, etc..

Is anyone surprised that changing the individual and supplying pills won’t work if you do not change the circumstances?

Clive Robinson November 24, 2023 7:54 AM

@ Winter, JonKnowsNothing, vas pup, ALL,

Re : Mortality increase in the US in lower middle class and below.

“Is anyone surprised that changing the individual and supplying pills won’t work if you do not change the circumstances?”

I suspect the general answer would be “No they are not surprised”

However a ridder should be added of,

“But such increases in mortality for some is highly profitable.”

Which brings us back to the question of,

“Why is such depression high in the US and not else where?”


“Why only in Whites not other ethnic groups?”

That is the statistics suggest that this is not due to “accidental side effects” but some form of “deliberative policy” and as importantly it appears to be politically focused in distribution across the US.

Winter November 24, 2023 8:17 AM


“Why only in Whites not other ethnic groups?”

Look at their figure 1 in your link.

The figure clearly shows that mortality in non-whites was much higher and fell over the years to close to that of WNH. Then they started to increase in lockstep with the low educated WNH.

So the answer is simple, non-whites had so much higher mortality that that swamped the education-dependend effects. Only after those other factors were removed did the effect of education show.

Clive Robinson November 24, 2023 8:56 AM


To continue,

In the article you link to we find two paragraphs.

The first,

“The increase in mortality in the U.S. is most significant among white adults in rural areas with low education levels and low income. In contrast, the mortality rates among Black and Hispanic adults have declined.”

Specifically states it is of a single socio-geo-racial group

1, White adults,
2, In rural areas
3, With low educational levels
4, With low income.

But immediately the following paragraph

“The increased mortality is related to a few different factors, including metabolic and cardiac diseases. These disorders are associated with obesity, which has dramatically increased among all racial and ethnic groups.”

They say amongst “all racial and ethnic groups”, so it would be logical to conclude that,

5, Metabolic and cardiac diseases,
6, Increased obesity,

Would not be the cause as it’s common to most similar socio-economic groups.

This apparently leaves geo-political.

That is where they live and who they have voted for / have had inflicted upon them over the last ten to thirty years of their lives.

However there is another view which you will find expressed in other places that, what has actually happened is they have “become the forgoton” thus “neglected”. That is the political focus has been on “other groups” who’s “lot in life” has improved by political attention to their needs thus their circumstances have improved in relation to what they once were.

However that is not right either because the trend is very definately downards in life expectancy in this particular socio-economic-regional-racial group (see points 1-4). The fact the regions are widely distributed but apparently political colour related.

Suggests it might be a “slice of the pie” problem. That is as you give a bigger slice / more to one group it is at the expense of other groups. With the result that the more groups that see some improvment even if it is small is at the expense of just one or two groups where the detriment is significant.

But there is another issue I’ve identified in past postings which is it is caused by “pain and fear”. That is physical stress such as back problems cause a very significant number of “lost hours” and it’s known in the US that if you take time of work then very quickly you won’t have a job, nor will you have health care and other benifits that in other countries are paid mostly by the state equitably for all and not dependent on if you are employed or not.

US Doctors are well aware of this, but they are also acutely aware that US health care funders won’t pay for “physical therapy”. Physical stress injury is not cured by pills or potions but by rest, physical therapy, and sometimes surgery. If your Doctor knows you won’t get physical therapy, and rest or hospital visits will get you made unemployed, then all they can realy do is “medicate the symptoms” by drugs that have no hope in hell of curing the real cause of the physical stress, but might help keep you employed, and maybe help you sleep such that you don’t drive off the road or loose limb/life due to tiredness inattention when using machinery and other “force multipliers”.

Winter November 24, 2023 9:09 AM


That is where they live and who they have voted for / have had inflicted upon them over the last ten to thirty years of their lives.

My reading is that the increase if mortality in this group is linked to despair. In other disadvantaged groups mortality decreased, but from a much higher level, due to overal improvement of living circumstances. That is, these other groups are slowly coming to the situation of low educated rural WNH.

The increased despair is indeed limited to a group of low educated rural WNH. But that is because there is less work for these people and it pays worse. Mortality in other disadvantaged groups does not decrease because they have already been in this misery for ages

JonKnowsNothing November 24, 2023 10:14 AM

@Clive, @Winter, All

re: Deaths of Despair by Education Level: Farmers


In USA Agriculture, older male farmers have an extremely high rate of suicide.

I have known a few older farmers who took that route. I also have known non-farmers who took that route too. Which proves nothing, in of itself.

In the USA, farming is not just about growing stuff. A good number of farmers have 4yr college degrees and many have post graduate degrees. Today’s high tech farming, and import export trade is part of the degree program. We have colleges and universities who’s main curriculum is Ag Business.

In the USA, a vast majority of farm labor is unskilled and uneducated. Many cannot read or write. They do labor that city folks won’t do for one hour.

So the question is this:

  • Why do educated old farmers die from despair “more than” their low paid back breaking stoop labor contracted workers?

The diagrams in the books show clearly a deep divide between education levels of all groups. You are less likely to die from despair if you have a degree and more likely to die from despair if you do not. It is anecdotal, but for farmers, it’s a big problem and education does not seem to be a preventative.

JonKnowsNothing November 24, 2023 10:40 AM

@Clive, Winter, All

re: Deaths of Despair v Metabolic, Cardiac diseases, Obesity

Deaths from metabolic, cardiac conditions and obesity are tracked in their own categories.

Deaths of Despair: suicide, alcohol, drugs is the catch basin for unnecessary deaths. It is this category that has the biggest shift going the wrong direction contrary to other markers.

Also fwiw, (USA) for low income people in most states they do get health care through a program called Medicaid(low income). Each state has their own implementation and name for the program. The primary criteria to get this health care is based on The Federal Poverty Level Index. The programs can be more comprehensive than open market health care (not company provided) and with more options than many retirement health care options offered by Medicare (65+).

It is the working people, making more than the Federal Poverty Index and younger than 65yo, that have no health care.

Healthcare for working people is provided as a corporate perk. You lose your job, you lose your healthcare. Once people see the cost of their corporate carry-over COBRA plan which allows people to keep that PPO plan provided they pay the full cost out of pocket, they drop it. Few can afford those costs when you have no income.

Winter November 24, 2023 11:15 AM

@Clive, JonKnows

re: Deaths of Despair v Metabolic, Cardiac diseases, Obesity

The reasoning is that cardiac arrest is one of the consequences of obesity. And obesity can be seen as the result of “substance” abuse, ie, overeating on high caloric food. Thus cardiac arrest could be analysed the same way like liver cirrhosis and drug overdosis.

There are good reasons to treat it like this. However, it is methodologically complicated.

Clive Robinson November 24, 2023 11:48 AM

@ Winter, JonKnowsNothing, vas pup, ALL,

“The increased despair is indeed limited to a group of low educated rural WNH. But that is because there is less work for these people and it pays worse.”

I think it’s actually rather more than,

1, Less work available.
2, Lower wages.

Consider “job security”.

I noticed when working with Far Eastern (Japan, South Korea, Taiwan, etc). That there is a significant “keep it to myself” culture.

Most manual labour can be done by those who can “pick up the tool” therefore as an employer I have no reason to care if you come or go because I can replace you in as little as half an hour (have a look at so called “gig-economy”). However as a job becomes more sophisticated lead in time becomes increasingly a problem. That is the “on the job training” asspect increases and this is productivitively “dead time” for the employer, yet it has the “full cost” for that time. Thus whilst the “body” might be near instantly replacable the “learned skills” are not. Thus the “fire and hire” cycle is reduced.

The more technically sophisticated, and especially when much of the job can be in an employees head not their muscles the more dificult it is to replace a person.

When you are talking about engineering design and similar to lose a “key employee” half way into a project can kill the entire project dead with massive attendent costs / losses.

In the Far East they knew this thus where they could they kept “secret knowledge” very much to themselves as firing them would be a major issue. Those who used to work with “Microsoft Foundation Class”(MFC) libraries, knew that using the “Microsoft approved entry points” was at best inefficient, and it was said by those that “reverse engineered” Microsoft applications that there were other entry points Microsoft used but did not publish so Microsoft had a “competitive advantage”. How true this was is open to debate, but it’s certainly true in the early days that “power coders” used to use many unofficial and undocumented entry points. They very certainly tried to keep such things very secret even in their own code, hence the “we know but don’t document” barrier to others who might replace them.

As a rule of thumb, the better qualified you are the more your job is “brain skill not brawn skill” and what you hold in your head is of vastly more worth thus your job security is very much higher. Which also counts for a lot when it comes to “pay negotiations”…

I found this out one day by accident. The boss had in the past given me a load of crap why I was not even worth an inflation rise let alone one for the effective promotion…

But some one in HR had ‘tipped me the wink” a couple of weeks earlier that others were getting 10% above inflation. So after a chat with a colleague I’d gone out and got another job offer that was the equivalent of a 40% rise. And I’d already accepted, unbeknown to my then boss.

In the evaluation meeting the boss pulled the “company is in difficult times etc etc and how very luck I was to get the very very small raise he had, had to fight senior managment to get for me”. I knew it was compleate cattle scat and just took the payrise letter and did not say very much. The following day I de-documented all my work and changed names in code and started spaghettifying not just code but other work so it was way to conveluted to get your head into. At the end of the week I handed in my notice not to my boss but to HR after my boss had gone early, and I’d requested I be payed for my outstanding leave entitlement. Come monday mid morning he opened his mail and got a note in the internal mail about me having handed in my notice… And it was then that he draged me into his office and stupidly demanded to know what game I thought I was playing in effect accusibg me of trying to “shake him down”… and he stupidly got the hump after I simply reiterated the facts, he made some comment about he would see if he could get me more money, and he later came back with a couple of percent more. I told him “to little to late” and that, “that boat had sailed” and I was going for a long holiday on “my boat”[1] for a month down into the Mediterranean to relax and sort things out. He was very very anoyed and told me I’d have to do a handover and document in what he incorrectly thought would be a months notice…

Later that day the afternoon internal mail got to peoples desk in trays and quite supprisingly some got photocopies of the entire depts pay rises, how I realy don’t know… Anyway he got to hear about it and assumed I was behind it and made the old “you’ll never work again” style argument in front of witnesses… I was advised to put a formal complaint by HR, for harrisment… They decided it was best they let me go home on full pay and owed holiday pay etc rather than work my notice.

A salutary lesson about “Knowledge is power” if you are prepared to use it.

I was later told[2] that after I left the project I had been running went compleatly sideways then crashed and burned and most thought I’d done well to get out when I did.

A few months after that apparently the boss took –enforced– early retirment, and the last I’d heard of him was in 98 apparently he’d set himself up as a “consultant” but was not bringing the money in as it were. The skill set he had, was suffering a “market turn down”. However others of us had hit the Y2K trail and were pulling in decent money. I jumped into a new proffession and full time job about mid 99 as I could see a “night of the long knives” comming up as accountants and their bosses would want to get their Y2K spend back. As it turned out the night of the long knives was a long time comming and 9/11 kept the money rolling in for those who were flexible enough to have dodged the IR35 rules that were shockingly applied retrospectively. It was the first Finacial downturn / recession in 2003 and Crisis four years after that, which finally stopped the ball rolling for “the independent contractors” in the UK.

It’s the IR35 rules that the likes of major Corps are exploiting on the “zero hours” and “gig economy” jobs like Uber-Eats. Basically they put in the workers contract a clause that says they have to be able to provide a relief / replacment and the contract will be terminated if they don’t. That way they are an “agent” not a “worker” so have no legal protections. This in turn has created a sub culture of people actually acting like an agent, or more correctly a “gang master”.

Which brings us back to a point @JonKnowsNothing made a couple of days back.

[1] I used to have a largish sail boat that I used to race and cruise with. One or two work colleagues had been out on it for a day sail or just to crew and I’d alowed one or two to borrow it for a weekend or week to “get away”. And they had stuck photos up on the company notice board of their happy and sometimes green faces. So it was well known that “He’s got his own yacht”…

[2] Oh and the company I went to work for, I’d got the job through one of my old work colleagues who’s brother worked there and needed a good engineer so it was a “good fit”. Apparently my old colleagues got to find out how much I was earning at my new job very quickly. I found this out one night through some of them at a Stag Night a month or so later. For the wedding of the girl from HR who was marrying the guy who sat at the desk next to mine. They were borrowing my boat for their honeymoon –which is why I’d sailed it down to the med on a “shake down”– as they were saving up for a house deposit. Apparently staff turnover in the old Dept had gone up dramatically as well.

Winter November 24, 2023 1:16 PM


The following day I de-documented all my work and changed names in code and started …

Clive, is this really you?[1]

Because, this is immoral behavior, and if you had done such a thing, you would know better than write it in a comment.

[1] Or AI impersonation?

JonKnowsNothing November 24, 2023 2:23 PM

@Winter, @Clive, All

re: Deaths of Despair v Metabolic, Cardiac diseases, Obesity

I think the point I was trying to make but clearly missed, was that metabolic, cardiac and obesity are their own separate categories, and deaths related or attributed to these conditions are counted separately and are not included in the numbers for Deaths of Despair.

It is based on what is entered on the Death Certificate. Whatever the MD chooses to put down in those boxes is what gets counted by the CDC and others. The accuracy of Cause of Death is variable and in the USA, there is nearly no follow up to determine the real cause of death or contributing factors. Autopsies are common only on TV. Getting one done privately is nearly impossible. Even if the rule of thumb states an autopsy should be done, often times it does not happen.

  • Kaiser Permanente HMO California, rule: An autopsy is optional only if the person dies while in hospital. If they are discharged to care or rehab and die at those facilities, Kaiser is not obligated to do an autopsy. It is also at the discretion of the MDs at the facility and even if the person dies while in hospital, the request may be declined. There is a special procedure and forms you have to fill out to request an autopsy.

The exception is if you have a legal case where the judgement maybe a large dollar amount and you are able to engage that law firm in a timely manner. In California, Kaiser Permanente HMO has a mandatory arbitration clause as part of their membership requirement; they also have a hard cap of $250,000 USD on any judgement outcome. No legal firm in California will take a case against Kaiser Permanente in California on a “contingency fee” basis (pay after the judgement) as the cost of mounting a case with medical and legal experts, court filings and medical evaluations, toxicology tests etc will far exceed the dollar value outcome.

So everyone just makes do with whatever is printed on the Death Certificate.

note: Cause of Death in hospice care is listed by the medical condition on acceptance to that program, not on the physical end of life processes.

JonKnowsNothing November 24, 2023 2:37 PM

@Winter, @Clive, All

re: de-documented all work and changed names in code

This is not all that uncommon in Silicon Valley, due to potential fights over intellectual property, patents and use rights.

Most of us Rats on the Ship, don’t know the ins and outs of where or how software is acquired for a project. Generally something like this happens:

  • We need a tool to do XYZ. Somebody: Company N, O, P, E all have such tools

So NOPE code lands on the desk, with directions to strip identifiable marks, alter code names to “match our naming convention”, selectively remove “unneeded or unnecessary” code segments. Verify the sanitized code will compile; if possible verify the sanitized code passes a sanity-check.

It also depends on the terms of your employment contract about code ownership. The base rule is anything you create working for Big Tech belongs to Big Tech. It’s safe to presume that any code landing on your desk is Big Tech Code and not some propriety code even if it has propriety marks in it. Companies do buy all and full software rights.

Rats is Rats.

Clive Robinson November 24, 2023 4:51 PM

@ Winter,

Re : De documentation are spaghettification.

“Clive, is this really you?[1]”

Yup it’s me.

I did bot say that the code stopped running or could not be built.

It was just not easy to read…

For instance in the original source a variable of


Is more or less clear. But a change to


In no way changes the functionality of the code. But it makes the redability that little bit harder.

You can make it even less readable by turning it into a local pointer to an entry in a ptr table that gets populated in the start up code. Also consider the likes of Stack Randomisation for security and the many ways you can do it.

This is actually not unusuall, especially if you write as I often do “stack based code” which means the code is that much closer to being fully reentrant and will happily run on 16bit or less microcontrolers that at best have no way to relocate code that gets interpreted.

Back in the 80’s and 90’s in the UK reentrant code that was reusable was the rage, and there were no laws that said the code you wrote had to be “poetry in execution”. Stack based code where you do your own memory managment can be increadably benifitial when things have to be ROM based and you have no MMU to obfiscate code position.

Worse it was the start of the craze for “overloading” so yeh code was less clear than that pit of snakes in a Harrison Ford Movie.

Have a look at the original C++ code it was actually C code using tables of structures hidden behind macros and the like…

As for documentation, then as now documentation was fairly optional…

Also the fun thing the UK copyright to the source code that produced the run time code that went into the compiler, belonged to an employer, the documentation and even macros and libraries that went into a pre-processor and got linked etc did not, the copyright on those belonged to the person who wrote the code behind them.

Winter November 24, 2023 4:58 PM


But it makes the redability that little bit harder.

Obfuscated C code! I know that

Like this one:

Jessie Jose November 25, 2023 1:27 PM

Managing the complexities of my coursework reminds me a little bit of the intricate details of AI surveillance that are covered in the article. I understand the constant observation as a student—it’s like an AI watching everything you do. In the face of obstacles, Edubirdie serves as my academic ally, making sure I understand the complex subjects’ algorithms. It’s similar to having a study partner who speaks the language of achievement. I’m a scholar with Edubirdie, unlocking the secrets of academia one skillfully written thesis at a time. I’m not just a student.

JonKnowsNothing November 25, 2023 2:05 PM

@Clive, @Winter, All

An interesting MSM article about the composition of OpenAI workforce and the implications for AI in general. AI development is heavily leaning on male perspectives; AI datasets are predominately male views, and repositories of historical data mined by AI for their training sets is male oriented history.

  • 702 (out of 750) employees who signed the letter demanding Altman’s reinstatement more than 75% were men [526 m 175 f ]
  • OpenAI’s newly established board of directors is now made up exclusively of white men
  • Generative AI (GAI) relies on processing vast datasets of text, images and video, all of which have featured overwhelmingly more men than women…
  • [YTD 2023] this year men have been quoted 3.7 times more frequently than women in news about AI in English-speaking nations
  • Leslie McIntosh, vice-president of research integrity at Digital Science, says: “If your perspective is not reported, you are not in the story. GAI takes those historical texts and is building and projecting our future. So where women’s missing voices were once crevices, they have now become large gaps.”

An interesting form of bias, not the one I normally associate with AI but certainly one I was not paying attention to.

Best Quote @SAL

  • “He just takes up a lot of oxygen.”


HAIL Warning

ht tps://www.theguardia n.c om/global-development/2023/nov/25/where-are-godmothers-of-ai-womens-voices-not-heard-in-tech-sam-altman-openai

  • Where are all the ‘godmothers’ of AI? Women’s voices are not being heard
  • Amid the coverage of Sam Altman returning to the helm of OpenAI, women are being written out of the future of AI

Winter November 25, 2023 3:01 PM


So where women’s missing voices were once crevices, they have now become large gaps.”

The internet is male dominated. Not least the English speaking part. But the same has been said of history and literature. For all of history and literature.

AI is a mirror, not a window obln the world.

Clive Robinson November 25, 2023 4:08 PM

@ JonKnowsNothing, Winter, ALL,

Re : Autism counts in ICT thus AI.

“AI datasets are predominately male views, and repositories of historical data mined by AI for their training sets is male oriented history.”

This is to be expected for many reasons but three you might want to consider,

1, Historically women were discouraged from STEM especially TE (something I’ve fought against all my working life. As I used to sarcasticaly say women were taught that the epitome of female achivment was the perfectly cooked backbeans on toast for her childrens tea).

2, The “voices” are not ephemeral and thus only current, they are doccuments that accumulate. Thus if every male “voice” stopped parity of comment in the corpus would be something like a half to full decade away at best.

3, In STEM high functioning autism goes a very long way in espectially in TE. Depending on who you ask there are four to five times as many males on the high functioning autism spectrum than there are females. Personally I think the difference is less. Because autism is a “social communications” personality disorder, the presented male symptoms, unfortunately further drive women out of the more tech heavy side of TE.

Have a look at “Gamer Gate” and “Cancel Culture” in TE and what was once called “Knuckle Dragers v. Pinko Hairs” later “SJW Doxing/bombing/gaslighting” conflicts and you actually see two sides of the same problem fighting each other.

Shockingly for some when told it shows the behavioural sex divide in high functioning autism spectrum presentation. In part because women learn to “mask more” than men do. In short the level of high functioning autism spectrum women is probably very much higher than currently diagnosed because the diagnostic methods are aimed more at male presentation charecteristics and women very much have to mask more than men just to be even on the edge of Neuro Typical female group dynamics.

Masking is something that is increadibly stressful, because in effect you have to lead a double life without having time to “decompress” and recover. Thus inability to perform at a high functioning task is degraded at best and mental health issues abound.

One of the things about “lockdown” that has scared politicians is that the workforce apparently are not going back to work. Whilst Long Covid is very real, there are significant mental health issues that have been drawn out by the issolationism that lockdown caused.

Whilst those “who need to live in other peoples heads” –micro managers etc– suffered, less recognised is that those “who need to live inside their own heads”, felt freedom like they had not felt it before. That is during lockdown they mostly did not need to mask and they very much decompressed and as a result realised they did not want to go back to the sheer fear and tourture of leading double lives fealing effectively like “enemy agents living behind the lines” in the country of the Neuro Typicals.

Keep your eyes on the scientific literature to see this start to get documented and the political battles it will cause… As the neo-con types seek to “dominate and suppress” by victimisation and blaiming, as you would expect from “strong-men” types with their significant personallity defects that are currently better documented due to their very significant harms and risks to others. It’s why the “Drive them back to work” right-wing dictitorial attitude is extrodinarily detrimental to society espexially for the vast majority who sit between the two ends of the bell curve.

Leave a comment


Allowed HTML <a href="URL"> • <em> <cite> <i> • <strong> <b> • <sub> <sup> • <ul> <ol> <li> • <blockquote> <pre> Markdown Extra syntax via

Sidebar photo of Bruce Schneier by Joe MacInnis.