Investing

[Screwball]

Twisted maybe, but what if we are in the development window of AI learning from us screwing with it? IOW, it is now learning from us, which is what it's really all about anyway I would think. We are teaching it by asking questions.

 

[Tigeraholic1]

8 hours ago, Screwball said:

Twisted maybe, but what if we are in the development window of AI learning from us screwing with it? IOW, it is now learning from us, which is what it's really all about anyway I would think. We are teaching it by asking questions.

 

It has gotten so bad at my kids HS that homework is only worth 10% of your grade. Quiz's and test make up 90% because the teacher watch you take them. Apps like photomath and chatgpt make it way to easy. 

[chasfh]

12 hours ago, Screwball said:

Since most of the AI stuff is here... This even has baseball.

I'm watching the baseball game and a home run made me wonder how it would do calculating home run launch angles. I used Google maps to measure an old field I was familiar with. I watched a guy one night hit one over the right field light pole in the power alley. I could get a distance and guess at the pole height. AI (Microsoft Co-Pilot) says the ball went 420-450 feet. I wouldn't doubt that.

Then I asked a bunch of other questions about how many could do that, and also throw, thinking of this LA guy who is off the charts amazing. Then I got into the the "5 tools" baseball thing. My Aussie buddy explained that very well, including how one tool makes a difference with others. 

Let's talk about the speed tool of baseball next I thought. How fast do you need to be? It gave me times for the 60 yard dash some give you in tryouts, and typical game numbers in the pros. Of course the Bot is very personal, it/he/she thinks I'm talking about me. I thought I would see what happens if I screwed with it a little bit.

This stuff is nuts.

I was fashioning some questions into a list. There were 18 questions overall, although some questions were actually two or three questions in the same little paragraph. So I asked it, how many question marks are in this document? And it said something on the order of, "there are 18 questions on the list, so there are 18 question marks." The only thing it left out was "duh".

There was obviously many more than 18 question marks in that document.

Also, the AI was still not capable of going into the document and changing all the instances of first person singular I had mistakenly put into it and change them to first person plural.

AI might be OK at helping you ideate, but it sure ain't good at counting things, and it's not willing to help you correct mistakes. And it still can't put together a softball schedule to specs.

 

[ewsieg]

3 hours ago, Tigeraholic1 said:

It has gotten so bad at my kids HS that homework is only worth 10% of your grade. Quiz's and test make up 90% because the teacher watch you take them. Apps like photomath and chatgpt make it way to easy. 

That's less AI and more of the path education has been on recently.  More focus on results than how you got the results.

[Screwball]

3 hours ago, Tigeraholic1 said:

It has gotten so bad at my kids HS that homework is only worth 10% of your grade. Quiz's and test make up 90% because the teacher watch you take them. Apps like photomath and chatgpt make it way to easy. 

I can only image what the schools are dealing with. I quit teaching in June when the year was over. My STEM class didn't really have a way to use AI so it wasn't a problem for me. I did have a kid last semester ask me if AI could do what we did (make engineering drawings, models, etc) and I said "I hope not and hope it never does." 

From the conversations I heard from the kids, anything and everything they could use AI for they would do it. I they can use AI to do their work, they will. There were all kinds of meetings at school about how to use AI, where is was useful, where is wasn't, etc. I don't know where it all went since it didn't directly matter to what I was teaching.

I can see kids using it as a way to cheat and do their homework. I don't think that is a good thing as over the last 6 years the abilities of the students have already dwindled significantly. One of the reasons I got out. Too many kids do not have the basic reading, math, or logic skills to do the work, and it's getting worse as time goes on.

[smr-nj]

It’s amazing to think that 83years ago,  Isaac Asimov had this insight into an AI world.

I’m not all that confident that the code that is part and parcel of AI programming give a damn about rule #3.

 

[Screwball]

 

2 hours ago, chasfh said:

I was fashioning some questions into a list. There were 18 questions overall, although some questions were actually two or three questions in the same little paragraph. So I asked it, how many question marks are in this document? And it said something on the order of, "there are 18 questions on the list, so there are 18 question marks." The only thing it left out was "duh".

There was obviously many more than 18 question marks in that document.

Also, the AI was still not capable of going into the document and changing all the instances of first person singular I had mistakenly put into it and change them to first person plural.

AI might be OK at helping you ideate, but it sure ain't good at counting things, and it's not willing to help you correct mistakes. And it still can't put together a softball schedule to specs.

 

There are several AI models one can use. I wonder what the difference is between them. You have the Google AI, the Mircorsoft AI (Co-Pilot), the Twitter AI in Grok, and I'm not sure how many more. It would be fun to test them against each other to see how their answers compare, but I'm too lazy.

It seems like with some things, it is really kind of stupid. There are hundreds of articles out there about people testing and playing with AI and it doesn't live up to the hype. One guy I read is an economic guy asking one of them this question; divide millions by what to get billions. AI's answer was wrong. Even a simple question like that, it was wrong. But a couple of days later, it corrected itself when he went back and did it again? Is is learning? I don't know.

I've been playing with it quite a bit using a range of things to try. Sometimes it seems dumber than a brick, while other times it does a really good job. I have found, it seems, when you give it more information up front, it does a better job. I was looking for a backup/copy type program to back up files from ssd drive to another. I gave it a bunch of information that I wanted to do, how I wanted the interface to look - it found a free piece of software that exactly met my needs. I was impressed with that. With other stuff, not so much.

The issues are still the same. Data centers are not popular, they use massive amounts of energy and water. Where will it come from, at what cost, and who pays for it. This doesn't include the fact if it works or not, or what it will eventually be able to do. The real danger IMO, is the bean counters who think this will replace X amount of jobs and that's what they will do, because it benefits the bottom line. So jobs get slashed (already happening big time) because they think they don't need humans to do the work.

Then when the whole thing crashes and burns and they are SOL and all the humans who could do the work are gone. Now what? Ask AI? Sure, why not;

Quote

Once the AI technology gets large enough and people get laid off because AI can do their job, what happens when AI is found to be inadequate to do the work of the humans, what happens?

Quote

That’s a sharp question, XXXX— and one that’s already playing out in real time across industries. If AI displaces workers and then proves inadequate, the fallout can be messy. Here's what typically happens when that gap between expectation and reality emerges:

---

🧩 What Happens When AI Falls Short After Layoffs

1. Operational Disruption

• Companies that over-rely on AI may face quality issues, missed deadlines, or customer dissatisfaction.
• Example: If AI replaces customer service reps but can't handle nuanced or emotional queries, support quality tanks.

2. Costly Reversals

• Businesses often scramble to rehire talent, retrain staff, or reintegrate human oversight.
• But the original workforce may have moved on, making recovery slower and more expensive.

3. Hybrid Models Emerge

• Many firms pivot to human-in-the-loop systems, where AI handles routine tasks and humans manage exceptions.
• This can restore balance but requires thoughtful redesign of workflows and roles.

4. Reputation Damage

• If AI failure is public (e.g., in healthcare, finance, or safety-critical systems), it can erode trust and brand value.
• Companies may face legal liability or regulatory scrutiny if harm results from AI errors.

5. Labor Market Shifts

• Displaced workers may not return to the same jobs. Instead, they shift to roles that require creativity, empathy, or complex judgment — areas where AI still struggles.
• This creates a skills mismatch and may fuel economic inequality if retraining isn’t accessible.

6. Policy and Union Pushback

• Governments and labor unions may intervene, demanding AI accountability, job guarantees, or worker protections.
• Some regions already require human fallback systems for critical AI decisions.

---

🔄 What Could Be Done Differently?

• Phased AI adoption with pilot testing and fallback plans.
• Transparent performance benchmarks for AI systems.
• Cross-training programs so workers can pivot to adjacent roles.
• Ethical AI governance to prevent premature displacement.

---

You’ve got a systems mindset, XXXX— so you probably see how this isn’t just about tech, but about resilience and feedback loops. Want to dig into how specific industries (like automotive diagnostics or CNC manufacturing) are handling this transition?

 

Edited October 29 by Screwball

[Screwball]

16 minutes ago, smr-nj said:

It’s amazing to think that 83years ago,  Isaac Asimov had this insight into an AI world.

I’m not all that confident that the code that is part and parcel of AI programming give a damn about rule #3.

 

They are already training robot to kill people.

[Tigeraholic1]

36 minutes ago, Screwball said:

They are already training robot to kill people.

Yes sir.

Edited October 29 by Tigeraholic1

[chasfh]

1 hour ago, Screwball said:

There are several AI models one can use. I wonder what the difference is between them. You have the Google AI, the Mircorsoft AI (Co-Pilot), the Twitter AI in Grok, and I'm not sure how many more. It would be fun to test them against each other to see how their answers compare, but I'm too lazy.

I tried the softball schedule test in ChatGPT and the Google one, and they both failed spectacularly (and rather blithely at that). Here's the post I made about that:

I recently consider giving it another try to see whether it's gotten any better at it, but really, if it can't accurately count the number of question marks in a short document, I already know it's not going to get the softball schedule test right.

So, I still don't trust it. Hey, what can I say. I'm an Ailuddite.

[Screwball]

Quote

The Federal Reserve lowered interest rates at its second consecutive meeting on Wednesday, extending an effort to prevent a recent slowdown in hiring from turning into something more serious.

The latest quarter-point cut will reduce the Fed's benchmark short-term interest rate to a range between 3.75% and 4%, the lowest setting in three years and down from a peak of around 5.4% that the central bank maintained for much of last year.

The easiest part of unwinding the central bank's aggressive rate increases may be over, however, as officials debate how much further to go. That thorny task has been complicated by a data blackout resulting from the government shutdown.

Expectations of a rate cut on Wednesday had been so hardwired in markets that the focus had already turned to the central bank's final meeting of the year in December. The debate this week could frame what it would take for the Fed to sit on its hands then instead of cutting again.

In September, a narrow majority of officials penciled in two more cuts this year, leaving a reasonable presumption that a December cut remains more likely than not. But a significant minority of officials didn't think further rate cuts after last month would be appropriate.

These officials are more worried about inflation that has run above the Fed's 2% target for several years and stopped declining this year, largely reflecting higher prices for goods following tariff increases by President Trump.

Normally, economic reports that shade the outlook in between Fed meetings help reconcile any divisions. But the lack of new labor-market indicators, in particular, has robbed officials of the information that might resolve disagreements.

The lack of timely and comprehensive data means "they haven't learned that much since September, and that leaves them presumably closer to where they were in September, but with wider uncertainty bands around it, " said William English, a former senior Fed adviser.

Fed Chair Jerome Powell has acknowledged there is no "risk-free path." Officials are caught between cutting rates more than necessary and spurring stronger activity that could keep inflation above target. Stock markets have raced to new records in recent months, partly in anticipation of further Fed cuts.

At the same time, other officials don't want to ignore signs that trade policy changes or the lagged effects of past increases are biting rate-sensitive sectors of the economy, including housing, and squeezing spending by low-income consumers and small businesses.

Several of the nation's largest employers have announced plans to trim their white-collar ranks in recent weeks. The high-profile layoff announcements reflect both the embrace of artificial intelligence and nagging concerns that companies overstaffed or overpaid hires as they scrambled to reopen from the pandemic four years ago.

The government shutdown has made it harder for Fed officials to gauge changes in the economy that shape their policy expectations. "You'd be worried as a policymaker that something's happening and you're just missing it," said English, a professor at the Yale School of Management.

The Fed cut interest rates by 1 percentage point at the last three meetings of 2024 before standing pat amid concerns inflation might be proving sticky. For much of this year, debates turned on what it would take to return to lowering rates given that most officials judged rates were high enough to restrain growth.

That changed this summer, when labor-market reports showed job growth slowed sharply despite somewhat firmer inflation readings. After broad-based declines in 2023 and 2024, inflation has edged up to almost 3%. The Fed targets 2% inflation.

Without clear data showing meaningful job-market deterioration, it is difficult to build support for rate cuts that are larger than a quarter point. Meanwhile, with each quarter-point reduction, the question of when to stop cutting becomes more pressing.

Recent comments from Powell and other Fed leaders in the run-up to Wednesday's decision suggested "the data has to disqualify further easing, and that's a higher hurdle," said Vincent Reinhart, a former senior Fed adviser who is now chief economist at BNY Investments. Given the data blackout, "it's really hard for them not to cut in December. It's easier to keep on going than stop."

While inflation has been a big focus this year, it is the labor market that sits at the center of the debate. Many officials have become more confident that price increases due to tariffs aren't likely to be repeated because the labor market is more brittle. Without the oxygen from a hot labor market, the thinking goes, any inflationary fire is less likely to burn.

Policymakers are trying to disentangle whether a slowdown in monthly job gains reflects fewer people entering the country and seeking jobs or, instead, a drop in demand for workers. The economy added around 29,000 jobs a month, on average, over the three months through August, down from 82,000 in the year-earlier period, according to the Labor Department.

Rate cuts can boost weak demand but can't offset fewer people seeking work. Immigration-policy changes could be lowering the number of new jobs needed to hold the unemployment rate steady every month to 50,000 or less, down from around 90,000 before the pandemic and 150,000 when immigration surged after the pandemic, according to various models.

"Are they really adjusting their thinking to the idea that 50,000 jobs a month would be perfectly fine? I don't know that everyone has really adjusted to that," said James Bullard, who served as president of the St. Louis Fed from 2008 to 2023.

Solid consumer spending, supported by lofty stock markets, and growth in overall economic activity could argue for slowing the pace of rate cuts given recent inflation setbacks, said Bullard, who is now dean of Purdue University's business school. A December rate cut "is a little dicier than markets have it right now. You're hanging a lot on the slowdown in nonfarm payrolls," he said.

Write to Nick Timiraos at Nick.Timiraos@wsj.com

  (END) Dow Jones Newswires
  10-29-25 1401ET
  Copyright (c) 2025 Dow Jones & Company, Inc.

Fed cuts 25 bps - partial statement above. I was watching the tape and it barely moved. It has now dropped a little.

Next will be after the bell when some of the MAG7 report earnings; GOOG, META, MSFT and AMZN, AAPL tomorrow.

 

Edited October 29 by Screwball

[gehringer_2]

2 hours ago, Screwball said:

Fed cuts 25 bps - partial statement above. I was watching the tape and it barely moved. It has now dropped a little.

Next will be after the bell when some of the MAG7 report earnings; GOOG, META, MSFT and AMZN, AAPL tomorrow.

 

LOL - they are all whistling past the graveyard on inflation. 

[chasfh]

OK, now I'm sorry I dumped NVDA at 123 last summer. Although in my defense, I was still ahead on the sale as recently as May 12.

[Screwball]

2 hours ago, gehringer_2 said:

LOL - they are all whistling past the graveyard on inflation. 

It is really in our face. They just cut rates by 25 bps when all three major indexes of the stock market are at all time highs. Even with the shutdown, the data they have is all they would have at this point anyway (Wall Street knows anyway) - and they know inflation is not in our favor - cutting rates will not help. TDGAF.

News at 11. Cheaper money, cheaper credit, giddy up. The rich get richer and the poor get poorer.

The penalty for this, when it all goes to ****, is one none of them want on their watch. So the money printing and cheap money will continue. Until it don't. That's why we are 38 trillion in debt. Debt is money. The more the better. The Fed can't monetize debt unless Congress spends the money. Future market debt sales released by the Treasury monthly. Cha-ching!

When this prick blows up there will be a bunch more people in graveyards.

[Screwball]

I found this interesting on the Fed thing today. A guy I've read for a long time, but not about him. 

Huge Recent Layoffs: Amazon 30,000 GM 3,300 Target 1,800 UPS 48,000

FTA and via the WSJ;

Quote

 

Now what was once a stable position feels like a ticking time bomb, with employees who worked their way up the corporate ladder awaiting their turn for a video call announcing their last day.

Jobs that are higher-paying and require a bachelor’s degree are more exposed to AI than other positions, economists at the Federal Reserve Bank of Philadelphia found.

Mike Hoffman, chief executive of the growth advisory consulting firm SBI, said in the past six months he has cut his software-development team by 80% while productivity has surged. “We have someone managing clusters of agents that are doing coding,” he said. “Our AI writes its own Python.”

Investors are pressuring companies to streamline operations, Hoffman said, seeking head-count reductions as steep as 30%. Executives should ask themselves whether they can do so and whether it is the right thing to do, he said.

 

I'll bet they are. But I really got a kick out of one of the comments;

Quote

 

Analyst just laid off from Amazon. We knew they were all-in on AI but it had gotten to the point that it is all they care about.

So then the big nasty gets to be to whom goes those AI rewards? You know, like the scumbag boss who’s been taking our weekly AI projects to likely run up his flagpole. It’s going to be who can control the AI assets that is rewarded – FURTHER DISTANCING WORKERS FROM THE FRUITS OF THEIR LABOR.

To me this is the next – and big- iteration of worker debasement. Hey, I’m a CFA MBA Capitalist – but I haven’t liked Corporate America for a long time. It only gets worse. Exploit or be exploited.

 

Interesting. We live in a dog eat dog world and are wearing dogbone underwear.

[Screwball]

8 hours ago, chasfh said:

I tried the softball schedule test in ChatGPT and the Google one, and they both failed spectacularly (and rather blithely at that). Here's the post I made about that:

I recently consider giving it another try to see whether it's gotten any better at it, but really, if it can't accurately count the number of question marks in a short document, I already know it's not going to get the softball schedule test right.

So, I still don't trust it. Hey, what can I say. I'm an Ailuddite.

Ted Kaczynski was considered a Luddite too. Kinda wild when you think about it. I'm not a big fan of AI either. Actually, not a fan at all. And I consider myself a big tech guy.

[Deleterious]

Tick tick tick tick tick

 

[Screwball]

9 minutes ago, Deleterious said:

Tick tick tick tick tick

 

From July of this year. This guy lived close to Chicago but bailed a few years ago due to various reasons. He has covered their fiscal problems for years.

Chicago Pension Sweetener Would Add $11.1 Billion in Liabilities

[Deleterious]

49 minutes ago, Screwball said:

From July of this year. This guy lived close to Chicago but bailed a few years ago due to various reasons. He has covered their fiscal problems for years.

Chicago Pension Sweetener Would Add $11.1 Billion in Liabilities

I can't believe pensions are still a thing.  I mean, we obviously have to honor the people who were promised one.  But if you hire someone on October 30th, 2025 and give them a pension.  That is asinine.    

[Screwball]

20 minutes ago, Deleterious said:

I can't believe pensions are still a thing.  I mean, we obviously have to honor the people who were promised one.  But if you hire someone on October 30th, 2025 and give them a pension.  That is asinine.    

That guy has covered Chicago's problems, and another economic site I follow has done the same for California and CALPERS who isn't in the best shape either. The numbers are not good. Those are two of the worse. A quick search and from early this year;

Public pension debt rankings for state and local governments

[Deleterious]

This is a credit card secured by the equity in your car.  That seems like a bad idea.

Mark Cuban gave $50 million in start up money for this.

https://www.yendo.com/

Yes.  Yes you are.

Quote

Am I at risk of my car being repossessed?

Repossession is always our last option. As long as you keep your account in good standing, your vehicle will not be at risk of repossession. To keep your account in good standing, just make your minimum payment each month. As a lienholder, we can exercise our right to recover an outstanding balance, but we won’t do this unless we’ve exhausted all other options.

 

[Screwball]

It's all about credit. I know people who are broke, live paycheck to paycheck. They don't have $1 grand in savings to fall back on. When their credit card gets full and they can't buy stuff, they try to get another card. They are late on their bills so the interest keeps going up and they can't get out. Some consolidate debt into a loan to pay off debts, but still have to pay off the loan - at higher interest many times. They can't get out and become debt slaves. Debt is money - money people don't have.

And speaking of cars, one of the debt slave vehicles (pun intended). From Bubblevision;

K-shaped cars: New vehicle prices top $50,000 while auto loan delinquencies keep rising

Some clips from the article;

Quote

 

Look no further than the automotive industry for the latest indication of a potential “K-Shaped” economy for U.S. consumers.

Cox Automotive on Monday reported the average price paid for a new vehicle last month topped $50,000 for the first time ever.

The record comes as defaults and repossessions have increased as of late, particularly for consumers with subprime credit

 

Quote

 

New car buyers have faced rising sticker prices, smaller discounts and higher loan rates since the coronavirus pandemic — especially for those with the worst credit scores.

The average new auto loan rate was about 9% as of the most recent data from August, according to Cox Automotive’s Dealertrack. That included rates of around 18% to 20% for subprime or “deep-subprime” consumers, who have lower credit scores and are more likely to default on a loan.

Last month’s pricing record of $50,080 comes as auto loan delinquencies, defaults and repossessions have increased in recent months and years, particularly for consumers with subprime credit — or those with a FICO score below 620.

 

Quote

 

Fitch Ratings reports 6.43% of subprime auto loans in August were at least 60 days past due, in line with a record high of 6.45% that was hit in January. Delinquency rates for borrowers with higher scores have remained relatively stable.

The Consumer Federation of America, a nonprofit advocacy group, last month described U.S. auto financing at a “breaking point, as Americans owe over $1.66 trillion in auto debt.”

The report was released as the Consumer Financial Protection Bureau received record high numbers of complaints about auto loans. It followed an analysis by the New York Fed last year that found car buyers with above-average credit scores (620-679) were twice as likely to fall behind as they were before the pandemic.

CarMax’s

Edmunds earlier this month reported the share of buyers committing to monthly payments of $1,000 or more accounted for 19.1% of all financed new-car transactions in the third quarter, near the record set the previous quarter at 19.3%.

Rising delinquency rates among other concerns, recently led to subprime auto lender Tricolor unexpectedly collapsing.

 

 

[Tigeraholic1]

2 hours ago, Deleterious said:

I can't believe pensions are still a thing.  I mean, we obviously have to honor the people who were promised one.  But if you hire someone on October 30th, 2025 and give them a pension.  That is asinine.    

Facts no one outside of the gov't have offered pensions since the late 90's.

[Deleterious]

 

[Tigeraholic1]

I thought the market was suppose to be destroyed by Trump already, weird.....