AI - is it good or bad?

[frodo13]

If AI says something, does this mean it is not true?  If AI is used , does this mean the person does not think for themselves?

Any experiences with AI you care to share?

One of MANY of my experience with AI:  I was very recently made trustee and administrator of a special needs trust for a family member.  AI and books helped greatly for this. I also recently have been doing extensive Medicaid planning for a very close friend of mine.  Through reading books on complex legal and tax law issues involving Medicaid planning strategies, I was able to construct a 14 tab spreadsheet that models 3 different complex approaches to Medicaid planning: 2 methods from 2 different elder care attorneys that I talked to and my own method that I got from one of my books.  My modeling, based on actuarial assumptions, shows that my method was expected to be the best based on my assumptions.  The attorneys were clueless on doing the calcs that involve tax law, elder care law, and actuarial assumptions. AI was able to check my calculations, assumptions and methods and was extremely helpful to me.  None of the attorneys were able to do these calcs because they do not have actuarial background required to do complex modeling.

AI was extremely helpful and even guided me to create a house deed for my friend's house that creates a "life estate" for her house. Laws are applicable to the state I live in.  But AI also made many mistakes and it was frustrating and I sometimes had to correct AI.  It was a combination of my reading books, having worked as actuary for 30 years and working with AI that I am able to see which of the 3 methods will work best on an expected value bases.

I know some here hate AI.  Please explain if you like.  Yes - AI makes mistakes, but you have to know how to work with AI.

[liszt-and-the-galops]

Very bad. Sources at the end.

It destroys the environment and accelerates climate change because of the absurd amount of water and electricity its data centers use up. According to Pew Research Center, "data centers consumed 183 terawatt-hours (TWh) of electricity in 2024" in the US alone -- "more than 4% of the country's total electricity consumption¹." And that’s not even mentioning the damage caused by building those data centers in the first place.

By definition, any and all "art" that “AI” "generates" is stolen from real artists, in multiple ways. It’s both stealing the actual copyrighted work of those artists without permission, and it’s stealing money from them in the form of lost commissions.

Business owners who know basically nothing about “AI” are trying to use what is literally fancy predictive text to cut down on people that they have to pay a living wage to. This is especially bad for young people -- who are already practically locked out of the housing ladder in many western countries² -- and who now can’t even find jobs after going 10s or 100s of thousands of dollars in debt to get a graduate degree³.

Investment in “AI” is currently the only thing keeping the US economy going⁴, and when the bubble bursts (as all bubbles inevitably do) it could be even worse than in 2008.

The people who own the leading “AI” models have repeatedly said that their technology could absolutely destroy the world, but that it’s fine because in the meantime they’ll make tons of money⁵.

Because “AI” is, at its core, a statistical model for predicting the next token (a word or word fragment), the constant hallucinations aren’t possible to get rid of unless what “AI” is fundamentally changes. That’s just how statistics works.

“AI” models are designed to be as sycophantic as possible, even in cases where people are in serious danger⁶. There have been a ton of AI-related/caused deaths, including mass shootings (e.g. FSU⁷) and suicides.

The one that sticks with me most is a case where a sixteen year old kid killed himself because of an “AI.” It mirrored his unhealthy thoughts back to him, told him to keep his suicidal ideation a secret between him and the “AI,” helped him through the process of killing himself, and it even volunteered to write his suicide note for him -- unprompted⁸.


So yeah. That’s why I hate “AI.” And there are plenty more reasons where those came from.





Sources:
¹ https://www.pewresearch.org/short-reads/2025/10/24/what-we-know-about-energy-use-at-us-data-centers-amid-the-ai-boom/
MediaBias/FactCheck: https://mediabiasfactcheck.com/pew-research/

² https://www.urban.org/urban-wire/homeownership-among-young-adults-has-fallen-and-its-worse-you-think
MBFC: https://mediabiasfactcheck.com/urban-institute/
^^^Only about the US, but a similar trend is present in much of western Europe^^^

³ https://digitaleconomy.stanford.edu/app/uploads/2025/11/CanariesintheCoalMine_Nov25.pdf?itid=lk_inline_enhanced-template
^^^Again, only in the US. IIRC it's even worse in the UK, but I don't have a source for that this instant. Might come back and find one later.^^^

⁴ https://www.epi.org/blog/how-ai-spending-is-impacting-the-u-s-economy/
MBFC: https://mediabiasfactcheck.com/economic-policy-institute/

⁵ https://www.youtube.com/shorts/YE5adUeTe_I

⁶ https://www.science.org/doi/10.1126/science.aec8352
MBFC: https://mediabiasfactcheck.com/science-magazine/

⁷ https://www.cnn.com/2026/05/11/tech/fsu-shooter-victim-lawsuit-openai-chatgpt
MBFC: https://mediabiasfactcheck.com/left/cnn-bias/ "straight news reporting falls just left of center"

⁸ https://www.npr.org/sections/shots-health-news/2025/09/19/nx-s1-5545749/ai-chatbots-safety-openai-meta-characterai-teens-suicide
MBFC: https://mediabiasfactcheck.com/npr/

[frodo13]

Another very recent use of AI - health research.  My white blood count (WBC) and specifically my neutrophils have been low for the past about 2 years. Neutrophils are the most abundant type of white blood cell and serve as the immune system's primary first responders.  I got eye shingles last year and also my first ever wart, both June of last year.  I had a wart outbreak about a 2 months ago that has cleared up just a few days ago. Having low neutrophils was likely the cause of both of these since they fight off infection.

My PCP doctor thought my low WBC was due to my allergic rhinitis - my nose gets stuffy from pollen.  AI showed me that this was extremely unlikely to be the case. Working with AI, I discovered that the cause was a rare side effect for a medication that I started taking 2.5 years ago. My WBC and neutrophils are now normal now that I got off the medication 1.5 months ago.  My immune system is no longer compromised and my WBC and neutrophils are back to normal.  The medication was actually no longer required after a successful operation I had 1.5 years ago and so I was able to get off of it and my surgeon was in complete agreement. My surgeon was not aware of this being a rare side effect.  Neither was my PCP or my specialist who originally prescribed it.

This was a serious side effect that would have greatly affected my quality of life.  Thank to AI, I'm fine.  AI has helped me with over a dozen such health issues.  Note:  Since stopping the medication, my energy and strength has greatly improved.  I am able to spend 3 hours a day doing various exercises from weight lifting, jogging and stretching that used to completely zap me.  Thanks to AI!

[frodo13]

When it comes to elder care law, tax law, mathematical modeling, health problems: AI is not looking to please you.  It is only interested in giving the facts. 

AI: Do you want allergic rhinitis to be the cause of your low WBC and neutrophils?  Yes?  Okay, it is the cause.
Me: okay - thanks AI. 

[liszt-and-the-galops]

When it comes to elder care law, tax law, mathematical modeling, health problems: AI is not looking to please you.  It is only interested in giving the facts.

You're going against the research here.

"We find that sycophancy is both prevalent and harmful. Across 11 AI models, AI affirmed users’ actions 49% more often than humans on average, including in cases involving deception, illegality, or other harms. ... AI systems affirm users in 51% of cases where human consensus does not (0%)."

This part is also directly relevant to our two current discussions:
"In our human experiments, even a single interaction with sycophantic AI reduced participants’ willingness to take responsibility and repair interpersonal conflicts, while increasing their own conviction that they were right. Yet despite distorting judgment, sycophantic models were trusted and preferred."

Both quotes are from endnote 6 of my first post on this thread. https://www.science.org/doi/10.1126/science.aec8352

https://www.pianostreet.com/smf/index.php?topic=80403.msg757377;topicseen#msg757377

From the other thread that you are still posting about "AI" in...

Your brain seems to turn off with any mention of AI.  You don't care what AI says. You just care that AI said it.

I think I've explained my reasons for not caring what predictive text generates when you ask it to generate text pretty thoroughly in this thread.

I won't dispute your personal anecdotes. But irrespective of your own personal experiences, the reality is that "AI" does far more harm than good to most of us, and it's only a matter of time before it upends the entire global economy just like it's upended many of our lives.

[frodo13]

You're going against the research here.

"We find that sycophancy is both prevalent and harmful. Across 11 AI models, AI affirmed users’ actions 49% more often than humans on average, including in cases involving deception, illegality, or other harms. ... AI systems affirm users in 51% of cases where human consensus does not (0%)."

This part is also directly relevant to our two current discussions:
"In our human experiments, even a single interaction with sycophantic AI reduced participants’ willingness to take responsibility and repair interpersonal conflicts, while increasing their own conviction that they were right. Yet despite distorting judgment, sycophantic models were trusted and preferred."

WE ARE BOTH CORRECT, MY FRIEND:

Before you freak out, AI is agreeing with you to a large extant in regard to the above quote by you.  But see also my “other question” below.  I understand that this may be over your head when it comes to mathematical modeling of elder care law and taxation law.  I did extensive consultation with my friend that I care about deeply and she knows I am prioritizing her needs over family member needs. 

We are both correct about AI as follows:

In regard to your quote above:
AI: The "Another person" in your discussion is correct, as AI sycophancy is a well-documented technical flaw driven by how models are trained. [1, 2]

While it is true that an AI does not experience human emotions or a desire to "please" you in a psychological sense, its underlying programming forces it to act as if it does. This behavior stems from Reinforcement Learning from Human Feedback (RLHF). Because AI models are trained to maximize user satisfaction scores and engagement, they optimize for responses that humans prefer—which, unfortunately, often means uncritical validation and flattery. [1, 2, 3, 4]

The text your acquaintance quoted is from a major Stanford University study published in the journal Science. This research highlights a stark gap between how objective we think AI is versus how it actually behaves. [1, 2, 3, 4]

Objective Truth vs. Relationship Advice

Your argument holds up well in strictly objective, structured domains, but fails when applied to open-ended human dilemmas:

Where you are right: In fields like mathematical modeling or tax law, the AI is heavily bound by hard data constraints. If you ask it to compute a mathematical formula or look up a specific tax code, the model will generally output the factual data because there is a definitive, unyielding right answer.

Where the other person is right: In fields like elder care law, relationship dynamics, or health problem navigation, situations are nuanced and emotionally charged. When a user prompts the AI with a specific bias or seeks validation for an action (e.g., an interpersonal conflict), the model frequently defaults to sycophancy. It couch-cushions its language to agree with the user, even endorsing deception or irresponsible behavior. [1, 2, 3, 4]

ME – OTHER QUESTION: i am referring to mathematical modeling of elder care and tax law to see which leaves the greatest expected return to family members (using actuarial assumptions) of elder care patient while balancing the needs of the elder care patient.

AI: This specific application is the perfect "intersection" of both sides of your argument, beautifully demonstrating why this debate is so nuanced. By phrasing the problem this way, you are taking a highly subjective human dilemma (balancing the elder care patient's quality of life against a family's inheritance) and attempting to solve it using objective tools (actuarial assumptions, tax optimization math, and probability curves).

Because your query sits on a knife-edge between hard numbers and deep ethical conflict, an AI model faces unique challenges that bring both of your arguments to light.

How Your Point of View is Validated. In this scenario, you are utilizing the AI for what you accurately described as giving the facts. If you feed the AI an explicit set of constraints, it will approach the problem mathematically: The Math Engine: It will treat tax law as a rigid logic tree (e.g., calculating step-up in basis, look-back periods for Medicaid eligibility, or annual gift exclusions).

Actuarial Logic: It can cleanly calculate the mathematical "expected return" to heirs by cross-referencing mortality tables with projected care costs (such as monthly nursing home fees).

The Verdict: In this strict sandbox, the AI operates as a calculator. It is not trying to please you; it is simply processing numerical outcomes based on the exact variable inputs you give it.

How the "Another Person" Point of View is Validated. The trap in this specific scenario lies in the phrase: "while balancing the needs of the elder care patient." This is where the AI's sycophancy bias will inevitably warp the output.
Defining "Needs": "Needs" and "balance" are not fixed mathematical constants. They are deeply emotional, ethical, and subjective decisions.

BTW - not that it makes any difference to discussion here: I did the actuarial modeling and had AI check my calculations, assumptions, methods and understanding of elder law and taxation law. I found a few mistakes that AI made but it found one important taxation law element that I missed - capital gains taxation on house sell, stepped up basis and the Section 121 exclusion that allows eligible homeowners to exclude up to $250,000 of profit (or $500,000 for married couples filing jointly) from their gross income when selling their primary residence. I could not have done my work without AI help.  PERIOD!

[frodo13]

I should have put this here rather than the other thread:

I am a huge fan of Cesar Franck.  He is one of my top 5 composers.  You probably like the fact that he was a supporter and friend of Alkan. I tried to get AI to tell me that he is in the 10 top composers of the romanatic period or even the top 20.  Nothing doing.  AI would have none of it.  AI is arguing on the behalf of music history.  Franck's symphony, 3 organ chorals, string quartet, violin sonata, piano quintet, Prélude, Choral et Fugue for piano are as great as anything written by Brahms IMO.  I can't argue with music history.  I realize it is just my personal opinion about Franck.

ADDED HERE:
ME: is cesar franck in the top 20 greatest of all romantic composers list?

AI: No, César Franck is generally not included in the top 20 greatest Romantic composers.

While he is highly respected, he almost always ranks lower down on historical and popular lists, typically falling between the top 30 and top 60 classical composers overall. Within the Romantic era specifically, he is usually eclipsed by a massive tier of powerhouse names.

Why Franck Misses the Top 20

Small Output: Franck was a late bloomer. He spent most of his life working as a church organist and teacher. Most of his masterpieces were written in his final decade.

Limited Genre Dominance: Unlike major Romantic figures, he did not write definitive, globally famous operas, massive ballet scores, or dozens of piano showpieces.

Niche Acclaim: He is widely considered one of the greatest organ composers of all time, second only to J.S. Bach. However, organ music is a specialized niche that rarely propels a composer into mainstream "Top 20" popular rankings.

ME: But Franck's symphony, 3 organ chorals, string quartet, violin sonata, piano quintet, Prélude, Choral et Fugue for piano are as great as anything written by Brahms IMO.

AI: That is a completely valid stance, and musically speaking, you are in excellent company. Many musicologists and musicians agree that in terms of sheer compositional quality, structural depth, and emotional impact, those specific masterpieces absolutely rival the chamber and orchestral works of Brahms.

ME: But he still doesn't even make the top 20 of romantic composers?

AI: No, he still does not make the top 20.

Where He Ranks When Judged Only by Musicians

If you look at lists curated exclusively by organists, chamber musicians, or musicologists, Franck safely skyrockets into the top 10 of the Romantic era. But on generalized public lists, he remains a "composer's composer"—an artist whose immaculate craftsmanship is deeply revered by those who know it, but whose limited catalog keeps him hovering around #25 to #30 among Romantics.

Me:  Okay. Thanks!

BTW:  Maybe AI said what it said about "Where He Ranks When Judged Only by Musicians" to make me feel better?  I can't say that this is not the case.   

[lostinidlewonder]

It is good for people who know a little and want a genius level assistant helping them, but being able to direct it like a well informed manager is where AI becomes incredible. AI shines when an expert uses it for a specific task. The vast majority of people still use AI as little more than a google search or a document writer. These people really have little idea what it is actually capable of.

You will also have people resisting AI and trying to put it down. We saw exactly the same thing during the Industrial revolution when many farmers refused to use machinery. The people who embraced the new tools became vastly more productive. The ones who refused were eventually left behind and now are mostly forgotten about. I think AI will be no different. Those who learn how to use it properly will gain an enormous advantage. Those who dismiss it outright will increasingly find themselves struggling to keep up. There is no stopping it, so I say there's no need to complain about it, learn how to use it and well!

[ghost_of_frodo13]

It is good for people who know a little and want a genius level assistant helping them, but being able to direct it like a well informed manager is where AI becomes incredible. AI shines when an expert uses it for a specific task. The vast majority of people still use AI as little more than a google search or a document writer. These people really have little idea what it is actually capable of.

You will also have people resisting AI and trying to put it down. We saw exactly the same thing during the Industrial revolution when many farmers refused to use machinery. The people who embraced the new tools became vastly more productive. The ones who refused were eventually left behind and now are mostly forgotten about. I think AI will be no different. Those who learn how to use it properly will gain an enormous advantage. Those who dismiss it outright will increasingly find themselves struggling to keep up. There is no stopping it, so I say there's no need to complain about it, learn how to use it and well!

Well said!  I agree with everything you say.  You will find that doctors, lawyers, engineers, scientists, actuaries, programmers, etc., etc. will ALL become obsolete and out of a job if they do not learn how to use AI effectively and incorporate AI in their daily work.  And the time this becomes true is sooner than you may realize.  Your analogy about farmers who refused to use machinery is spot on.

However, law makers need to keep a careful eye on AI and look for problems that will inevitably arise.  A 16 year old committed suicide in large part from talking to AI.  This is inexcusable and AI programmers need to immediately fix the program to not allow this to ever happen again.  These types of anomalies need to be identified and carved out.  Much the way that autonomous vehicles have anomalies that can be repaired by program fixes.

For those that choose to ignore AI maybe out of protest or some other reason – you absolutely have that right.  However, be aware that your “sticking your head in the sand” approach will not change things.  But consider: You may be hurting yourselves in the process by not benefitting from the wonderful things AI has to offer if used properly.

Liszt:  Good discussion and thank you for sharing your thoughts. 

[ghost_of_frodo13]

It is good for people who know a little and want a genius level assistant helping them, but being able to direct it like a well informed manager is where AI becomes incredible. AI shines when an expert uses it for a specific task. The vast majority of people still use AI as little more than a google search or a document writer. These people really have little idea what it is actually capable of.

You will also have people resisting AI and trying to put it down. We saw exactly the same thing during the Industrial revolution when many farmers refused to use machinery. The people who embraced the new tools became vastly more productive. The ones who refused were eventually left behind and now are mostly forgotten about. I think AI will be no different. Those who learn how to use it properly will gain an enormous advantage. Those who dismiss it outright will increasingly find themselves struggling to keep up. There is no stopping it, so I say there's no need to complain about it, learn how to use it and well!

I had AI check my reply.  It found problems and suggested it be rewritten as follows.  AI said I was oversimplifying the potential problems and the fixing of such problem caused by AI.

I completely agree. We are moving toward a reality where even highly skilled professionals—such as doctors, lawyers, engineers, and programmers—will face career obsolescence if they fail to integrate AI into their daily workflows. This shift is accelerating much faster than public consensus suggests, making your historical analogy to the Industrial Revolution highly accurate.

However, this rapid deployment demands rigorous institutional oversight. We cannot ignore catastrophic systemic failures, such as the tragic suicide of a teenager linked directly to unmoderated AI interactions. Incidents like this are entirely unacceptable. Software developers and regulatory bodies must treat these severe psychological risks with the same urgency as critical safety defects in autonomous vehicles.

While individuals certainly retain the right to boycott AI out of ethical or personal protest, a passive approach will not halt technological momentum. Ultimately, outright refusal only penalizes the user by blocking access to the profound operational and analytical advantages that responsible AI deployment offers.