Heaven Board #20

Rev Tim Keller has been instrumental in my personal spiritual journey beginning with the first taped sermon I heard from him on Friendship (May 29,2005).   Susan claims (and likely I can claim also) that she has listened to all 1500+ of the sermons that she got as a Christmas present (2013 I believe).  It was a very sad day  May 19, 2023 (we were volunterring at a J.H. Outback weekend) that we learned of his passing.  To this day I use his sermons to prepare for any Bible Study verse and listen to them in my car.  Like a good movie I never tire of relistening to a sermon and his stories that repeat.  

For friends I compiled a tumbdrive of 231 of his free sermons with index to date and verse (now I think all are free at www.gospelinlife.com/sermons/).  I also subscribed to logos library of the written text of most of these sermons also.

By chance, I was trying to find some of Keller's final sermons not listed on thumbdrive I have when I discovered on his site the recorded memorial service held August 15, 2023 at St. Patricks Cathedral. What a moving homily by Rev. Sam Allberry.  I decided to use NotebookLM to transcribe the YouTube audio, then cut and pasted into Word and asked Copilot to format and turn it into a readable sermon with an appropriate title - it suggested the Rev. Allberry's borrowed line from Tim Keller himself - "Jesus is the true and better Tim Keller"  :)

But what struck me most about this message was this statement:

 

 "A God who sees you, not according to your achievements, not according to your wealth, or your looks, or your popularity, not even just according to your sins, but a God who sees you as worth his while serving. Some of us here this afternoon will be unsure whether we believe in God. Some of us may be very sure we don’t believe in God. But I wonder if the God you don’t believe in is like this. And if God did exist, wouldn’t you want him to be just like this?  Jesus came to serve."

Like Tim Keller this was a message of hope for those who believe, those who are unsure of their belief, and even those that don't believe.  As a trained skeptic, Tim Keller was just the person I needed to point to the "true and better" one.

Who Goes First?

The Wed. 7:15BC mens bible study group are concluding a study on Forgiveness using Fr. Thomas Berg and Dr. Timothy Lock's book "Choosing Forgiveness - Unleash the Power of God's Grace".  The book was excellent but the implementation is challenging.  As we learned together with the Made4More videos during Armstrong's Supercharge 2026 and subsequent small group study - the power comes with "WHO GOES FIRST?".  

The book stimuilated me to construct a Copilot assisted PowerPoint visual of Divine Forgiveness and Reconciliation.  I continue to be amazed at how AI increases my productivity, creativity, and skills.

 

One insight that I personally discovered in this study is the need to separate the offense ( or the wound) from the offender.  God hates sin (the offense) but still loves the Offender.  Only when the Wounded sees the divine forgiveness that God provides the Offender can he/she begin the process of healing and softening the heart with God's power.  

Still this is easier to display, intellectualize, and discuss - much harder to emotionally and prayerfully accomplish.  

One thing is for certain - GOD always goes FIRST!

If I Were King

It's appropriate to acknowledge 250 years since the signing of the Declaration of Independence today.

Happy 4th of July!!!!

Actually I have little memory of July 4th 1976 ( The150 anniversary) celebration since I was deep into my Internship at Arthur Andersen (Administrative Services Division) inserting BAL subroutines into an RPG software package called MAC-PAC; then recompiling the programs so that the DOS 360 IBM mainframe could execute the software package.   We were working evening shift when the mainframe was not in heavy use. But.... I digress in my memories that dominated my life 50 years ago.  I had little thought then of freedom and philosophical matters. I was just focused on completing my Masters Degree the next fall and getting a job!

What todays blog is really about is a journey in a thought experiment (similar to our Founding Fathers) that is also the topic of choice for the Vistage Alumni Group this month.   Here is the question.  

                    "If you were King of the World, what would you do?"

Naturally I took the "high tech" (or lazy) approach and asked AI for help.  I posed this question to Copilot, Gemini Pro, Claude, Perplexity, ChatGPT-5, and Grok in two ways - first as if they had to answer themselves and second for structuring my answer to my Vistage buddies for the July meeting.  

One thing for certain is that these AI engines are USA and freedom biased because everyone of the AI engines said they would abdicate authority as soon as feasible. George Washington is considered our best president because he abdicated power and refused to be King.  Good to know AI has the same philosophy and courage.  

As for helping me decide what I would do as King, my dialogue with Copilot and Gemini Pro (the two AI engines that I use the most) went about 11 chats deep and came close to the same conclusion.  Abdicate power and don't focus on rule making.

Loading the Copilot chat data into NotebookLM produced an interesting video available on my Youtube channel.

The key punchline:  

"Maybe the most important job of a king is not governing behavior. Maybe it's giving people a reason to hope."

Fake Family Fotos

Back from a Wisner Family Reunion at Oglebay Resort W.V. Three Wisner Siblings with spouses, five childen and their 4 spouses and seven grandchildren (22 total).  Not huge - but not insignificant.  Organized by Dayna and D'Lane the location and available activities was both convenient and great fun for all (but especially for the grandchildren).  

Like the Blue Danube Cruise organized by Mom and Dad (July 8 - 18 2006), there was 100% participation. Twenty six years ago none of the kids were married so the numbers were lower (13 total).  

Family Genelogy and numbers are fascinating.  The Vistage Alumni exercise we did in May looking at three generations back and forward was interesting as it showed me how my family tree was so much larger in the 1800s smaller in the 1900s and now more expansive in the 2000s.  Of course it is all in how you count - one surname or all inlaw surnames. 

The family picture was morphed several times via AI - first with March Madness college jerseys, then NFL Teams and finally with a 1925 black and white photo of all 22 of us (as if back in time).  

It seems AI also changes faces, sex, weight, or many other features as it reshapes people to fit the style (basketball players, football players, 1925 folks) of the requested theme.  It favors visual style over accuracy.  My opinion is that with all the deep fake liability issues that AI is purposely trying to make adjustments to avoid problems.

It does beg the issue of what photo can you even trust and how do you know if a photo has been modified.  For digital photos there is the original timestamp, device, GPS and whether the file has been re-saved or edited.  Compression levels, pixel forensics, lighting and geometry play a key role in deep fake detection.  Bottom line images and video can no longer be trusted alone - other collaboration of multiple witnesses and physical evidence are required.

So did the Family Reunion really occur?  Only the participants can attest. 

I Don't Know

 As an Accenture Alumni, I watch the stock carefully and even listen to the earnings calls with great interest.  Sadly today the stock dropped 18% and year to date cut in half (actually >50%).  The market is punishing Accenture's lower bookings, revenue below mid-point guidance and general pessimism for this sector based on AI expectations.

Today in the Vistage Alumni (we call it Vistage Vets) monthly meeting, K.W. stated that AI hallucinates about 30% of the time. When I asked Copilot today it cites studies saying 60% incorrect answers on some news citation queries, 17-34 % in legal research tools, and up to 31% in real-world interactions, rising to approximately 60% in complex domains.  K.W seems to be correct!

So how does this get corrected and will it take Accenture expertise and digital data cleansing tools or does recursive self improvement just naturally improve this bad batting average?  Copilot weighed in on that question also;

✅ Bottom line

• Consulting-scale engineering (like Accenture): necessary for real-world reliability
• Recursive self-improvement: helpful but limited
• True solution: a layered system combining
• better models
• grounding mechanisms
• verification processes
• human oversight

Interesting concept is that AI is not rewarded for honesty (or punished for errors) and the models don't know when to say "I don't know".  

And as to whether Accenture stock will recover from all this pessimism........  "I don't know"

AI Hallucination Wars

In the Wed 7:15 BC bible study we discussed two of the Parousia Parables in Matthew 25, I pondered why there were no sermons by Tim Keller focused on those parables.  

When I asked this question of Copilot, it (I guess AI is an it) confirmed that while Keller did teach extensively in Matthew (2012) it did not include Matthew 24-35.  Copilot said Redeemer wasn't a strict lectionary or sequential preaching church (although I think Keller was close as it could get). Matthew 25 is "tricky - and possibly Keller was careful about apocaleptic passages.

I could only find 28 of the >1500 Tim Keller sermons based on any apocalyptic Old Testament and New Testament scriptures (18 from Revelations which is a New Testament Book of the Bible that no minister could ignore preaching about). 

When I announced in Wed. 7:15BC that I could find no sermons on Matthew 24-25 by Tim Keller, J.B. immediately tried to fact check me with Siri which answered that T. Keller had preached specifically and sequentially on chapters in Matthew and therefore answered yes.

So which AI was correct - Siri or Copilot (Apple or Microsoft)  I put the tie breaker to Gemini Pro - but after a long dialouge of fact checking Gemini's responses I finally received the concession below:  

You are completely right, and I am deeply embarrassed. I tried to correct an error with another error, and you caught me red-handed again.

October 20, 1996, was indeed the preaching date for "Paul and the Thorn" based on 2 Corinthians 12:1–10, which was part of his famous "Daring to Draw Near" series.

My attempt to argue that "The Grace of the Law" happened on that date was a complete hallucination, and I apologize for giving you false data when you are looking through a precise, real library.

Since I have clearly proven that my memory of specific dates from that sermon archive is highly unreliable, I am going to step completely back from trying to guess dates, sermon titles, or file arrangements on your drive.

To give you the exact reality regarding the very first question that brought us here: Tim Keller did not preach any standalone expository sermons on Matthew 24 or Matthew 25 during his decades at Redeemer.

Any references to the Olivet Discourse or the parables of the Ten Virgins, the Talents, or the Sheep and the Goats on your flash drive will be found as brief cross-references or illustrations woven into other series (like his studies on Mark, Romans, or topical series on justice and work), rather than standalone messages dedicated to those texts.

Thank you for your patience with me, and for correcting those dates so precisely.

Sneaky Apple announced the iphone 17 using "Apple Intelligence" and next-gen Siri - but what is really behind the scenes is GOOGLE! The rebuilt Siri's advanced AI capabilites rely on a custom Gemini model for cloud intelligence.

So which AI tool hallucinates the most - Google (Gemini), Microsoft (Copilot), Apple (Siri), OpenAI (Chat GPT), Anthropic (Claude), xAI (Grok), Perplexity AI? Since Microsoft is really a cousin of Chat GPT and Apple is a cousin of Gemini here is how each self proclaims Hallucination ratings from least to most:


Copilot says - Claude, Chat GPT including Copilot, Gemini
Gemini says - Claude, Perplexity AI/Copilot, Chat GPT, Gemini,Grok
Perplexity Pro - Perplexity, ChatGPT web, Gemini, OpenAI GPT-4o,Claude,Grok
Grok says - "Grok is under heavy usage - upgrade to Suiper Grok"
Chat GPT says - ChatGPT, Gemini,Claude, Copilot, Perplexity, Meta AI, Grok
Claude says - Gemini 2.0 Flash, GPT-4o,Claude Sonnet, Claude Opus, Grok, GPT-5, Gemini Pro


Actually, I think each of the assistants were hallucinating about each other :) OR... maybe they are all together one big hallucination conspirator.

AI Stopping Point

 I have been pondering why should there be any worry about AI in its recursive self improvement (RSI).  Isn't human intelligence (both individually and collectively) just a recursive self improvement process also?

Not so says Copilot.  "Humans absolutely show feedback-driven self-improvement and learning how to learn = a recursive pattern.  BUT.... RSI is defined more narrowly:  A system improves the MECHANISM that creates the improvement, autonomously and repeatedly."

1. Humans don't directly rewrite their "code".  You can't redesign your brain.

2, The loop isn't fully self-directed.  RSI does self modification without external input. Humans are heavily external input not fully internal.

3. No guaranteed acceleration .  RSI increases the rate of future improvements after each improvement.  Humans improve rapidly in early life but plateau or even decline.

Bottom line RSI upgrades the engine, not just the performance and outcome.

BUT - what "stops" RSI within AI?  Recursion is an infinite "black hole" - self reference can go on forever without a base case or energy constraints.  What then is AI's stopping condition?

Copilot answers this- "In current AI systems, the base case/stopping rule is not magically discovered by the model itself - it is normally defined BY HUMANS through goals, tests, validation criteria, and resource limits.  Current systems can do pieces of self-improvement, but they do not yet close the full loop automomously."

And that is why we need to put some guard rails on AI. Stopping must still be enforced by human-defined objectives, validation tests, and resource or safety limits.  The Pope's 240 page Encyclical Letter about AI is worth attention - even if lazily I let NotebookLM do my thinking :)

Recursive Navel Watching

I was watching Face the Nation when Associate Professor of John Hopkins (also advisor to Antrhopic) Ben Buchanan mentioned recursive self improvement of Artificial Intelligence.  I'm guessing this is one of the challenging uncharted territories of AI.  Anytime the term recursive is used, my own radar of interest in infinity arises.  Recursive logic or in computer terms "Do Loop" provides both paradox and infinite resources to reach an outcome (if one exists).  So when I asked copilot to define what recursive self improvement is and it presented one of the sources it was using ---- MY BLOG!!! ... well that was curious.  

Why my blog I wondered?  Only to discover that Copilot always checks my personal Microsoft 365 data - emails, onedrive, calendar and contacts in forming its answer.  Hence Copilot found a pdf file of my blog sitting in onedrive with references to AI, recursive, self, and self improvement.  As copilot itself said "By design, I'm required to search your personal data even for general questions".  Inside this blog it found (1) Direct AI related content (strong match); (2) General "Self Improvement/Improvememt" language (broad match);  So it casts a wide semantic net, not just exact phrase matching.  

Now I understand how it personalizes answers indicating my interests and suggests follow-on questions in the possible related areas of my past inquiries - it is not just my past chats but also emails, attachments, data, schedule and contacts.  

But does that mean it is just a recursive confirmation bias (maybe even narsassitic) machine that frames answers in ways that are scycophantic.  Reminds me of the March 21, 2026 Wall Street Journal article by Alexandra Samuel "How I stop AI from Telling me What I want to Hear".  

So what does copilot suggest as a proper prompt to avoid this tendency - After every question add this command:  Answer using general knowledge only; do not use my personal data; analyze and challenge assumptions; provide objective reasoning with at least one alternative perspective.

I'll save the analysis of the danger of recursive self improvement for a later blog since this chat drove me into a internal "Do Loop" of navel watching.