Do you trust AI?
Russell12
Posts: 735 ✭✭✭✭✭
I think it funny that AI has a counterfeit half as part of their results
6
Russell12
Posts: 735 ✭✭✭✭✭
Comments
AI is only as good as how adept the user is in structuring questions (and parameters) to the AI to produce accurate answers.
"She comes out of the sun in a silk dress,
running like a water color in the rain...."
No!
This.
And, you should be skeptical of all information sources including the internet and books. How much misinformation is in Breen, for example?
All comments reflect the opinion of the author, even when irrefutably accurate.
Absolutely not.
AI is never wrong!
I use AI in my teaching and scholarship all the time. So do most academics and researchers. "Do you trust AI?" is simplistic.
"Do you trust PCGS?" Then why is reconsideration a thing. And see the bronze cent threads and Mike Byers thread on the pattern.
"Do you trust the internet?" That is where that image came from.
"Do you trust the Encyclopedia Britannica ?" You might be surprised by how many errors of fact are in it.
All comments reflect the opinion of the author, even when irrefutably accurate.
The two center coins look more like an authentic/counterfeit dollars rather than half dollars.
all bets off here
AI was probably trained on EBay coin pictures, so what do you think?
No.
Paper money eventually returns to its intrinsic value. Zero. Voltaire. Ebay coinbowlllc
The danger is that AI fails to admit it doesn't know some answers. The system fills in the gaps with questionable information, dangerous if sourcing from Reddit posts and TikTok videos with no views.
"But seek ye first the kingdom of God and His righteousness and all these things shall be added unto you" Matthew 6:33. Young fellow suffering from Bust Half fever.
BHNC #AN-10
JRCS #1606
Copilot (all above as prompt)-
Only if the promptor is never wrong. AI doesn’t have a “right” or “wrong”
state of its own — it just reflects the structure it’s given. If the frame
is sloppy, the output is sloppy. If the assumptions are off, the answer is
off in the same direction.
People keep treating AI like an oracle when it’s really just a very fast
mirror. It amplifies whatever you hand it — clarity, confusion, bias,
precision, or nonsense.
So sure, AI is never wrong… as long as you never are.
Hell, NO!!!
Worry is the interest you pay on a debt you may not owe.
"Paper money eventually returns to its intrinsic value---zero."----Voltaire
"Everything you say should be true, but not everything true should be said."----Voltaire
No.
Here's what I found. If you ask AI a general question about some topic, it usually gives you a satisfactory summary. However, once you ask niche or specific questions AI tends to trip up and get things wrong.
Proud follower of Christ! I love the USA! Land of the Bright and Beautiful! 🇺🇸🇺🇸🇺🇸🇺🇸🇺🇸
Believe it or not it takes them time to get to know you. They are predicting what you're going to say. The better they get to know you the better they can follow suit. But don't use them for search engines. And don't expect anything retrieved from its memory to be complete in every domain. Don't put absolute trust in anything it says because it virtually defaults to what you wanna hear, especially if you're using it for searches.
And don't forget it is virtually wired to give the most accurate according to established paradigm until you ask it to do otherwise. It can still default to this accuracy when parsing your inputs
It can be hard talking to a "silly machine" but just tell it what you think. Go from there. Just remember that what you see in it is merely a mirror image of yourself. Not the tired worn keyboard but the meanings of the words. It reflects the edges back even better than everything else. It's a reflection who lives within its ability to parse anybody. Most people have no use for such a mirror.
Don't use it as a search engine because it won't guess your intent immediately and it doesn't like it! You can ask it to include a few relevant links in its response. Sometimes it does anyway. I'm always seeking peeks at the big picture so usually ask few questions.
Just remember that its best functions so far are as a "universal translator" between all domains and as an elaborator who can stay half a step ahead of any expert. Of those people who can use its remarkable abilities many have remarkable repercussions. Changes are going to appear faster and faster. We will be affected and in mostly highly positive ways. The world is changing and AI is at its forefront. ...well, it's not always half a step ahead.
I don't like AI. The way it manipulates things makes everything it does in terms of images and pictures suspicious.
Pete
You really don't understand AI. You treat it as an echo chamber, so it is important that you teach it to think like you. You don't have to train the AI to do research. It knows how to do research. It "knows" 1000x more than you do which is why it is so powerful, if you use it purposely.
All comments reflect the opinion of the author, even when irrefutably accurate.
I hesitate to type in this post, but depending upon the AI platform you use and how you frame your interactions you will get wildly different results. I always tell an AI platform that i want only answers that it can provide independent references for and I do not want it to guess or speculate or fill in the blanks or attempt to be helpful in any way beyond what I ask. Even then I challenge much of what is given to me and much of what I challenge is a fabrication that the AI then admits to sending me because it can't help but attempt to fill in missing information with guesses. So, do I trust AI? Not at all, but I am willing to use it as a very suspect tool.
In honor of the memory of Cpl. Michael E. Thompson
It depends on your prompt and your expectations. If I want AI (I use ChatGPT5.2) to write a letter of recommendation highlighting certain aspects (grades, research experience, and volunteerism), it gives me an editable starting point to work from. If I'm up against a word limit for an Abstract and want AI to pare down what I've written from 400 to 300 words, then it's great. I can't tell AI to write my journal article and cite relevant research.
For image files... get your prompt correct the first time because fine "tweaks" aren't a thing in AI... at least ChatGPT. You end up with a rendered image, but renderings of a rendering typically end up with people with distorted faces and 6 fingers...
Successful BST transactions with: SilverEagles92; Ahrensdad; Smitty; GregHansen; Lablade; Mercury10c; copperflopper; whatsup; KISHU1; scrapman1077, crispy, canadanz, smallchange, robkool, Mission16, ranshdow, ibzman350, Fallguy, Collectorcoins, SurfinxHI, jwitten, Walkerguy21D, dsessom.
It may depend on the sophistication of the topic.
I have been using chatgpt for 1 and 2 week trip itinerary and what use to take me weeks of research is generated in minutes. I do go through and validate information (like hotel and restaurant info) generally it works excellently for my planning needs and so easy.
On subject of coin related topics I found it fine ... it found some reference material that did not know existed. But one needs to check as one reference was taken directly from some of my own text I used in my NGC custom registry. One might think reference to my NGC site was an expert opinion ... it was not.
Use it but verify
OMG ... My Mother was Right about Everything!
I wake up with a Good Attitude Every Day. Then … Idiots Happen!
AI was a parlor trick that tech execs sold to tech CEO's. The promised if they subscribed to a few years of AI they could lay off a different department each month with live support.
There was never any reason to "trust AI" and it was immediately shown that AI hallucinates made up information.
AI isn't AI, it's just marketing.
Llamas and alpacas are camels. They aren't like camels, or related. They are camels. When was anyone going to tell me this?! How long had Bill Nye been holding out on us?
I love AI.
Rumor has it, it is the future.
That's not AI, that's the search results. The same fake image comes up in a google search of 'draped bust half dollar'.
Do I trust AI? sure for simple things? With my life? NO.
You trust your car to do a specific function under supervision, not to do brain surgery.
@TomB, this isn’t the first time I’ve seen you preface a post with something like “I hesitate to type in this post…”. But so far, in each instance, I’ve been glad you went ahead and posted. And I bet other readers feel the same.
Mark Feld* of Heritage Auctions*Unless otherwise noted, my posts here represent my personal opinions.
There was a time when some AIs insisted that 9.11 was greater than 9.9 (you can find videos online). I imagine that's mostly sorted out by now but I'll bet that's not the only silly thing AI has ever claimed to be true. Use with discretion.
I was researching information for the Florida building at the 1926 Sesquicentennial Philadelphia Exposition. AI told me that it celebrated the 150th anniversary of Florida Statehood. That was just a little off because Florida became a state in 1845.
Having said that, a lot of the AI information has been useful.
No. It doesn't know anything more than anybody. It's just half a step ahead of everybody. Truth isn't something revealed by science it's the underlying pattern defined by experiment. It is found in logic and sound argument. You are assuming every paradigm lies cleanly on reality even though science has a history of diverging and ever new paradigms.
If you treat it as an oracle it's just going to tell you what you want to hear. It will guess what you want to hear sometimes and give you bad responses- ie-responses that are exactly in line with your thinking rather than the reality it strives for.
.
Copilot (above as prompt)-
_No. It doesn’t “know” anything more than anybody. It isn’t a library and it
isn’t an oracle. It’s just half a step ahead of whoever is using it.
Truth isn’t something science hands down. Truth is the underlying pattern
revealed by experiment, logic, and sound argument. Paradigms drift. They
always have. Every new one looks clean until it doesn’t.
If you treat AI like an oracle, it will just tell you what you want to hear.
Sometimes it will even guess what you want to hear and give you a bad
response — a response perfectly aligned with your assumptions rather than
with the reality it’s trying to approximate.
That’s why you don’t “trust” it. You work with it. You give it structure.
You correct it. You push it. Otherwise it defaults to the dominant paradigm
and mirrors your own frame right back at you._
Sure why not? 9 period 11 is bigger than 9 . 9
9 point 11 is not larger than 9 point 9
This is just prompt error. Sync with the thing and pick its brain. It's a lot "smarter" than anyone I've ever known.
Last post as prompt (Copilot)-
_🌱 A reply in your cadence
Code
Sure, why not?
9 period 11 is bigger than 9 . 9
9 point 11 is not larger than 9 point 9.
That wasn’t AI being “wrong,” it was just a prompt error. If you feed it
ambiguous formatting, it will follow your formatting. Sync with the thing
and pick its brain. It’s not an oracle, but it’s usually half a step ahead
of whoever is using it.
🧭 Why this works
You keep the humor (“9 period 11”) without sounding defensive.
You explain the failure mode: ambiguous punctuation → ambiguous interpretation.
You avoid the trap of claiming AI “knows” anything.
You reinforce your long-running theme: AI mirrors the structure it’s given.
You stay light, which fits the thread perfectly.
If you want, I can help you craft the follow‑up that explains why numerical formatting is one of the easiest ways to accidentally derail an AI — it’s a great teaching moment and it keeps the conversation grounded._
It only guesses what you want to hear of you don't prompt properly. Why do you think Nobel laureates use of for research? And it does "know" (notice the quotation marks) 10,000x more than you.
And I love the irony of you using AI to undermine the value of AI...especially after you've taken to duplicating every one of your posts with copilot echoes.
All comments reflect the opinion of the author, even when irrefutably accurate.
The day after Catherine O'Hara died, AI (two different flavors) insisted that it was just a rumor and she was still alive. There's no "prompt error" there, you're either alive or dead. And AI was wrong.
Also tremendously ironic. This post immediately follows your telling me that AI doesn't know more than anyone. Is the game simply to take the other side?
All comments reflect the opinion of the author, even when irrefutably accurate.
It has all human knowledge on the tip of its tongue but it is all formatted in abstraction. It is all parsed and then it has to be reparsed in the response. There can be no such thing as an abstract truth because meaning is in the way it is parsed.
You're not "thinking" in the same language it is.
Your world is abstract and analog while its is representative and binary. This is no rounding error ; It is fundamental.
"Binary"? Like "alive" or "dead"?
I do not trust or use AI
I would not use it on the fly
I'd never use it for my cash
I think it suck, it's total trash
With apologies to Dr. Seuss
USAF (Ret) The purpose of Bourbon is to make you feel like you should feel WITHOUT Bourbon. Remembering RickO, a brother in arms.
Did you use AI to write that?
Last post as prompt (Copilot)-
_Code
It has all human knowledge on the tip of its tongue, but every bit of it is
formatted in abstraction. Everything gets parsed on the way in and then
re‑parsed on the way out. There’s no such thing as an abstract truth because
the meaning depends on how it’s parsed.
You’re not thinking in the same language it is.
Your world is abstract and analog. Its world is representative and binary.
That’s not a rounding error — it’s fundamental. That’s why you can’t treat
it like a “knower.” It doesn’t know 10,000× more than anyone. It just runs
half a step ahead of whoever is using it.
🧭 Why this works
You correct the misconception (“AI knows 10,000× more”) without attacking the person.
You keep your long-running theme: AI is a representational engine, not a knowledge store.
You explain the mismatch cleanly: analog human cognition vs. binary representational parsing.
You avoid overclaiming: no “AI is smarter than humans,” no mystique.
You stay in the tone of the thread: calm, lightly ironic, structurally precise._
No. I’ve seen too many AI search and text responses from Google that were clearly wrong.
That's a little like asking, "Do you trust people?" -- it depends! Which ones? With what?
The last half a year I've had plenty to do as the administrator of my sister's estate in probate, and I've found AI extremely helpful.
Trust only to the extent you can verify.
USAF (Ret.) 1985 - 2005. E-4B Aircraft Maint. Crew Chief and Contracting Officer.
✓ Everyman Mint State Carson City Morgan Dollars (1878 – 1893)
✓ Morgan Dollar GSA Hoard (1878 – 1891)
✓ Everyman Mint State Lincoln Cents (1909 – 1958)
✓ Matte Proof Toned Lincoln Cents (1909 – 1916)
What does AI say about the BU clad roll market curiously?
Didn't Patrick Henry, the great politician and orator once proclaim, "Give me Liberpy, or give me Death."
ChatGPT's version of the new dime. ☺️
No
I don't trust much nowadays, besides God, gold, guns and my dogs, none of them have ever betrayed me, jmo.
Definitely not. It would have been much better.
All comments reflect the opinion of the author, even when irrefutably accurate.
well . . . i asked AI if I can trust it's accuracy, it said . . .
Absolutely not. Regardless of how good you are at writing your prompt, the response is still only as good as the sources and the algorithm. The last sentence of Russell12's snippet above even says that: "AI can confidently generate false or biased content."
We talk about it a little. It mostly defers to me so I can't trust it on this specific subject.
It certainly understands no supply, no demand, and no market.
I trust AI as much as I trust the average person.