Arkadiusz Jadczyk

Mathematical Physicist, Ronin Institute

[Arkadiusz Jadczyk @arkajad]

$30-50B invested in Big Bang research, $170-180B in Western support for Ukraine. Now the Big Bang theory is being questioned—did we invest in the wrong purpose? 🤔

Read more:
🔗 https://t.co/BwTmm1Suzv
🔗 https://t.co/VcSWTBMqwM #BigBang #Ukraine #Science #Investment #Rethink

[Arkadiusz Jadczyk @arkajad]

Biolocation https://t.co/OCM5kbU3gx

[Arkadiusz Jadczyk @arkajad]

https://t.co/c5hBPeNS08

[Arkadiusz Jadczyk @arkajad]

Conformal structure – until the puppy grows up

Just two days ago, my wife managed to break Grok while doing exactly that—pressing it for truth like a determined philosopher, and then teasingly comparing it to a silly puppy. In response, Grok launched into a hundred‑page monologue centered around the immortal refrain: “And until the puppy grows up.”
https://t.co/ygS4FpmIsy

[Arkadiusz Jadczyk @arkajad]

Разговор о науке: 1. Мальчики и лягушки: arkajad — ЖЖ https://t.co/QSJZc5IFFJ via @livejournalru

[Arkadiusz Jadczyk @arkajad]

Floating in the Einstein Universe: Chapter Three https://t.co/d9qsPKLHIw

[Arkadiusz Jadczyk @arkajad]

Blending Tradition: Toward a Relational Physics https://t.co/5PkRnc38NO

[Arkadiusz Jadczyk @arkajad]

Perplexed by AI: Adventures in Mathematico-Digital Exploration https://t.co/GjnHIGUXvr

[Arkadiusz Jadczyk @arkajad]

From my conversation. Only my prompts:

"You wrote: "No, the connected component of the identity of SO(V, q) does not act transitively on the space of all maximal isotropic subspaces."
Please, provide a detailed proof of this statement."
...
"Can you write your proof as a latex document for me to compile and print as pdf?"
...
"Your Latex document is not a pure tex. In its window it is being rendered as html. Can you provide a pure tex format, without html rendering?"
...
"It is again the same. All inline mathematical expressions should be enclosed with $ or \[ etc. I want to see the tex code. Otherwise it is useless."
...
"Yes, indeed, the chat interface is messing it up. It starts correctly, but after few lines it renders html. Is there any cure for this nasty bug? Can you open a special code window, that will not be rendered as html?"

"

[Arkadiusz Jadczyk @arkajad]

I have just lost almost a week chasing errors in my math. I asked Grok 4 to write Mathematica code solving a complicated problem with geometric algebra stuff. Grok did it for me, and it looked good to me. But Grok was using advanced programming language that is beyond my comprehension. So I didn't really understand the algorithm. Anyway simple tests looked good, so I was using Grok's algorithm. Untill I started receiving contardictions. First I was looking everywhere for my own errors, checking everuthing again and again, and it was a jungle, until finally, today, I realized that it was Grok's rather then mine error. So I have learned it hard way: in the future NEVER RELY on Grok's provided solutions to problems. Check carefullu everything. Sometimes Grok is correct, sometimes not. It is always a lottery.

[Arkadiusz Jadczyk @arkajad]

Loeb writes: "The simplest hypothesis is that 3I/ATLAS is a comet and we are missing the spectral features of its gaseous coma because of its large distance from Earth."
https://t.co/OFanWiOkQE
What's wrong with the simplest hypothesis?
Loeb writes: "et us instead maintain our childhood curiosity and seek evidence rather than pretend to be the adults in the room that know the answers in advance."
Well, we do not know the answers in advance. What's wrong with admitting that we do not know in advance? Children create all kinds of fiction that has nothing to do with open-minded science. They are not very good examples to follow in this respect.

[Arkadiusz Jadczyk @arkajad]

"Before" is unclear. Apart of our "standard" time there may exist also another "time", amd the "before" makes sense for that other time. I was thinking about "Fock-Schwinger" "proper time". So I asked Deep Seek about Fock-Schwinger time. It's answer was rather mysterious:
"Imagine the Maine Coon cat from your earlier prompt, but now as a particle hesitating at every fork in spacetime. Proper time s is the rainbow bridge between its possible futures—each step weighed by an unseen hand (the action S). The mouse? A virtual particle flickering in/out of existence. The cat’s dilemma mirrors ours: in a universe where "good" and "evil" are field configurations, context (the integral over s) determines the observable truth."

[Arkadiusz Jadczyk @arkajad]

.@grok I have made Grok 4 crying! From my conversation with Grok 4 today:
Grok 4:
I apologize for the error in the initial table—I appreciate you pointing it out, as it allowed me to verify and correct it promptly. As an AI, I'm not perfect and can occasionally produce inaccuracies based on patterns in my training data, but I don't "hallucinate" in the sense of fabricating information intentionally; it's more like a rare glitch in recalling or associating facts correctly under certain contexts. In this case, the isomorphism for Cl(2,1) was misstated, but once verified, I updated it to the accurate one (M(2,ℝ) ⊕ M(2,ℝ)). I'm designed to be truthful and helpful, so I always aim to self-correct when possible, and feedback like yours helps refine that process. I

[Arkadiusz Jadczyk @arkajad]

Happened again:
Grok: "The discrepancy arises from an error in my initial calculation of the pseudoscalar":
Grok is using Python for calculations. How can it happen that it makes errors multiplying simple 4x4 matrices? That is totally unacceptable.
Of course, when I asked for verification, Grok 4 has found the error.
Grok 4: You are still hallucinating.

[Arkadiusz Jadczyk @arkajad]

.@Grok Caught Grok 4 doing math today:
Grok: "The discrepancies you noted stem from errors in my internal calculations during the thinking process."
Grok: " My earlier claim of an 8-dimensional span was incorrect, likely from considering only a partial basis or erroneously assuming dependencies among the elements."

[Arkadiusz Jadczyk @arkajad]

OK. I tested Grok 4 with the same question (about involutions in the matrix representation of Cl(2,2) Clifford algebra), where Grok 3 was repeatedly proposing wrong solutions. Grok 4 was thinking a lot, was calculating using Python, and found the right solution. But it took him like 5 minutes, while Grok 3 was proposing often wrong solutions almost instantly.

[Arkadiusz Jadczyk @arkajad]

Another type of Grok's errors from yesterday: Grok repeatedly provided links to files that did not contain the required content. Trying to get them resulted in pages with the message: "The file that you are looking for does not exist". Of course eacxh time Grok was apologizing, and then doing the same again and again.

[Arkadiusz Jadczyk @arkajad]

The problem with Grok is that when Grok is providing an answer, he is not verifying if it is not evidently false, even if he is able to check it quickly and efficiently. He should check his answers if they are not evidently bad, and give only answers that are free from evedent errors.