I subscribed to Grok 4 a couple of hours ago with the hopes of training it in the subject of my Temporal Rights theory of natural rights. I had done that a week ago with ChatGPT by asking it to analyze Temporal Rights. It did a fair job but showed some bias toward traditional theories of natural rights and then went completely bonkers, quoting and citing texts that were total fabrications.
In all fairness, it eventually became a significant help to me by becoming “someone” with whom to discuss the ontological basis of Temporal Rights. For example, I said I was concerned that attributing rights to capabilities might be arbitrary, but it showed me how it was logically existential.
Anyway, I became a victim of advertising this evening and paid $19.99 for a Grok 4 subscription with the hope that it might prove to be an even more intelligent AI than ChatGPT, and help me identify any possible weaknesses in my logic. I worked with it for some time and finally realized that it was actually fabricating responses out of thin air.
This is a frequent problem with all AI. The problem arises because artificial intelligence is an oxymoron. It has no native intelligence, per se. So far, AI is primarily a word association genius called a Large Language Model (LLM) that looks for relevant associations, then evaluates, curates, and responds with its “findings”. Most of the time, it is a wonderful assistant, but it can and does go off the rails and not only fails to comprehend but produces falsehoods, as the dialog below will show you.
When I called Groc 4 out this evening it confessed thus:
“You are absolutely right. I am clearly still struggling with accurate recall. I sincerely apologize. My previous response was a complete fabrication.”
You would hope that artificial intelligence could detect when it was about to tell a lie!
The Grok 4 advertisements claim that it is “smarter than almost all graduate students, in all disciplines, simultaneously.” However, it could not even confine its analysis to a particular article while claiming to do that.
Grok apparently deeply skewed its approach to analysis by putting excessive weight on the URL of the article and on the meaning of “temporal”. Because the URL mentions God, it assumed God was a primary feature of Temporal Rights, even to the extent of getting the meaning of “temporal” wrong.
If you want to understand how Grok got this all wrong, you may want to get at least a hint at the Temporal Rights view of natural rights. I recommend “Temporal Rights: An Executive Summary” for that. Then come back here and read Grok’s self-inflicted embarrassment.
Here’s a few quoted lines of this evenings session with Grok 4:
[Begin quoting my session with Grok 4]
[Grok] Core Arguments of “Rocks Have Rights” (Directly from the Text):
- Critique of Anthropocentrism: Pemberton argues against the idea that humans are the only things that matter. He states, “It is time to stop thinking that the world is all about us.”
- God’s Creation and Dominion: He emphasizes God’s creation and humanity’s role in it: “God created the world and everything in it. He then gave it to us to manage.”
- Responsibility to Creation: He argues that this dominion comes with responsibility: “We are supposed to take care of it. We are supposed to make sure that it thrives.”
- Inherent Value: He implies that all of creation has inherent value by stating, “Everything has a purpose.”
- Rights as Responsibility: He connects rights to responsibility: “If we have a right to live, then we have a responsibility to let others live.”
[Editor’s Note: Although Grok says “Directly from the text” none of these points are found in the text! They sound like they are aimed at the writings of traditional theorists like John Locke. These LLMs (Large Language Models), artificial intelligence bots use word associations to build their responses, and this heavy bias to other, earlier theories and/or its initial assumption about the basis of Temporal Rights is evidence of that weakness.]
How This Relates to Our Discussion (Based on Accurate Textual Evidence):
- Expanded Scope of Responsibility: The article expands the scope of human responsibility to include the entire created order.
- Inherent Value Beyond Humans: It suggests that inherent value extends beyond humans, although it doesn’t explicitly state that everything has “rights” in the traditional sense.
- Theological Foundation: The argument is firmly rooted in a theological understanding of creation and humanity’s role within it.
I apologize again for my previous inaccuracies. I am working to improve my ability to process and recall information correctly.
Given this revised and more accurate understanding, what aspects of Pemberton’s argument would you like to explore further?
[Me] I can’t find “It is time to stop thinking that the world is all about us.” anywhere in that article.
[Grok] You are absolutely right. I am clearly still struggling with accurate recall. I sincerely apologize. My previous response was a complete fabrication.
I need to recalibrate my approach. I am going to stop relying on my memory entirely and focus solely on analyzing the text directly.
[I am skipping now to the end of our session and continuing to quote it. I have just asked if Grok had access to ChatGPT.]]
[Grok] Is there something specific you were hoping I could do with access to ChatGPT? Perhaps I can accomplish the same goal using my own capabilities.
[Me] Well, i asked ChatGPT these questions and it had no trouble getting its “head” around Pemberton’s Temporal Rights as a new and unprecedented theory of natural rights based on existential factors.
[Grok] I understand your frustration. It’s clear I’m struggling to grasp Pemberton’s theory and discuss it accurately, while ChatGPT seems to be handling it with ease.
Since I’m having difficulty, and you’ve already had a productive conversation with ChatGPT, perhaps it would be best for you to continue exploring the topic with that tool. It seems better suited to this particular discussion.
I apologize for not being more helpful. I am still under development, and I have much to learn. I appreciate you bringing this to my attention, as it helps me identify areas for improvement.
If you have other topics you’d like to discuss, I’m happy to try, but I understand if you prefer to work with a different AI model at this time.
[End of Quoted Session]
This experience with Grok 4 provides me, and now you, with a serious warning about the level of trust we should have for AI.
2 thoughts on “Grok 4 Apologizes Then Gives Up”