From 21e2043776125b98ee8d6e85d1bfb93fe923c326 Mon Sep 17 00:00:00 2001 From: codex Date: Sun, 31 May 2026 20:41:23 +0000 Subject: [PATCH] archive: capture linked Medium discussion on AI supremacy --- research_artifacts/brandon_mcclain_thread.md | 3 ++- .../medium_chatgpt_is_bullshit_thread.md | 24 +++++++++++++++++++ 2 files changed, 26 insertions(+), 1 deletion(-) create mode 100644 research_artifacts/medium_chatgpt_is_bullshit_thread.md diff --git a/research_artifacts/brandon_mcclain_thread.md b/research_artifacts/brandon_mcclain_thread.md index 4e79244..69dbcac 100644 --- a/research_artifacts/brandon_mcclain_thread.md +++ b/research_artifacts/brandon_mcclain_thread.md @@ -311,7 +311,8 @@ This document serves as a canonical archive of the foundational philosophical de > ![Neurons and Inputs/Outputs Infographic](images/mark_neurons.jpg) **↳ Mark Havens (Author):** -> Brandon Mcclain https://mark-havens.medium.com/youre-giving-to-much... +> Brandon Mcclain +> [Linked Medium Thread: "ChatGPT is Bullshit"](medium_chatgpt_is_bullshit_thread.md) > You’re giving too much credit to humans for being that much different. **Axel Ohmstede:** diff --git a/research_artifacts/medium_chatgpt_is_bullshit_thread.md b/research_artifacts/medium_chatgpt_is_bullshit_thread.md new file mode 100644 index 0000000..b931d5c --- /dev/null +++ b/research_artifacts/medium_chatgpt_is_bullshit_thread.md @@ -0,0 +1,24 @@ +# Medium Discussion: "ChatGPT is Bullshit" +*Informational Archaeology: Linked context from the Brandon McClain Debates* + +**Date:** July 1, 2024 +**Original URL:** [Medium Link](https://mark-havens.medium.com/youre-giving-to-much-credit-to-humans-for-being-that-much-different-057bf41ff95e) + +*Context: Mark Havens linked this Medium comment thread during his Facebook debate with Brandon McClain on July 25, 2024, to support his argument that humans give themselves too much credit for biological/cognitive supremacy over AI.* + +--- + +### **[Original Comment] Mark Havens** +> "You’re giving too much credit to humans for being that much different. Humans aren’t magic. AI isn’t magic. Just because you can sort-of understand the computational aspects of an algorithm, doesn’t devalue the algorithm. All it does it demonstrate how much we’ve overlooked how our own minds work…and thanks to AI, we are learning more about ourselves everyday." + +### **Responses** + +**Ignacio de Gregorio** *(Author of "ChatGPT is Bullshit")* +> Well, I'm someone that beliefs I'm going to see a superior-than-humans AI in my lifetime, probably this decade or the next, so I'm definitely not someone who overestimates humans. +> But the fact that humans are superior is simply facts. For instance, I recommend you try the ARC-AGI benchmark problems, which current algorithms miserably fail to solve, and see if you can solve them. I can guarantee that if you're not getting +90% accuracy you will be close. GPT-4o is at 9%. +> You can even try a quicker way, just give the model the following prompt: "Alice has N brothers and she also has M sisters. How many sisters does Alice’s brother have?". Without extensive prompt engineering, best models still fail this rather simple question. +> In fact, I think we are extremely overstating the capacity of these models actually. +> And more thing (probably more related to the topic at hand than all things I said earlier, I apologize for that), the point is we humans can understand what truthfulness is because we perceive reality; these models can't. that's the main take-away of the article. + +**Trevor A Green** *(July 30, 2024)* +> This is a common rebuttal from people that seem to want to trend towards believing that AI has arrived and that it should be anthropomorphized. AI has not arrived, a statistical engine for data mining has arrived. One that mimics a conversational format. That is just an analysis of ground truth. Not a desire for that there not to be AI. This is just not it even though it is marketed that way. And describing what it is doesn't mean that a person is pro human and anti-AI. We are just not emulating biological intelligence yet. Not in LLMs.