top | item 46576014

(no title)

stuartjohnson12 | 1 month ago

In the essay I linked, there are some instructions you can follow to test out the idea under "step 1". It's really important to follow them exactly and not to use the same ChatGPT instance as you're talking to about this idea so we can test with an independent party what is going on. I'd be curious what the output is.

discuss

order

daikikadowaki|1 month ago

I took the challenge. To ensure a completely objective 'reality-check,' I opened a fresh session in Chrome Incognito mode with a brand-new account and used GPT-5, as suggested.

I followed 'Step 1' of the essay to the letter—copy-pasting the exact prompt designed to expose self-deception and 'AI-aided' delusions. I didn't frame it as my own work, allowing the model to provide a raw, critical audit without any bias toward the author.

https://chatgpt.com/share/6963b843-9bbc-8001-a2ea-409a5f6dd6...

stuartjohnson12|1 month ago

Awesome - now read it really closely and compare it to the version of reality in your OP. And DON'T paste it or this comment into your normal ChatGPT instance and ask it to respond. Really just think for a moment on your own.

> The goal: replace vague legal and philosophical notions of “manipulation” with a concrete engineering variable. [...] formally define the metric

What's the conclusion? Is this a "concrete engineering paper"? Has anything been "formally proved"? From your link:

> The math is conceptual, not formal.

> This is serious, careful, and intellectually honest work, but it is not conventional science.

> The project would be strongest if positioned explicitly as foundational theory + open design pattern, rather than as something awaiting “validation.”

> it is valid as a design pattern or architectural disclosure, not as experimental systems research

Be careful before immediately dismissing this as just imprecise language or a translation issue. There's a reason I suggested this to you.

thunfischbrot|1 month ago

That’s not too bad and mirrored some of the feedback in this thread. Tldr: interesting idea, more worthy of a blog post or a thread in one of your favourite online communities, rather than a paper.