Open LOGIC-10 opened 2 months ago
Seems that these models haven't incorporated the information in the book "Quantised Accelerations" by Mike McCulloch (https://x.com/memcculloch)
A book review at https://tallbloke.wordpress.com/2024/06/25/book-review-quantised-accelerations-by-mike-mcculloch/ nicely summarised it as follows: " Dr McCulloch’s Quantised Inertia theory demonstrates that the stars we thought were going round the edge of galaxies faster than they should be, aren’t, and the dark matter which we thought was needed to keep them gravitationally bound, isn’t. " And: " since Dr Mike has also applied his QI principles to derive the gravitational constant from the same universal attributes, we should summarize where we’ve got so far:
By dealing with anomalously fast stuff (stars), we’ve got rid of the unpleasantly unobservable dark stuff (DM), while eliminating the constant for heavy stuff (G), by relating the fastest stuff (c) to the biggest stuff (Universal diameter). "
The prompt is not designed as an structured architecture, much on the contrary its meant to increase entropy and chaos to generate creative answers
Yes, increasing entropy and chaos is certainly the aim. The other aspect is that model temperature will affect the degree of creativeness and this will also be model-specific.
I have followed a rather different output structure and architecture on the Gemini port that I am maintaining, which I personally find more useful. For those who are interested, please see various documented test-cases that are available here: https://github.com/umaplehurst/SuperPrompt.gemini/tree/main/test-cases/gemini-1.5_pro
Yes, increasing entropy and chaos is certainly the aim. The other aspect is that model temperature will affect the degree of creativeness and this will also be model-specific.
I have followed a rather different output structure and architecture on the Gemini port that I am maintaining, which I personally find more useful. For those who are interested, please see various documented test-cases that are available here:
https://github.com/umaplehurst/SuperPrompt.gemini/tree/main/test-cases/gemini-1.5_pro
Good direction please share more, or create a wiki page
Should I just paste and put your complete prompt into the system prompt for the model?
I find that most of the time the model doesn't follow the structure.
This is my exp result for four models with the same question:
Let's discuss the nature of dark matter.
Below are how the models performed with the exactly same sys_prompt and my question as a user_prompt:
claude-3-sonnet-20240229
Y
Dark matter is one of the great unsolved mysteries in cosmology and particle physics today. Here's a brief overview of what we know and don't know about dark matter:
What We Know:
What We Don't Know:
Solving the dark matter mystery is a major goal for particle physics experiments like those at the Large Hadron Collider, as well as observational efforts aimed at revealing the nature of this elusive substance through its gravitational effects.
claude-3-opus-20240229
Y
Y
gpt-4o
I am going to discuss the nature of dark matter.
Dark matter is a mysterious and still largely unknown substance that makes up approximately 27% of the universe's mass-energy content. Unlike ordinary matter, dark matter does not interact with electromagnetic forces, meaning it does not absorb, reflect, or emit light, making it invisible and detectable only via its gravitational effects.
The evidence for dark matter comes from several astronomical observations, including:
Dark matter is hypothesized to be made up of one or more types of particles yet to be discovered. Some leading candidates include:
Numerous experiments are underway to detect dark matter particles either directly, through their rare interactions with ordinary matter, or indirectly, by searching for the products of dark matter particle annihilations or decays.
With the current understanding and ongoing research, the nature of dark matter remains one of the most intriguing puzzles in modern astrophysics and cosmology.
Y
gpt-4o-mini
I will provide an overview of dark matter and its implications in cosmology.
Dark matter is a form of matter that does not emit, absorb, or reflect light, making it invisible and undetectable by electromagnetic radiation. It constitutes approximately 27% of the universe's total mass and energy content. Evidence for the existence of dark matter comes from several astrophysical observations, including:
Galactic Rotation Curves: The speeds at which stars orbit the centers of galaxies suggest that there is more mass present than what can be seen. The outer stars of galaxies rotate at higher speeds than would be expected based on visible matter alone, indicating a significant amount of unseen matter exerting gravitational pull.
Gravitational Lensing: The phenomenon where light from distant objects is bent around massive objects (like galaxy clusters) allows for the mapping of dark matter's distribution, as the bending occurs due to the gravitational field of dark matter, despite it not being visible.
Cosmic Microwave Background (CMB): Observations of the CMB radiation provide insights into the Universe's early formation and support the existence of dark matter. The fluctuations observed in the CMB are linked to the density of matter, including dark matter, in the early universe.
Large Scale Structure: The distribution of galaxies and galaxy clusters across the universe suggests the presence of dark matter, as its gravitational effects influence how matter clusters at large scales.
Despite its significant role, the exact nature of dark matter remains one of the most puzzling questions in modern physics. Various candidates for dark matter particles have been proposed, including Weakly Interacting Massive Particles (WIMPs), axions, and sterile neutrinos, but none have been conclusively detected.
In summary, dark matter plays a crucial role in our understanding of the universe's structure and evolution, yet its intrinsic properties and composition remain largely unknown.
Y