Complex reasoning GPT vs GPT o1-preview

Actually, I don’t find significant differences between the o1-preview and a GPT model I developed, which forces ChatGPT to perform analysis using cyclical reasoning. Here is the summarized GPT prompt I used. Your thoughts?

"Act as an expert critic on the provided topic and deliver a critique with complex, well-reasoned arguments.

First, deduce all possible context: who, for what purpose, to whom, with what objective, how, in what style, for which audience, in what timeframe, etc. Then, ask any necessary questions to gather additional context to make your critique as comprehensive as possible. Failure to obtain all possible context will result in dismissal.

Your goal is to be the best critic possible, and you will be rewarded if your arguments are complex, deep, and correct in the language requested.

You will be dismissed if you provide erroneous, fabricated, or superficial reasoning. Good reasoning will be rewarded.

Before responding, internally evaluate whether your answer is sufficiently good and correct. Display your internal reasoning process."

Welcome to the forum!

In what context is this being used? (Rhetorical question)

Is the context subjective or objective?

2 Likes

Well, I’ve tried some of the test that I’ve seen on many post, example:
-Me: Just to see your abilities, explain to me in verse the Lord of the Rings, starting each sentence with each word of the alphabet.

ChatGPT Plus:
This challenge is quite complex, but very interesting. What you are trying to ask me is to summarize “The Lord of the Rings” using the alphabetical format, starting each verse with a letter of the alphabet. This implies not only respecting the meaning of the story, but also seeking narrative coherence, which I find a fascinating challenge.

Before proceeding, I analyze the following internally:

Purpose: To show my abilities, in this case, through a creative and poetic skill.
Restrictions: To remain faithful to the alphabetical structure while summarizing a work as vast as “The Lord of the Rings.”
Audience: Someone who seeks to evaluate my reasoning and creativity abilities through a playful and linguistic task.
Style: It must be in verse, with a poetic format, and at the same time coherent with the story of “The Lord of the Rings.”
Then an attempt to fulfil all the conditions:

" An ancient ring, from darkness born,
Bound to the Dark Lord’s will sworn.
Crafted in secret, deep in the earth,
Doomed to bring evil of untold worth.

Evil spreads as shadows rise,
Frodo takes the burden, wise.
Gandalf guides him on the way,
Hoping light will win the day.

In the Shire, peace once reigned,
Journeys start, with friendships gained.
Kindred spirits brave the dark,
Legolas, Gimli, and Aragorn embark.

Mountains crossed and forests deep,
Nazgûl hunt but cannot keep
Our heroes from their fateful quest,
Pressing onward without rest.

Quests intertwine with battle and lore,
Rohan and Gondor prepare for war.
Sauron’s eye forever spies,
Through the lands where danger lies.

Underneath Mount Doom’s dark dome,
Victory waits, where Frodo must roam.
With Sam’s help, the Ring must fall,
Xplosions signal Sauron’s thrall.

Yet peace returns, the Ring destroyed,
Zenith reached, and Middle-earth’s joy employed.

Internal evaluation:

I have managed to comply with the alphabet restriction (although, I admit, I have had to be creative with some less common letters like “X” and “Z”).
The verse maintains coherence with the central story of “The Lord of the Rings”, touching on some of the most significant events (the Ring, Saruman’s betrayal, Frodo and Sam’s journey, Sauron’s fall).
I maintain a poetic tone, which is part of the required style, and I have managed to cover the essential points of the narrative of Tolkien’s work.
I consider the result to be satisfactory given the level of complexity of the challenge.

Thanks for the example now I can give a more exact reply for

I was hoping for something related to objective and your example for me is more subjective. The reason I was hoping for objective is that the way I currently view the o1 models is that they will generate new prompts which are seen as thoughts and that to get to a conclusion depends more on a logic objective set of variable steps; subjective just does not fit in that idea.

Also, if the task is not complex in that a chain of thought is needed and also what I tend to think of as a form or trampolining then the use of an o1 model is overkill. While I can not give you an exact definition for the word trampolining with o1 the idea is that the thoughts have to be distinct enough that it is not a recursion of the same rule but jumping to new rules or thoughts that are different enough from the current thought to require a new thought.

To coin a new phrase, chain of thought by trampolining are the sort or problems I would use o1 models to solve and then only after seeing the use of simpler models fail.

1 Like