Apple’s ‘The Illusion of Thinking’ is shocking – but what did you remember here

But enough about me. I have to tell you all the tea on “The Illusion of Thinking” … You will not believe how many people are missing the big picture.

Confusion of thinking

Before we jump, let’s define what a logic model is. A large regioning model (LRM) is an AI system designed to solve complex problems by working through each step, essentially showing its logic process before providing answers.

Unlike the standard language models that predict the next word, these models are trained to implement structured arguments and logic, which, more human-like, breaks problems in a step-by-step manner.

Also: I gave more information about myself – here is told how AI used that personal information

Apple’s “The Illusion of Thinking” kept some top big rational models such as O 3 Mini, Dipsec R1, and Cloud 3.7 sonnet for testing with four classic puzzles: 🎮

• Tower of Hanoi
• Checkers jumping
• Across the river
• Block world

As the testing became difficult, the model was Vile e. Hit a wall -like wall.

Womp wompppppppppppppppppp.

The performance of the model decreased with an increase in complexity, and adding more computing strength or token did not help.

The more complicated the puzzles, the less these models thought, even though they had the ability to continue. He originally accepted the defeat – like humans!

Anyway, Apple broke the results into three performance areas:

• Low complexity: The standard model improved and used low tokens.
• Moderate complexity: Thinking model showed an advantage.
• High complexity: All models failed. The accuracy fell rapidly.

Even when Apple answered LRMS as an algorithm, the models still failed and could not follow a multi-step plan to complete.

LRMS’s ability to argue was incompatible. A model can complete more than 100 moves in the tower of Hanoi, but can manage only less than 5 correct moves in the river crossing.

Also: This new chat trick lets you download your canvas work as PDF in second

This behavior suggests pattern matching, not the correct argument or intelligence.

Everyone said, “confusion of thinking” began a turf war and got a good amount of criticism. It is important to note that this paper is a preprint and has not reviewed the colleague.

In other words, Apple kept the paper out and it was, “Believe me, brother.”

Another criticism is about to make online rounds that Apple used puzzles instead of real -world scenarios. Critics also argue that the puzzles are more than the model tokens and step range, so “failures” reflect these obstacles rather than lack of logic.

But the most inflicting criticism is to be behind, “Apple is behind AI, so it is pointing out the weak places of rival systems after losing the ground.”

Completely random, but when I read that criticism, I cannot help, but think of Danzel Washington in “Training Day”.

“King Kong did not love me!”

Apple is probably very classy for such colorful language, but the idea still cracks me. Hehehe

The movie quotes one side, let’s discuss our next step.

Your next trick

Let’s become serious for a second.

Apple evaluated, batted his ball and batting and going home.

If true, it is good on Apple to call AI, equivalent to “natural flavors”, stating that one of the most important AI progress can be just a product of marketing department.

Also: Why I am switching to VS code. Indication: This is about AI tool integration

Meanwhile, the Internet is left to take out it. One camp claims that AI is all publicity, while the other argues that the test was rigged from the beginning.

Do you buy in publicity, or you ignore AI completely?

Here’s the matter: If you have ever used AI in a meaningful way, then you know that this report showed that there is some truth. Even if Apple pushed it a little bit barely, it does not change the fact that, based on our conversation, we believe that AI is still lacking and a long way is to go.

Currently, AI performs the best when we are subject matter specialist and can identify our mistakes.

I don’t think this report changes the big picture. In fact, it doesn’t matter a little.

The logic model can be more publicized than the substance right now, but this does not mean that AI will not take jobs or replace industries. Today, dismissing AI would be like rejecting the initial computer just because it was slow.

Even if the AI ​​does not even lasts a brochure, its capacity is incredible. Speaking for myself, AI has enabled me to effectively consider-churn and automate tedious tasks, eventually I have been given a level of productivity that will require to work on 26 hours of days.

Also: Best AI for coding in 2025 (and what not to use)

Although this is not correct and sometimes reaches my final nerve, it strengthens everyday people like us and to be more independent and creative when working on our businesses or careers.

So what should you do?

First, continue to be informed, but also apply what you learn. In addition, when it comes to AI, there is not much or too low at the height.

Outside him, here are some more recommendations:

• Use AI as a tool, not crutches: Think “do with me,” No “do it for me.”

• Do not dismiss AI: Just because it is flawed does not mean that it cannot be helpful.

• Be subject expert: Continue learning and growing in your area. The more you know, the more useful AI becomes.

• Automatic boring goods: Free your time by handling repetitive tasks to AI so that you can continue to learn and implement.

• Double on human strength: AI is powerful when combined with human abilities such as leadership, sympathy, problems and cooperation.

• Take action, even if it is a mess: Do not wait for the correct or correct signal of AI. If you are not already, start playing with AI. People who win with AI are those who are using it.

Finally, it acts as a valuable life lesson: be conscious of the potential agenda. My father once told me, “A carpenter will always ask you to build a house using wood.”

my two cents

As I step down from my soapbox, I want to clarify something clearly …

I am not reporting Apple. I believe that Apple’s findings are on the basis of my conversation with these models, but what is so?

The job will still be lost, and industries will be replaced by AI. This is just a fact.

While the report is practical, it does not fundamentally change the reality we are facing. At the same time, I am not here to defend LRMS here.

If anything, I am on “Team You”, and as a partner of your team, I am here to remind you: don’t throw the child out with a batht.

😇 Hope it helps. I listen to you.

PS If you want more easy and useful AI tips and tricks, then sign up for my free newsletter,No full just facts,