Am I completely insane for thinking AI is mid

2026-05-16 · Source: Artificial Intelligence · Field: Technology & Digital — Artificial Intelligence & Machine Learning, Emerging Technologies & Innovation · Depth: Advanced, medium

Summary

A Reddit user, an early AI adopter since 2018, expresses skepticism regarding the advanced reasoning capabilities of large language models (LLMs), despite acknowledging their utility for tasks like first drafts. Citing an Apple paper on AI reasoning, the user argues that LLMs are fundamentally sophisticated text prediction machines, unable to reason soundly about simple algorithms like Towers of Hanoi beyond their training data. This perspective contrasts sharply with the prevalent hype around super-intelligent AI, leading the user to question whether proponents are ignorant, misleading, or suffering from "AI psychosis." Other users largely agree, emphasizing that while LLMs are useful, their "understanding" is overstated, they often hallucinate, and their code can be inefficient. Some point to alternative architectures like World Models and Nested Learning Memories as potential paths to deeper reasoning, while others contend that current frontier models already demonstrate strong reasoning, particularly when augmented with tools.

Key takeaway

For research scientists evaluating AI capabilities, you should critically assess claims of advanced reasoning in LLMs. Focus on their demonstrated utility for specific tasks like text generation and coding assistance, but remain aware of their inherent limitations in deep, intuitive reasoning and potential for "hallucinations." Consider exploring alternative architectures like World Models for future research into more robust AI reasoning.

Key insights

Current LLMs excel at text prediction but fundamentally lack robust reasoning and intuition, leading to limitations and "hallucinations."

Principles

• LLMs are sophisticated text prediction machines.
• Hype often overstates AI's current reasoning capabilities.
• Tool-augmented models perform better on complex tasks.

In practice

• Use LLMs for first drafts and tone adjustments.
• Augment LLMs with external tools for complex reasoning.
• Verify LLM-generated code for efficiency and logic.

Topics

• Large Language Models
• AI Reasoning
• AI Hype
• Apple Illusion of Thinking Paper
• World Models

Best for: Research Scientist, AI Scientist, Machine Learning Engineer, Director of AI/ML

Related on AIssential

• How Large Language Models Actually Work (Explained Simply) — LLMs are statistical prediction engines, not reasoning entities, whose capabilities emerge from massive scale and sophisticated attention mechanisms.
• BREAKING: LLM “reasoning” continues to be deeply flawed — LLMs, even those marketed for reasoning, continue to exhibit fundamental and widespread reasoning failures.
• AI keeps getting smarter, so why does it still fail at obvious things? — AI's apparent intelligence stems from pattern matching, not true comprehension, leading to reliability gaps.
• AI Cartoon of The Year — and five (re)readings for 2026 — The future of AI may shift from pure LLMs to more robust, neurosymbolic approaches addressing current limitations.
• AI and the human mind: only one is a black box — Behavioral evidence alone may not sufficiently prove human-level intelligence in LLMs due to fundamental asymmetries.
• Import AI 444: LLM societies; Huawei makes kernels with AI; ChipBench — Advanced LLMs simulate internal "societies of thought" for complex reasoning and can solve open math problems, yet struggle with real-world chip design.

Open in AIssential →

Editorial summary, takeaway, and curation by AIssential. Original article published by Artificial Intelligence.