What is something LLMs do poorly when it comes to writing?

Large Language Models (LLMs) excel at generating fluent text, but they consistently struggle with producing nuanced, original, and deeply researched content. Their writing is often characterized as "flat" or "lifeless," producing a median style of prose that lacks a unique voice or personality.

What do LLMs do poorly when it comes to writing?

LLM writing often avoids specificity. It refers to ideas without defining them and makes claims without evidence. E.g., “Some experts say prompt engineering is becoming less important. The ability to simply prompt LLMs can have a major impact on productivity.” But who are the experts?

What does LLMs struggle with?

LLMs struggle with tabular data

Every LLM was trained on billions of words, learning to predict what comes next in a sentence. But analytical data doesn't follow narrative patterns. A 50-column table isn't a story, it's a multidimensional relationship map where correlations matter more than sequence.

What are the weaknesses of LLMs?

Limited Reasoning Skills

While LLMs are good at understanding and generating sentences, they're not great at solving complex problems. For example, if you ask an LLM to solve a multi-step math problem or a puzzle, it might get confused and make mistakes along the way.

What are the limitations of LLMs?

Key Limitations

• Hallucinations and Inaccuracy. ...
• Bias and Fairness. ...
• Limited Understanding. ...
• Context and Memory Constraints. ...
• Security Vulnerabilities. ...
• Multimodal and Agentic Challenges. ...
• Resource and Environmental Impact. ...
• Need for Human Oversight.

CERN Insider Speaks Out: “Something Is Wrong with Reality”

What can LLMs never do?

LLMs can't stop, gather world state, reason, revisit older answers or predict future answers, unless that process also is detailed in the training data. If you include the previous prompts and responses, that still leaves the next inference starting from scratch as another single pass.

What is the biggest weakness of AI?

Bias and Discrimination: Generative AI models can inadvertently learn biases present in the training data, which can perpetuate or amplify existing biases and discrimination. This can lead to biased outcomes in areas like language generation, image synthesis, or decision-making systems.

What are the 5 biggest AI fails?

• Volkswagen's Cariad Billion-Dollar AI Fail.
• Taco Bell's Drive-Thru AI Gone Wrong.
• Google AI Overviews: The Hallucination Problem.
• Arup Deepfake Heist: $25 Million Stolen.
• Replit "Rogue Agent": Complete Database Deletion.
• McDonald's & Paradox.ai: 64 Million Records Exposed.
• UnitedHealth & Humana: Algorithmic Care Denial.

Why do LLMs fail?

Unfortunately, LLMs are not yet great strategic planners. They often create plans that seem logical on the surface but are deeply flawed. They might miss critical steps, put tasks in the wrong order, or fail to account for potential problems.

What are the 7 problem characteristics of AI?

The 7 key problem characteristics in AI help define how to solve them: decomposability, ignorable/reversible steps, predictability of universe, absolute vs. relative solutions, state vs. path solutions, role of knowledge, and need for human interaction, which categorize problems like chess (hard to decompose, non-reversible, uncertain) versus the Tower of Hanoi (decomposable, reversible) to guide algorithm choice. 

Why do LLMs make so many mistakes?

Large language models (LLMs) sometimes learn the wrong lessons, according to an MIT study. Rather than answering a query based on domain knowledge, an LLM could respond by leveraging grammatical patterns it learned during training. This can cause a model to fail unexpectedly when deployed on new tasks.

What is the 8 problem in AI?

The 8-puzzle problem involves a 3x3 grid with 8 numbered tiles and 1 blank space that can be moved. The A* algorithm maintains a tree of paths from the initial to final state, extending the paths one step at a time until the final state is reached.

What are 5 disadvantages of AI?

What are the potential disadvantages or risks of AI?

• Privacy concerns and ethical problems.
• Cost of implementation and maintenance.
• Environmental issues.
• Hallucinations.
• Lack of transparency.

What are the 3 C's of writing?

The 3 Cs of writing are most commonly Clarity, Conciseness, and Coherence, fundamental principles that ensure your message is easy to understand, brief, and logically structured, with variations sometimes substituting Consistency, Compelling, or Correctness for one of the core three. They guide writers to eliminate jargon, get straight to the point, and maintain a consistent flow, making writing effective for any audience.
 

How does LLM handle typos?

At its core, the ability of LLMs to handle spelling mistakes stems from their focus on meaning over exactness. They aren't tied down by the rigid rules of spelling; instead, they're guided by patterns, probabilities, and context.

What are large language models bad at?

Despite their impressive capabilities, LLMs struggle with contextual understanding and common-sense reasoning. They can process language based on patterns in the data they were trained on, but they often miss nuances and subtleties. This limitation can lead to outputs that seem out of place or lack logical coherence.

What is the 80 20 rule in machine learning?

By applying the 80/20 Rule to your learning process, you can prioritize the most impactful concepts, algorithms, and techniques to achieve 80% of the results in a fraction of the time. Focus on the fundamentals, master core algorithms, implement effective data preprocessing, and continually refine your models.

Why do 95% of AI projects fail?

As evidence shows, most AI initiatives fail — not because the technology lacks promise, but because it is misapplied, overhyped, left unchecked, or deployed without sufficient training. With 95% of pilots failing in 2025, the message is clear: generic approaches lead to generic failures.

What is a major limitation of LLMs?

Limited Knowledge Update

Another significant limitation of LLMs is their inability to acquire new information after their initial training phase. This static nature means that the information they provide can become outdated, potentially leading to the dissemination of stale or inaccurate data.

What is the 30% rule in AI?

The 30% rule in AI is a practical framework that says you should start by automating roughly 30% of your repetitive tasks—the ones that eat up time but don't require human creativity or judgment. This focused approach delivers the biggest ROI while avoiding the chaos of trying to automate everything at once.

What is something AI can never replace?

Focus on developing skills AI cannot easily replicate, such as critical thinking, creativity, emotional intelligence, problem-solving, and leadership.

What was Stephen Hawking's warning about AI?

In a BBC interview in 2014, Hawking warned: “The development of full artificial intelligence could spell the end of the human race.” While the technology was only just beginning to emerge, Hawking had the foresight to theorise how it might develop and impact our lives, especially if it exceeds human intelligence.

What does AI struggle with the most?

AI excels at pattern recognition, but real-world problems often lack clear-cut solutions. For instance, AI can easily solve structured math problems but struggles with open-ended questions that require interpreting ambiguous data, ethical considerations, or creative reasoning.

What is one example of weak AI?

Whether you are aware of it or not, you experience artificial intelligence almost every day. Most likely, it's Weak AI (also called Narrow AI). For example, if you use a virtual assistant technology such as Siri or Alexa that you access through a smart device, you are tapping into Weak AI.

What does the Bible say about artificial intelligence?

The Bible doesn't mention Artificial Intelligence (AI) directly, but its timeless principles guide how Christians should approach it, emphasizing that humans, made in God's image, can use their God-given intelligence for good (stewardship), while exercising wisdom, avoiding pride, and ensuring AI serves humanity rather than becoming an idol, with some interpretations linking Revelation's "beast" imagery to potential deceptive AI, though this isn't a universal view.