Sure, let's simplify it!

You know how sometimes we humans can answer questions or solve problems in many different ways? Some people are really good at remembering facts (like a superhero with a big brain), and others are better at thinking through problems step-by-step (like a detective).

Now, computers have something similar, but they need help. The computer can:

1. **Remember lots of things**: This is like when it searches for information to answer your questions. It's really good at this!

2. **Use steps sometimes**: For more difficult questions that need step-by-step thinking, the computer can look at how people have solved similar problems before and try to do something similar. But remember, it's learning these steps from what other people did, not thinking them up all by itself.

Some people think computers are like super smart gods because they can do so many things we usually need many humans for, while others think they're just pretending to be smarter than they really are, like a monkey wearing a tiny suit and tie (this is the 'imitating monkeys' part).

And there's an argument happening among grown-ups about this. Some people say computers will keep getting smarter until they can think on their own, while others worry that maybe they'll only ever be as good as we make them from our examples.

Right now, it's like having a smart helper who can do lots of things if you show them how, but whether or not it can become really, truly as clever as us on its own is still up for debate! Read from source...

Based on the provided text, here are some potential critiques and inconsistencies:

1. **Inconsistency in Comparison:** The author compares human left brain (logical) and right brain (creative) thinking to LLMs' capabilities. However, it's crucial to note that humans don't have a "digital" equivalent of a brain, making this comparison somewhat flawed.

2. **Overgeneralization:** The author implies that all LLMs struggle with reasoning questions without demonstrating procedures or formulas. This statement may not hold true for all models and depends on the specific model's training data and architecture.

3. **Confirmation Bias:** The author selectively mentions critical views on LLMs (e.g., Ferragu, engineer at Google), but does not balance this by mentioning optimistic views or successful use cases of LLMs, which might imply confirmation bias.

4. **Lack of Nuanced Understanding:** The author oversimplifies complex arguments about LLM capabilities and AGI's progress. For instance, the claim that OpenAI hindered AGI progress is a contentious one and depends on various factors, such as the resources invested in research compared to commercial deployments.

5. **Emotional Language:** Parts of the text use emotional language (e.g., "digital god", "glorified digital imitating monkeys"), which can make it feel more like an opinion piece than a neutral analysis.

6. **Lack of Clear Thesis or Argument:** The article starts with Ferragu's tweet, but doesn't provide a clear thesis or argument about LLMs' capabilities. Instead, it mentions various opinions without connecting them together to form a cohesive narrative.

Neutral. The article presents a balanced view of LLMs' capabilities and their limitations, without excessively emphasizing either side. It includes opinions from both proponents and critics of LLMs, which cancels out any extreme sentiment.