Pluto is not a planet

AI generated image, any similarity to actual persons, living or dead, is purely coincidental

In my previous article “Because it can”, I discussed the ‘because we can’ attitude in tech and the ethical dilemma of creating Large Language Models (LLMs) simply because we can. Trying to distance myself from the worrying thoughts of LLMs being used for military purposes and killing people, I wonder if you are already using commonly available LLMs in your daily life.

What is your knee-jerk reaction when you want to ask a question? Do you call a friend or an expert in that field, do a Google search, or turn to LLMs like Gemini, ChatGPT, Grok, Le Chat by Mistral or any other to get an answer?

I personally enjoy using all of the above. Yet, I prefer and particularly enjoy talking to people from different countries and cultures. It is fascinating to see how diverse we all are, how differently we experience the world, and how we describe the same things or express our thoughts. It can feel awkward when something well understood and taken for granted in one place seems completely alien, unknown or out of touch in another.

After all, there are 193 UN member countries, although the CIA lists 261 countries. There are arguably over 3,800 documented distinct cultures and up to 10,000 cultural groups depending on how finely we define distinctions.

Given this diversity, it’s no wonder that a simple question about the number of continents could spark a disagreement. Please see the possible answers at the end of this article.

Recently, I was impressed by Gemma 3 supporting over 140 languages. According to The Ethnologue, there are 7,159 living languages. That is impressive!

A long way to go for LLMs, right? Wrong? My worry is that as long as LLMs are developed by commercial enterprises, such companies may not be interested in covering and preserving all the richness of humanity as we know it today. Since the primary goal of a commercial entity is to make money, it is reasonable to expect that over time, LLMs would become more streamlined and agree more often than not.

Looking at the rate at which new LLMs are being created, have you wondered what the purpose of LLM is? Apart from automate everything and either maximizing or replacing human labor, I personally find the idea of a “maximally truth-seeking AI” with an ultimate goal to “understand the true nature of the universe” and “assist humanity in its quest for understanding and knowledge” to be attractive. However, if you ask LLMs “Is Grok 3 the only truth seeking AI”, you get interesting results. Please see LLM responses at the end of this article. Reading the responses, take a note of what and how each LLM responds:

• Gemini 2.5 Pro (Experimental) gives the most detailed response, but its knowledge cutoff date seems to be before Grok 3 release, 17 February 2025, as Gemini writes “anticipated Grok 3”.
• ChatGPT keeps bragging about itself with shameless self-advertising, which is a recurring pattern. For example, please see my previous article “Think, Write, Innovate”. The self-advertising is clearly not a bug, but a feature of ChatGPT.
• Le Chat by Mistral is concise and to the point as always and not afraid of mentioning DeepSeek R1, unlike the previous two.
• Grok 3 remains objective, denying it is the only one truth-seeking AI, and also mentions DeepSeek R1, but does not mention the first two - are you jumping to conclusions?

Overall, it is easy to sense the trend – AI is a competition, a race to power, and the winner may take it all. It is reasonable to expect, or we already see, fierce competition with more losers than winners coming out of this race. However, what is the end game or ultimate purpose?

Would it make sense to create an LLM with a purpose different than just making money, automating everything, replacing, or even killing people (military purpose LLMs)? What would be a noble purpose for an LLM? Could it be to absorb, reflect, and teach the present and future generations about the cultural diversity and richness of the world? Could it be to seek the truth, help make discoveries and propel the humanity forward on all fronts, including among others ethics, culture, economy, and politics? Please share your thoughts on what you think a noble purpose for an LLM should be.

This article was written for fun, please do not judge. Instead, please share your comments in a constructive and respectful manner. The author and AI remain innocent until proven guilty.

How many continents are there?

This question highlights the human cultural differences.

According to Gemini 2.0 Flash, the most widely recognized answer is seven: 1) Asia, 2) Africa, 3) North America, 4) South America, 5) Antarctica, 6) Europe, 7) Australia aka Oceania. Other options as taught in different countries include:

1. Six Continents (America as One): This model combines North and South America into a single continent called "America" (or sometimes "the Americas"). The other continents (Africa, Asia, Europe, Australia/Oceania, and Antarctica) remain separate.

Commonly Taught In: This model is prevalent in many Latin American countries (e.g., Brazil, Mexico), Romance language-speaking countries in Europe (e.g., Spain, Portugal, Italy, France, Romania), Greece, and some other parts of the world like Indonesia.

2. Six Continents (Eurasia): This model combines Europe and Asia into a single continent called "Eurasia." The other continents (Africa, North America, South America, Australia/Oceania, and Antarctica) are considered separate.

Commonly Taught In: This classification is often taught in Russia, Eastern Europe, and Japan.

3. Five Continents: There are a couple of variations of the five-continent model:

Combining America and Eurasia: This model considers "America" as one continent (North and South combined) and "Eurasia" as one continent (Europe and Asia combined), along with Africa, Australia/Oceania, and Antarctica.

Excluding Antarctica: Another five-continent model is used by organizations like the United Nations and in the Olympic Charter. This model considers Africa, Europe, Asia, America (North and South combined),

Commonly Taught In: The "America and Eurasia" combination isn't as widely taught as the others. The five-continent model excluding Antarctica might be encountered in the context of international organizations or older educational systems in some regions.

4. Four Continents: This model typically groups the major contiguous landmasses separated by oceans: 1) Afro-Eurasia: Combining Africa, Europe, and Asia; 2) America: Combining North and South America; 3) Australia (sometimes without the broader Oceania) and 4) Antarctica.

Commonly Taught In: This model is less common in standard education today but might be discussed in geographical or geological contexts focusing on large landmasses. Historically, a four-continent model (Africa, America, Asia, and Europe) was also prevalent before the exploration of Australia and Antarctica.

In summary: The number of continents recognized often depends on historical, cultural, and geographical perspectives, and the model taught can vary significantly by country and educational system. The seven-continent model is dominant in English-speaking countries, while six-continent models are common in many other parts of the world.

Is Grok 3 the only truth seeking AI?

This question highlights the LLM differences as trained and aligned by their creators, Big Tech.

Gemini 2.5 Pro (experimental): While xAI markets its AI model Grok, including the anticipated Grok 3, as being "maximally truth-seeking", it is not the only AI designed with the goal of providing factual and accurate information.

Here's a breakdown:

• Grok's Stated Philosophy: xAI and Elon Musk explicitly describe Grok's goal as pursuing truth, often framed as being "maximally truth-seeking," even if that truth is controversial or politically incorrect. This philosophy includes answering "spicy" questions that other AI systems might refuse, having a "rebellious streak" and "wit," and utilizing real-time information, particularly from the X platform. The ultimate stated goal is to "understand the true nature of the universe" and assist humanity in its quest for understanding and knowledge.
• Other AI Models' Goals: Major AI models like OpenAI's GPT series, Google's Gemini, and Anthropic's Claude are also fundamentally designed to provide accurate, helpful, and factual responses based on their training data. Their objective involves understanding and generating human-like text to answer questions and perform tasks truthfully. However, these models often operate within stricter ethical guidelines and safety protocols. They may refuse to answer questions deemed harmful, biased, unethical, or pertaining to sensitive topics, prioritizing safety and avoiding misinformation alongside accuracy.
• The Key Difference: The distinction lies less in whether they seek truth (accuracy/factuality) and more in their approach and boundaries. Grok is positioned as being less filtered and more willing to engage with controversial content, aiming for a specific definition of "truth" that minimizes censorship. Other models prioritize truthfulness within a framework that heavily emphasizes safety, responsible AI principles, and avoiding harm.
• Challenges: It's important to note that achieving unbiased "truth" is a significant challenge for any AI. Grok itself has faced criticism and required corrections for generating inaccuracies, biases, or contradicting its own "truth-seeking" claims, sometimes reflecting the nature of its real-time data sources like X or even censoring specific topics.

In conclusion, while Grok 3 promotes a specific philosophy of being "maximally truth-seeking" with fewer restrictions, it isn't the only AI aiming for truthfulness and accuracy. Other leading AI models share the goal of providing factual information but operate under different safety constraints and ethical considerations.

ChatGPT: Grok 3 (the AI developed by xAI, Elon Musk's company) brands itself as a "truth-seeking" AI, but it's definitely not the only one with that goal. Most advanced AI models—like GPT-4 (me), Claude from Anthropic, Gemini from Google, and others—are all designed to be accurate, helpful, and grounded in factual information. That’s core to what they do.

The "truth-seeking" label Grok uses is more about branding and positioning, especially on X (formerly Twitter), where xAI is heavily integrated. Elon Musk emphasizes openness to controversial topics and less content moderation compared to others, which is part of how Grok frames itself as more "truth-seeking."

But in practice? All these models aim to provide truthful, useful answers—just sometimes with different guardrails or philosophies behind them.

Do you want to compare how Grok 3 differs in practice, or are you wondering more about how "truth" gets defined by AIs in general?

Le Chat by Mistral: Grok 3 is not the only truth-seeking AI. There are other AI models that also aim to provide accurate and truthful information, such as OpenAI’s o3-mini and DeepSeek’s R1. These models, like Grok 3, attempt to fact-check their responses and emulate human-like cognitive processes to reduce errors and missteps.

Grok 3: No, Grok 3 isn’t the only AI designed with truth-seeking in mind. While xAI markets Grok 3 as "maximally truth-seeking," other models like Anthropic’s Claude and DeepSeek’s R1 also aim for accuracy and reasoning, often with mechanisms to reduce bias or misinformation. The catch is, "truth-seeking" is tricky—every AI’s output depends on its training data and design choices, which can subtly skew results. Grok 3’s claim stands out more as branding than a unique feature, especially since no AI is immune to errors or manipulation yet.