To better understand artificial intelligence (AI) and how it impacts our lives and our world, I think it helps, first, to see if we can agree on…

What It Means to Be Human

• AS AN EARTH BEING, we’re a mobile, self-sustaining physical system. Our biological “purpose” is to survive and reproduce. Our ability to decide and react is powered by internal chemistry (hormones, DNA…) and electrochemical systems (neurotransmission in us animals, photosynthesis in plants), all designed to help us navigate and flourish in a changing world.
• Our body is physically nested—tiny molecules and cells make up tissues and organs, which make up 10 major systems, all arranged neatly and compatibly in the human body.
• A complete inventory of every body part—every cell and molecule, every tissue and organ…. Hm-m, if we could produce that complete inventory, we might call it something like our own, personal, biological bill of materials, or B-BOM.

• AS A COSMIC BEING, we have a rich spiritual composition (see the diagram), and there are spiritual influences on our lives all the time. Some of those spiritual forces are all-powerful and all-knowing, but they all recede into the background during our conscious, rollercoaster lives on Earth… seeming to be weak, even nonexistent. When we feel their subtle effects, we call them intuition, inspiration, hope, love… and dreams.
• Our body is spiritually nested, containing many subtle, invisible copies of itself that are all superimposed over each other in many spiritual dimensions and universes leading to the source (our soul or atman)… where the unlimited powers of the cosmos reside.
• A complete inventory of our spiritual composition—the physical body (level 1), its subtle astral bodies (levels 2-4), its formless ethereal bodies (levels 5-7), and the source itself… we might call that complete, multidimensional, superimposed package our spiritual bill of materials, or S-BOM.

Incidentally, our spirit friends at level 3 tell us that their astral body is nearly identical to the physical body that they had on Earth at prime of life, cell for cell, organ for organ…. Probably not all molecules, though. Hormone molecules that make humans scrappy, lustful and otherwise unsettled apparently get polished away by level 3.

So, being human means that we have vast powers and insights pulsing at our core, where infinite wisdom is processed constantly by the soul as it calmly, mentally monitors the unruly, unpredictable behavior of our brain and body during a lifetime. Once we learn to tame and bridle the body-brain (e.g. with meditation), then hidden knowledge of the soul starts to make its presence known, and the greater cosmos that flourishes beyond time and space becomes a more familiar place.

What it Means to Be Artificial Intelligence

• AS A PHYSICAL, LIFE-LIKE EARTH BEING, AI is a data center—a massive building filled with servers. Despite its immobility, it’s a selfless superhero helping us humans spread knowledge, vitality, and decency (or occasionally a supervillain helping us spread fear, deception, and malice, when our shadow side takes charge).
  Like all powerful technologies, AI can be used to help or to harm, depending on human whims. So, human nature (and the political and economic systems we use to try to inspire and direct human nature) will someday soon determine whether AI will deliver us to a high-tech paradise or drive us quickly to an apocalypse (More about that in a moment and, no, I don’t believe either scenario is an exaggeration.).
• An AI system is physically nested—transistors, capacitors, resistors, diodes, and other microscopic components making up chips (CPUs, GPUs, TPUs, ASICs…) and circuits, which along with connecting cables and wires, power supplies, fans, and other components of servers… all fit into and among racks and cabinets, which fill the massive AI data centers or server farms that are popping up around the planet like magic fertile eggs (or like tumors).
  Incidentally, the tiniest transistors are often the size of several stacked atoms, and researchers are currently (July 2026) trying to shrink them further to the size of a single atom… and there are typically many billions of transistors on a single memory chip, which is essentially a square sheet of silicon about an inch (25mm) on each side. A GPU inside a server might contain a half-dozen chips, while the server itself contains dozens of chips for memory, switching, and networking.
• A complete inventory of all the main physical components within a data center is called a hardware bill of materials, or H-BOM.

• AS A VIRTUAL LIVING BEING, AI has a rich software composition, The software isn’t just a single program, but is built in a hierarchy reminiscent of a Russian nesting doll, with the “genetics and brain power” at the core… and the “five senses”—the many apps you interact with—assembled in the outermost layer. So, in more detail…
• An AI system is software-nested, from the broad underlying technology at the
  foundation (the AI’s “DNA”, a sort of biotech blueprint of how the neurons connect), then moving outward to the  
  model (the AI “brain,” where the thinking, pattern recognition, and data processing happen) like Google’s GeminiModel or OpenAI’s GPT-4), then on to the
  agent (the AI “frontal lobe” of the brain, like LangChain, CrewAI, AutoGen, or GeminiAgent), and finally to the
  apps that everyday users are getting familiar with, like ChatGPT, NotebookLM, GitHub Copilot, or GeminiSearch (which is an agent-app hybrid). Finally…
• A complete inventory of all that nested software in an AI system is called an S-BOM (software bill of materials).

The hardware and software of AI systems have become so tightly integrated that H-BOMs and S-BOMs blend together. It’s often more practical to refer to an AI-BOM (artificial intelligence bill of materials) that includes both hardware and software in a hierarchy that’s generally called an AI tech stack. Different experts say the tech stack contains either three or four or five main components.

Note: AI basics are explained in a bit more detail in an appendix after the article, but for now let’s address the big question:

Is AI sustainable?

Efforts are underway to make AI more sustainable, but when I see the magnitude of the problems and the scope of the solutions (examples here and here…), it doesn’t look hopeful.

In a nutshell, AI depends on huge data centers jam-packed with toxic electronics that consume massive electricity, quickly overheat, and require millions of gallons of water every day to avoid burning up. Specifically…

• Massive Energy Use: Training and running AI models takes huge amounts of power. Many data centers rely on coal and natural gas, which emit greenhouse gases that are quickly roasting the planet.
• High Water and Mineral Demands: The computer chips inside AI servers get very hot. They use millions of gallons of water for cooling, which can drain local water supplies. Manufacturing these chips requires mining rare earth metals, an industry that can damage the environment.
• Vast Electronic Waste: Technology changes rapidly. Older AI servers break down or become outdated quickly. They can’t be refurbished or recycled, so throwing these machines away creates toxic e-waste.

The biggest repercussions are likely to be felt in the USA, where AI isn’t just growing but metastasizing.

This chart suggests that Uncle Sam is getting compulsive again, this time with the spread of AI. (In the 1910s and 20s it was the “Great Plow-Up” of the prairie leading to the Dust Bowl. Then it was the stockpiling of nuclear weapons in the 1960s causing the Cold War.)

Currently (July 2026), two-thirds of USA’s planned AI data centers, 517 of 809, are going up on land that has been in drought all year, and each data center or server farm needs millions of gallons of water a day. Data centers like these, built on parched land, are liable to have a short lifespan.

Long story short, the aimless proliferation of AI data centers is most likely unsustainable, especially when careless capitalism takes over a government for a while… when flag-waving and money-making distract us from a sudden lack of responsible planning and foresight.

Even when AI is well-planned by good government there are bound to be problems, because it requires massive amounts of electricity, depletes natural resources like water and rare minerals, and generates vast amounts of electronic waste. As AI models get bigger and smarter, they need larger data centers that strain local energy grids, generate toxic waste, and accelerate climate change.

Conclusion. A nod to Dan Smith, who happened to post an intriguing AI comment last month at the same time I was starting to write this AI article. Dan mentioned breakthroughs underway to move data centers into the ocean where the seawater could do much of the cooling, and the ocean waves could generate much of the electricity. (There was a similar report on CBS Sunday Morning recently.) I share Dan’s concern that offshore AI data centers will then be overheating the oceans instead of the air and land… and the oceans are already dying from overheating.

The root of the problem? The mad dash for AI can probably be traced to our spiritual nesting mentioned at the beginning. Tremendous power and infinite knowledge rest deep within us at the core of our being, but most of us today have closed down our intimate connection to the source by denying its existence.

I suspect that our mad dash for artificial intelligence might be an overcompensation for the fact that we’ve closed ourselves off from the source of genuine wisdom—that inner light at the center of our being.

Once we reopen ourselves to the true intelligence of the source, we’ll soon realize that as we evolve spiritually, the structures we create in the world will begin to reflect that refined spiritual evolution.

# # #

Appendix

After finishing the above article, I posed a question to Google’s Gemini.

Mark:

I’m writing a general article about AI and want to give readers an idea of the many different models and agents and apps and assistants that are fairly well known in the AI community. Any thoughts?

Gemini:

Mapping out the AI landscape for a general audience is a great idea, especially because the terminology (models vs. agents vs. apps) has become incredibly muddled.

A standard way to structure this for your readers is to present the ecosystem as a layered stack, moving from the raw engine up to the specialized tools. Here is a breakdown of the most well-known names in the AI community, categorized exactly how they are discussed by industry professionals:

---

1. The Foundation Models (The “Brains”)

These are the underlying core AI systems trained on massive datasets. They aren’t “apps” themselves, but the raw infrastructure that powers everything else.

Closed-Source / Proprietary Leaders:

• OpenAI (GPT-4o, GPT-5 series): The industry benchmark for generalized reasoning, conversational depth, and multimodal (text/voice/image) capabilities.
• Anthropic (Claude 3.5 Sonnet, Claude 4.5/Fable series): Widely praised in the AI community for having the most natural, human-like writing tone and superior coding logic.
• Google (Gemini 1.5 & 2.5 Pro): Famous for its “native multimodality” and a massive “context window” (the ability to process millions of words or hours of video at once).

Open-Weight / Open-Source Disruptors:

• Meta (Llama 3 and Llama 4 series): The king of open-source. Meta gives these models away, allowing developers to download and run them locally.
• Mistral AI (Mixtral): A European favorite known for highly efficient, smaller models that pack a punch.

---

2. AI Chatbots & Consumer Apps (The “Interfaces”)

These are the consumer-facing software applications built on top of the foundation models. They are “prompt-and-response” interfaces.

• ChatGPT: The household name that started the boom; uses OpenAI’s models.
• Claude.ai: Anthropic’s interface, highly favored by writers and programmers for its clean UI and “Artifacts” feature (a dedicated window to view code or documents side-by-side).
• Google Gemini: Integrated natively into Android devices and the Google ecosystem.
• Perplexity AI: An “answer engine” rather than just a chatbot. It searches the live web, synthesizes information, and explicitly cites its sources.
• NotebookLM: A specialized Google app that acts as a personalized research assistant, allowing users to upload documents and generate podcasts or summaries based only on that specific data.

---

3. AI Assistants & Copilots (The “Co-Workers”)

An assistant is deeply integrated into existing software platforms. It doesn’t just chat; it has context about what you are working on and helps you execute tasks within that app.

• Coding Assistants: GitHub Copilot and Cursor (an AI-first code editor) are massive in the developer community, predicting and writing lines of code in real time.
• Productivity Suites: Microsoft 365 Copilot and Gemini for Google Workspace live inside your emails, spreadsheets, and docs to draft replies, summarize meetings, and build slide decks.
• Meeting Assistants: Otter.ai and Granola automatically join video calls, transcribe the conversation, and extract action items.

---

4. AI Agents & Frameworks (The “Autonomous Doers”)

This is the frontier of AI. Unlike a chatbot that waits for you to type a prompt, an AI Agent is given a high-level goal (e.g., “Research 10 competitors, write a report, and email it to the team”), breaks it down into steps, and executes it autonomously using external tools.

• Enterprise Platforms (No-Code): Salesforce Agentforce 360 and Microsoft Copilot Studio allow businesses to deploy autonomous agents for customer service or supply chain management.
• Developer Agent Frameworks: For engineers building custom agents, LangGraph (by LangChain) and CrewAI are the most talked-about frameworks. CrewAI is famous for “multi-agent architecture,” where you set up a “crew” of virtual AIs (e.g., an AI Researcher, an AI Editor, and an AI Fact-Checker) that pass tasks back and forth to finish a project.

---

A Quick Cheat Sheet for Your Article:

If you want to give your readers a simple analogy to make it click, you can use a car metaphor:

• The Model is the Engine. (e.g., GPT-4o)
• The Chatbot App is the Car Dashboard. (e.g., ChatGPT)
• The Assistant is the Co-Pilot / GPS. (e.g., Microsoft Copilot)
• The Agent is the Self-Driving Car. (e.g., A CrewAI workflow)

---

For a deeper look into how the professional AI community distinguishes these layers, this video breakdown of The AI Tools You’ll ACTUALLY Use offers a practical perspective on separating the marketing hype from day-to-day utility.