AI Critical Thinking: Will AI Save Us or Destroy Us?
- Karina
- Nov 16
- 12 min read
We ask AI to solve our biggest problems - but isn’t it quietly making others worse?
That’s the paradox of our moment.
On one side, AI looks like the ultimate tool: fast, tireless, capable of analyzing more data in an afternoon than a human could in a lifetime. On the other, the very systems we celebrate are energy-hungry, resource-intensive, and strangely entangled with the environmental and social issues we hope they’ll fix.
And when you zoom out, there’s a strange tension in the air.
We’re rushing to adopt a technology that feels intelligent, sometimes even wise, yet we know disturbingly little about the hidden costs of relying on it. From the water it consumes to cool its servers, to the subtle ways it shapes our thinking, to the research suggesting that frequent AI use may actually reduce our cognitive effort and increase dependence, there’s an undercurrent we’re only beginning to acknowledge.
Before we go further, maybe we should pause and ask a more foundational question:
AI stands for Artificial Intelligence - but is it actually intelligence?
A traditional definition of intelligence is:
“the ability to acquire and apply knowledge and skills.”
AI certainly acquires vast amounts of knowledge. But can it acquire skills? Can it truly apply what it knows? Or does it only simulate application well enough to convince us? At least in this generative phase, the answer isn’t obvious.
And that uncertainty is exactly we will explore: the promise, the risk, the beauty, the contradictions.
So let’s look at the whole picture: the problems AI tries to solve, the ones it may be creating, and the deeper question of what happens when we build a non-intelligent intelligence and start treating it as if it were our own.
What Do We Mean by “AI”?
When most people talk about AI today, they’re really talking about generative AI, which is powered by large language models (LLMs). These systems ingest vast amounts of text, learn statistical patterns, and then generate responses that look intelligent. They can acquire staggering amounts of information, far beyond human scale - but they don’t actually understand it. They don’t reason, contextualize, or apply knowledge the way humans do. What feels like critical thinking is more like extremely advanced pattern prediction. Useful, yes. Impressive, absolutely. But fundamentally different from human intelligence.
It also matters how these models work behind the scenes. AI has two phases: training and inference. Training is the resource-intensive phase where the model “learns” from huge datasets. This is where the majority of energy consumption, water usage, and environmental cost is generated. Inference is what happens when you use it, when you ask a question or generate text. Inference is cheaper, but it scales infinitely with global demand. That’s the trade-off: even if each request becomes more efficient, billions of daily requests multiply the impact. Understanding this distinction helps us see why AI feels magical at the surface yet carries enormous hidden costs underneath.
The Promise and Peril of AI: A Snapshot
AI is not inherently good or bad, it’s a tool with enormous potential on both sides of the spectrum.
Its benefits can be transformative, but its costs, environmental, cognitive, social, are becoming too large to ignore.
So here is the quick, honest snapshot: what AI gives us, and what it quietly takes away.
AI Pros & Cons - In a NutshellAI Pros & Cons - In a Nutshell
The Pros of AI
Here are some of the most compelling benefits of AI — and real-world examples that show how they’re already being used for good.
1. Accelerated Data Analysis & Research
AI can analyze enormous datasets far more quickly than humans, enabling breakthroughs in science and medicine. A flagship example is AlphaFold, developed by DeepMind. AlphaFold predicts the 3D structure of proteins with very high accuracy, which has dramatically accelerated biological research and drug discovery.
In one study, researchers used AlphaFold’s predictions together with generative chemistry to identify a novel inhibitor for CDK20 — a target for drug development — in just 30 days.
This kind of acceleration would be nearly impossible without AI’s ability to process and model molecular structures at scale.
Another high-impact case: the 2024 Nobel Prize in Chemistry was awarded to researchers who used AI to design new proteins.
These AI-driven techniques could lead to new medicines, enzymes, or materials.
2. Enhancing Productivity
AI tools automate many repetitive or time-consuming tasks, freeing people to focus on strategic, creative, or higher-value work.
For example: in writing, marketing, coding, or design, generative AI can draft, iterate, and polish much faster than a human alone.
In the healthcare domain, companies like Tempus use AI to evaluate medical data (genomic, clinical, imaging) to help doctors make faster, more informed decisions.
By reducing the time spent on data analysis and interpretation, AI helps practitioners focus on patient care.
3. Augmenting Human Intelligence / Decision Making
AI doesn’t just replace labor; in many contexts, it augments human expertise —- helping doctors, researchers, and decision-makers do things they couldn’t easily do alone.
In healthcare, OpenEvidence is an AI-driven platform that helps physicians navigate medical literature by synthesizing and organizing research from clinical journals. This makes clinical decision-making more evidence-based and efficient.
Another company, Aidoc, develops AI systems to assist in medical imaging: their algorithms help detect conditions such as stroke or hemorrhage in CT scans, speeding up diagnosis in hospitals.
4. Democratizing Access to Knowledge
AI lowers the barrier for people to engage in skills like writing, strategy, and analysis — making expert-level tools more widely available.
For instance, generative models (like language models) let non-experts draft complex reports, essays or proposals. Instead of needing years of training in writing or research, people can use AI to generate outlines, polish their text, or brainstorm ideas - thereby equalizing access to certain types of cognitive labor.
5. Optimizing Systems: Energy Grids, Logistics, Infrastructure
AI is already making infrastructure smarter and more sustainable by optimizing how energy is generated, stored, and used.
AutoGrid, for example, uses AI and reinforcement learning for smart microgrid management. Their system forecasts renewable generation (solar, wind), predicts demand, and optimizes dispatch to reduce costs and maximize clean energy usage.
AI platforms like those described by BCG can improve renewable energy companies’ operational efficiency by 15–25%, through better forecasting, maintenance, and virtual power-plant strategies.
Tools such as Grid4C and Bidgely predict demand and consumption patterns in real time, helping utilities reduce waste and better match supply to demand.
6. Potential Climate & Environmental Benefits
Beyond just optimization, AI can help accelerate the transition to renewable energy and improve efficiency in clean-energy systems.
IBM’s Smarter Energy platform, for instance, leverages AI to forecast weather and adjust the operations of wind and solar farms, helping maximize their output and minimize curtailment.
AI is also used for predictive maintenance: thanks to data from sensors on solar panels or wind turbines, operators can detect anomalies early, prevent failures, and extend asset life.
On a system level, AI helps integrate decentralized energy sources and microgrids, increasing resilience and reducing reliance on fossil-fuel backup.
7. New Opportunities & Industries
AI isn’t just making existing jobs more efficient — it’s spawning entirely new sectors.
AI-driven biotech: Companies like Owkin combine patient data, genomics, and clinical research to develop new diagnostics and therapeutics using AI.
Cleantech / energy AI: Startups such as Capalo AI are building “virtual power plants” – AI platforms that predict renewable generation and consumption, optimize battery storage, and help flatten demand curves.
Smart buildings / industrial IoT: Firms like Verdigris Technologies use AI + IoT to monitor building energy usage in real time, identifying inefficiencies and reducing consumption.
The Cons of AI
For all the breakthroughs AI promises, its costs: environmental, cognitive, economic, and social, are becoming impossible to ignore.
1. Environmental Footprint
Elon Musk’s xAI is building a massive compute facility outside Memphis - a project that triggered investigations into air pollution, unpermitted gas turbines, and environmental risks for nearby residential communities.
Why It Matters
Generative AI scales exponentially, not linearly.
Data centers require huge amounts of electricity and freshwater.
Many are built in economically vulnerable areas with limited environmental oversight.
2. Cognitive Decline & Mental Health
Early research shows that frequent use of AI tools can weaken problem-solving stamina and deepen passivity - people stop thinking before they prompt.
Articles:
Why It Matters
Over-reliance dulls critical thinking “muscles.”
Constant comparison to machine outputs increases anxiety and self-doubt.
Affects younger users most, especially students who replace learning with prompting.
3. Hallucinations & Inaccuracy
A New York federal judge sanctioned attorneys who submitted fabricated citations generated by ChatGPT.
Why It Matters
AI sounds authoritative even when wrong.
In fields like health, law, and finance, errors = real harm.
Infrastructure failures (e.g., AWS outages) ripple across everything built on top.
4. Bias & Discrimination
AI models repeatedly reproduce gender and racial stereotypes. Even in high-stakes contexts such as hiring or judicial analysis. UNESCO report & Stanford “Foundation Models” bias analysis.
Why It Matters
AI doesn’t merely mirror society’s biases - it amplifies them.
Unequal representation → discriminatory outputs.
Deploying biased models at scale multiplies structural inequity.
5. Job Displacement & Economic Shockwaves
2023–2025 saw hundreds of companies publicly cite AI as a reason for cutting roles - from media writers to support agents to coding teams.
Why It Matters
Middle-skill workers are hit fastest.
New jobs require specialized skills that many displaced workers lack.
Transition period = increased inequality and social instability.
6. Concentration of Power
Cloud capacity, cutting-edge chips, and top-tier models are controlled by a tiny cluster: OpenAI/Microsoft, Google, Amazon, Meta, and Nvidia.
Why It Matters
Democratic oversight becomes nearly impossible.
Pricing, safety standards, and access are set by private interests.
Centralization → systemic fragility (one outage = global disruption).
7. Public Health & Localized Harm
Investigations into multiple U.S. and EU data centers found:
Noise pollution violating residential limits
Diesel backup generators raising particulate levels
PFAS contamination concerns (The Guardian)
Why It Matters
Local communities pay environmental and health costs.
Often affects poorer, nonwhite neighborhoods (environmental injustice).
Expansion of AI = expansion of these risks.
Trade-offs and Tensions
AI forces us to face a series of uncomfortable trade-offs. It promises astonishing efficiency, but efficiency almost never comes without a shadow.
The paradox of efficiency.
Even as AI models become more efficient per query - less energy, less water, smaller carbon footprint, the total demand increases radically. This is the classic Jevons Paradox: when you make something cheaper or easier to use, people use much more of it. AI might one day use half the energy per request, but if the world makes 100× more requests, the net impact still grows.
Short-term vs. long-term cost.
Training a cutting-edge model requires staggering compute, energy, and water. But once trained, inference scales cheaply - which means more people, more industries, more governments lean on AI for everything from homework to national security. The long-term challenge is not simply “how expensive training is,” but how infinite our appetite becomes once the model exists.
Local vs. global impacts.
While AI benefits are global, the environmental burden is strikingly local. Data centers concentrate water consumption, heat pollution, chemical contamination, and noise in specific regions. Communities in Arizona, Oregon, Chile, Ireland, and rural Spain have protested water shortages and pollution tied to hyperscale data centers. Some towns bear the environmental cost so the rest of the world can enjoy frictionless digital convenience.
Who benefits? Who pays?
As with most technologies, access is not equal. A small number of very wealthy companies and individuals own the infrastructure; the rest of the world becomes dependent on their systems. Equity becomes a central tension: can AI truly benefit all, or will it widen the gap between the powerful and everyone else?
And yet… I
t’s worth remembering a broader perspective: despite everything, humans live longer, healthier lives than at any point in history.
Technology has always amplified both our wisdom and our recklessness. Used wisely, it enhances our healthspan, safety, and quality of life.
After all, life itself remains the most sophisticated intelligence we know, and technology is ultimately its extension. The question is not whether AI is good or bad, but how we choose to use it.
Mitigating the Risks
If we want AI to lead us toward progress, we need to be intentional. Solutions already exist, and many more are emerging.
Technological Solutions
Efficient cooling: immersion cooling, closed-loop systems, and on-site heat recapture dramatically reduce water use.
Greener data centers: companies are beginning to power AI with solar, wind, geothermal, and even captured industrial waste heat.
Low-carbon AI research: new benchmarks track not only model performance, but energy and water footprint (e.g., Energy / Carbon / Water benchmarks for inference).
Policy & Regulation
Governments are starting to demand transparency on AI’s physical footprint:
Energy reporting
Water usage disclosure
Local environmental compliance
Emissions accountability
The Guardian has reported extensively on data center impacts, particularly in the UK and Ireland, where local communities face measurable strain.
Ethical AI
Developing models that are:
Bias-aware
Transparent
Auditable
Accountable
This includes building models that can explain reasoning, track data lineage, and evaluate their own uncertainty - the building blocks of machine “critical thinking.”
Social Solutions
Reskilling and workforce transition programs
Safety nets for displaced workers
Universal basic income discussions
“Human-in-the-loop” systems to prevent full automation loss
Education reforms emphasizing creativity, synthesis, physical-world skills
Your insight is crucial here: AI also creates entire new career paths. Strategic AI management, prompt engineering, AI governance, data ethics, synthetic data design, and alignment research are already fast-growing fields. Every technological disruption simultaneously destroys old jobs and generates new ones.
Is AI Our Savior or Our Doom?
There is no simple binary here. AI is neither salvation nor apocalypse - it is a magnifier.
It will amplify the intelligence, wisdom, and integrity of those who use it well… and amplify the shortcuts, ignorance, and recklessness of those who do not.
The real question is not whether AI can think critically. It’s whether we can.
Critical thinking becomes the determining factor. Everyone is responsible for deciding how AI enhances or diminishes their life. We need to understand:
what parts of AI help us think better,
what parts make us mentally softer,
and where we risk outsourcing too much of our agency.
Global cooperation becomes essential. Climate, energy, labor, ethics: none of these challenges are solvable within borders. AI forces us into a shared destiny, whether we like it or not.
AI will not save or doom us - our relationship with it will.
Conclusion: If you want to make an omelet you gotta break some eggs.
Recently, I went through a fractional CO₂ laser treatment. For anyone unfamiliar: it’s a procedure that literally burns microscopic layers of your skin so it can rebuild itself - producing fresh collagen, repairing damage, and ultimately making you look younger.
Why am I mentioning this here?
Because transformation often requires destruction.
Muscles grow only after micro-tears. Forests regenerate after fires. Systems evolve when outdated structures collapse. There are moments in human progress when burning something down - is the only way to build something better.
Maybe AI is going through that phase now.
A season of mental-health crises, loneliness, job displacement, environmental strain, and deep societal confusion - before we reach a future of true General Intelligence.
Let’s imagine the world it could help create.
A world where we shut down what harms us: CO₂-heavy concrete plants, unnecessary mass production, and the constant travel pollution. Where we use nuclear-powered propulsion to send probes beyond our solar system - a step toward the Kardashev vision of civilizations that harvest energy from stars.
Where financial wealth is shared fairly, where everyone has access to similar comfort and stability, and where we no longer drain ourselves comparing lifestyles or chasing status. Instead, our attention shifts inward - to creativity, art, learning, and connection.
A world less obsessed with material difference and more focused on human depth, planetary balance, and cosmic curiosity.
Just like the harsh aftermath of a CO₂ laser, before the skin renews itself, perhaps the turbulence we experience with AI is the painful precursor to something more sustainable, and more deeply human.
Unlike the laser, where results are almost guaranteed, there is no certainty with AI.
But we shouldn’t underestimate ourselves.
After all, we began as microscopic organisms floating in ancient oceans - and here we are, building technology that attempts to mimic ourselves.
What’s your take on AI?
What does your “ideal world with AI” look like - one where it truly solves problems rather than creates new ones?
Author: Me.
Yes, I used ChatGPT as a tool, but it did not write this on its own. It would not have written any of this without my direction, my prompts, my ideas, and my voice. The authorship is mine entirely; AI was simply the instrument.
Comments