Showing posts with label Artificial Intelligence. Show all posts
Showing posts with label Artificial Intelligence. Show all posts

US-China Tech Race: China Has Not Won

A cracked silicon wafer marked with US semiconductor text sitting beside glowing digital citation graphs and shipping cranes in a port setting.



 I sat across from a commercial banker at a noisy coffee shop near I.I. Chundrigar Road last November. He stirred his tea, leaned over his laptop screen, and pointed to a colorful research chart showing Beijing's massive surge in international patent filings under the World Intellectual Property Organization. "The Americans lost," he told me, clicking through the ASPI Critical Technology Tracker to highlight China leading in 66 out of 74 advanced sectors. I watched his finger scan the screen while traffic roared outside on the street. That conversation stuck with me because it exposes a deep flaw in how Western analysts measure technological dominance today.

Academic publishing metrics create an optical illusion for people looking at raw totals. China built an extraordinary paper factory over the past two decades. The Chinese Ministry of Education tied faculty promotions and cash bonuses directly to publication volume in indexed journals, driving a massive wave of scientific citations. That policy flooded databases like the Nature Index with high-impact research papers. Paper volume shows state incentive structures. It does not reflect actual industrial control on the factory floor.

I spent decades tracking cross-border financial settlements and trade documentation flows from terminal desks. Real power in global technology does not live in academic PDF files. It lives in physical assembly nodes and complex industrial supply chains that take decades to construct.

The Silicon Bottleneck in the US-China Tech Race

The Australian Strategic Policy Institute counts citations, but citation counts cannot forge a silicon wafer. High-level research papers on advanced semiconductors mean very little when you do not possess extreme ultraviolet lithography machines. ASML in the Netherlands remains the sole global producer of those vital EUV lithography systems, maintaining an intricate supply web that takes years to master. American intellectual property sits inside every single machine that leaves their factory floor in Veldhoven. China can publish thousands of research documents on quantum physics, yet its advanced foundries still struggle to manufacture sub-three-nanometer chips at commercial yields without those Western tools.

Consider electronic design automation software, a sector where three American firms command over ninety percent of the global market. Cadence and Synopsys build the digital drafting tables that engineers must use to design complex microchips. Chinese engineers rely heavily on those precise tools to layout their circuit paths. Strip away that underlying software layer, and the domestic patent surge slows down rapidly.

I tracked this exact dynamic playing out when international trade restrictions hit Huawei back in 2019. The company possessed thousands of 5G patents, yet its smartphone market share collapsed within months because it lost access to foreign foundries.

Paper Citations Do Not Equal Industrial Control

Statistical volume acts as a powerful propaganda tool in modern trade negotiations. Bureaucrats in Beijing track international patent counts to secure local government research funding, hitting state quotas designed to project economic progress across overseas media outlets. Western outlets repeat those high paper figures without evaluating real commercial execution on the ground. The United States maintains a decisive grip on frontier artificial intelligence architectures and specialized cloud compute backbones.

This gap between paper metrics and actual industrial strength creates confusion. Analysts mistake academic SEO for sovereign mastery. I have observed how China excels at scaling existing industrial hardware and capturing lower-tier manufacturing markets across emerging economies while still failing to replace foundational hardware bottlenecks. Western alliance networks retain control over these critical chokepoints.

A state can subsidize thousands of academic journal submissions every quarter. It cannot easily subsidize the foundational physics capabilities required to build advanced lithography optics from scratch.

The Fragile Reality of Digital Dominance

Standing at the harbor in Karachi, you see thousands of steel shipping containers stacked beneath heavy iron cranes. You realize quickly that real commerce relies on physical hardware and deep financial rails. Press releases matter little here.

China has not won the US-China tech race. The battle continues as the global economy splits into two separate, uncomfortable tech stacks that refuse to talk to each other. I wonder how long developing markets can balance between these competing digital ecosystems before the underlying infrastructure snaps completely.

When AI Stops Doctors from Thinking: A Quiet Risk in Medical Training

 I keep thinking about the first time a young doctor is alone with a patient.

No supervisor hovering. No WhatsApp group lighting up. Just a stethoscope, a story that doesn’t quite fit the textbook, and that uncomfortable pause where you realise… you have to decide. That pause is medicine. Or at least it used to be.

Lately, that pause is shrinking.


The Quiet Risk Isn’t AI. It’s Obedience

The editorial in BMJ Evidence Based Medicine is careful, polite, academic. But underneath the cautious language is a sharper warning: medicine is drifting from thinking to accepting.

AI doesn’t bully young doctors. It doesn’t shout. It just sounds calm, fluent, certain. And confidence, especially early in training, is intoxicating. You stop asking why because the answer arrives fully dressed.

That’s not efficiency. That’s habit formation.

Medicine was never about retrieving information. It was about holding uncertainty without panicking. AI is brilliant at pattern-matching. Diagnosis, however, is pattern-breaking. The danger begins when students confuse the two.


Deskilling Doesn’t Look Like Failure. It Looks Like Smoothness

Here’s the uncomfortable part. Overreliance doesn’t produce bad doctors overnight. It produces doctors who look fine. Efficient. Up to date. Until something unusual walks in.

Cognitive off-loading sounds harmless. We’ve outsourced navigation to GPS, spelling to autocorrect. But when you outsource reasoning before it fully forms, you blunt it permanently. A trainee who never wrestles with ambiguity won’t suddenly develop that muscle at 35.

And AI’s mistakes are especially dangerous because they’re polite. Hallucinations don’t announce themselves. Bias doesn’t wear a warning label.


Why Training Must Teach Doubt, Not Just Output

One idea from the editorial actually made me pause. Training students on intentionally flawed AI outputs. Forcing them to argue back. To reject confidently delivered nonsense using evidence.

That’s closer to real medicine than most exams.

Grade the reasoning. Not the final answer. Make students explain what they don’t trust and why. Because in the real world, patients don’t care how elegant your tool was. They care whether you noticed what didn’t fit.


A Personal Turn: Watching a New Doctor Step In

My daughter, Maryam Jamal, has just passed her MBBS. Watching that moment was pride mixed with something heavier. Relief, yes. But also awareness.

Young doctors today are stepping into a system flooded with tools their seniors never had. That’s not fair or unfair. It’s just reality. The question is how they use them without letting those tools quietly reshape who they become.

So if I were speaking directly to new doctors like her, I’d say this:

  • Use AI after you’ve thought, not before. Make your own differential first. Then check yourself.

  • Never outsource first principles. Anatomy, physiology, clinical reasoning. These are non-negotiable.

  • Be suspicious of confidence, especially your own.

  • Treat AI like a junior assistant. Helpful, fast, occasionally wrong. Never in charge.

  • Protect the bedside. Communication, examination, judgement. These don’t scale. And that’s the point.


The Question Medicine Has to Answer Now

This isn’t about banning AI or pretending we can roll the clock back. That ship sailed. It’s about deciding what kind of doctors we’re training.

Because when something goes wrong, it won’t be the algorithm sitting with the family, explaining a choice. It will be a human being. A doctor. Alone with that pause again.

The only question is whether we’re still teaching them how to live inside it.

The "Nuke Play" — AI’s Physical Reality Check

 Watching the sunset over the Isar in Munich, one might easily forget that our seamless digital world is tethered to a brutal, industrial reality. While I spend half my year here in Bavaria, the heart of the AI revolution actually beats across the Atlantic. If you ever fly into Washington D.C.’s Dulles Airport, look out the window as you land. Those massive, warehouse-like buildings are not for storing packages; they are data centers, the biggest concentration of them anywhere on the planet. This is the physical heart of the Artificial Intelligence boom. It is a tangible reality that challenges the common perception of AI as a purely digital, ethereal concept. We often treat algorithms as ghosts in the machine, yet they require a heavy, industrial skeleton to function. AI physical infrastructure is the anchor that prevents the digital dream from drifting into irrelevance.




The Industrial Scale of AI Physical Infrastructure

The scale of the physical infrastructure boom ignited by AI is difficult to overstate. Loudoun County, Virginia, currently possesses more data centers than any other region on earth. This concentration represents an accelerating capital investment that defies historical precedent. Fueled almost entirely by retained earnings rather than debt, the quarterly expenditures of Meta, Microsoft, Google, and Amazon are projected to approach a combined total of $100 billion by 2025.

This spending spree has become a critical pillar of the American economy. Investment in information processing equipment accounted for over 90 percent of economic growth in the first half of 2025. It isn't just a tech story. It is the engine keeping the U.S. economy afloat. However, this mountain of silicon casts a long shadow. Can our current power grid actually sustain this level of unbridled growth?

Energy Constraints and the Looming Power Wall

The serenity of the Bavarian landscape stands in stark contrast to the aggressive industrialization occurring in Virginia's data center hubs. This construction is colliding with the finite limits of our electrical grid. We no longer measure AI capacity in abstract code; we measure it in megawatts and gigawatts. The energy demand is immense. It is growing at an exponential rate.

Consider the sheer scale. A single server rack can consume more power than a small village. A single new data center may require 1 gigawatt of energy. According to Epoch AI, the power required to train frontier models is doubling annually. This relentless pace is creating "transmission bottlenecks" in hubs like Northern Virginia. To understand the gravity of this shift, imagine a high-speed train attempting to run on wooden tracks designed for a horse and carriage. The friction is inevitable.

CompanyCapex-to-Revenue (%)
Meta35%
Microsoft28%
Alphabet21%
Utility Average28%

This capital intensity is forcing a metamorphosis. Big Tech is no longer "asset-light." These firms now spend on AI physical infrastructure at rates that rival or exceed traditional utility companies. They are becoming the very industrial giants they once replaced.

The Unavoidable Physical Reckoning

The AI revolution's primary challenge is no longer computational; it is fundamentally physical. We are witnessing a collision between exponential digital demand and the slow, stubborn reality of the global power grid. It is a reckoning that cannot be avoided by cleverer coding or more efficient software.

The search for reliable energy has become a strategic obsession. Radical ideas, such as building dedicated nuclear reactors for data centers, are now mainstream boardroom discussions. This is the "Nuke Play." It is the startling, logical endpoint of a boom that has outgrown the virtual world. Whether in the laboratories of Munich or the server halls of Virginia, the message is clear: the future of intelligence is, and always will be, a matter of physical power.

The AI Bubble: Golden Goose or Expensive Pigeon?

 


Artificial intelligence was introduced as capitalism’s next miracle. Generative systems were expected to write emails, reinvent customer service, and even reshape whole economies.

Yet as 2026 approaches, the picture looks different. The goose is not golden. It feels more like a very costly pigeon.

A new MIT study confirms the gap. Ninety-five percent of enterprise AI pilots fail. They produce no meaningful revenue. That single fact changes the story.


Hype Moves Faster Than Reality

Silicon Valley often lives on exaggerated promises. With AI the cycle is faster. OpenAI, Google, and Anthropic release their latest models. Executives rush to adopt them. Budgets are shifted. Road maps are rewritten. Teams lose weeks of focus.

The result is disappointing. Only five percent of pilots bring measurable revenue. For the rest, the dashboards glow brightly while the bottom line stays flat.


Building or Buying

The study also found a pattern. Companies that buy specialized tools succeed twice as often as those trying to build their own. Yet ambition clouds judgment. Many leaders insist on internal builds.

The outcome is predictable. Projects remain stuck in beta for years. By the time they function, a newer model already exists, leaving the tool obsolete. It feels like chasing a train that never stops.


The Quiet Truth About Spending

Another secret hides in the numbers. Most corporate budgets are directed toward sales and marketing experiments. Email writers, pitch decks, and lead generators consume resources.

The genuine return appears elsewhere. AI improves back-office work. It reduces repetitive tasks. It makes operations smoother. That is where the technology saves money. The arithmetic is clear, but psychology is not. Leaders love the shine of dashboards more than the reality of efficiency.


A Bubble With a Kernel of Truth

The question has been asked often: is AI a bubble? Even Sam Altman of OpenAI admits it is. He calls it a bubble built around a kernel of truth.

That phrase matters. The same thing happened during the dotcom years. The internet was real, but not every dream survived. Pets.com collapsed. Webvan disappeared. Yet giants like Google, Amazon, and eBay emerged. The NASDAQ lost nearly eighty percent of its value between 2000 and 2002, but the survivors became household names.


Déjà Vu in Silicon Valley

The AI cycle feels familiar. Trillions of dollars flow into data centers and startups. Every corporate memo contains the word “AI.” The belief resembles the dotcom fever of two decades ago.

There is no doubt that new tools will appear. They may transform areas we cannot yet imagine. Still, there is a mismatch between expectations and adoption. That gap is the bubble.

Consider OpenAI itself. It leads the industry, but it is still not profitable. That fact should speak louder than hype.

The bubble has not burst. It is stretching. And stretched bubbles seldom deflate with grace

Report Reveals Harvard MBAs Struggling to Get Jobs

Did you know that a Harvard MBA, once considered a ticket to a top job and a six-figure salary, isn't as foolproof as it used to be? Despite being one of the most prestigious business schools, 23% of Harvard graduates in 2024 were still jobless three months after graduation. This number has been increasing, up from 20% in 2023 and just 10% in 2022. It's not just Harvard facing this issue; other top business schools like Northwestern Kellogg and Chicago Booth see similar trends.

A career officer at Howard put it well, "Going to Harvard is not a differentiator anymore; you need the skills to back it up." In the past, simply getting into a top B-school was enough to secure a lucrative career. These days, companies seek additional skills and specialties.

Take McKinsey, for instance. They hired 71 graduates from Chicago B in 2023, but only 33 in 2024. Companies are changing their hiring strategies, opting for smaller, more targeted rounds rather than large campus recruitments.

In India, top B-schools like the Indian Institutes of Management (IIMs) also face challenges. Previously, they reported 100% placements within days, but last year, even after two months, not all placements were complete.

So, what's causing this shift? Three main factors:

1. Artificial Intelligence: AI is reducing the manpower needed for many jobs, including management roles.

2. Oversupply: There are too many MBAs in the market, leading to decreased demand.

3. Specialization: Companies now prefer candidates with specialized skills in areas like digital marketing, data science, or AI.

If you're considering an MBA, remember that an MBA alone might not be enough. Think about upskilling and specialized degrees with strong job prospects. Also, be mindful of the costs involved, especially if you're taking out a loan.

Ultimately, weigh these factors carefully before deciding to pursue an MBA.


Women Experience Diabetes Differently Than Men

  I remember sitting in the waiting room at Civil Hospital Karachi three years ago, watching an exhausted woman in a black chador argue with...