How AI Is Changing the World

Every prediction below came from a real conversation about where AI takes us next. They deserve a proper answer rather than a cheer or a dismissal.

So each one gets the same treatment: the strongest evidence for it, the strongest evidence against it, and a verdict.

Some hold up better than the optimists think. Some do not survive contact with the data at all.

Robots will do all the work

The claim: machines do everything, humans give commands, governments hand out the money.

What supports it. SHRM's 2026 analysis estimates 20% of US employment is already at least half automated — roughly 31 million jobs. That share rose in nearly every occupational group in a single year.

The World Economic Forum projects 92 million jobs displaced and 170 million created by 2030. Goldman Sachs puts around 25% of work tasks within reach of automation.

What argues against it. Note the shape of those numbers. Displacement and creation, not displacement alone.

SHRM found that 60.4% of US employment has at least one non-technical barrier to being automated — licensing, physical presence, legal liability, plain human preference. Only 5.1% is both highly automated and has no barrier at all.

BCG's 2026 work lands in a similar place: 50–55% of US jobs reshaped in two to three years, but only 10–15% genuinely vulnerable.

• Reshaped means your job changes. Most people.
• Replaced means your job ends. Far fewer, and concentrated in clerical and routine white-collar work.

Verdict: directionally right, wildly early on timing. McKinsey's own midpoint for automating half of all work tasks is around 2045, with a range stretching to 2060.

Universal robot-funded income is a political choice, not a technical one. No major economy has legislated anything close.

AI music will end singers

The claim: anyone generates a song in seconds, so singers lose their value.

What supports it. The generation part is already true. Text-to-music tools produce finished, listenable tracks from one sentence. Production scarcity is gone.

Session work, library music, jingles, background scoring — those are commodity jobs, and commodity jobs go first.

What argues against it. People do not only buy songs. They buy artists.

Recorded music has been effectively free since Napster, and near-free on streaming since. That did not kill the industry — it moved the money to live shows and fandom, which are the two things a generator cannot supply.

Concerts are not a delivery mechanism for audio. Twenty thousand people do not stand in a field for the waveform.

Verdict: half right. Generic music becomes worthless. Specific artists become more valuable, because scarcity moves to the one thing that cannot be synthesised — a person other people care about.

AI will end the movie industry

The claim: type "Iron Man vs Superman", get a great film, cinema collapses.

What supports it. Generated video has moved from unwatchable to usable in about three years. Storyboarding, VFX, dubbing and background extras are already being done this way.

Cost per minute is falling fast. That reshapes who can make something.

What argues against it. Two problems, and neither is technical.

• You cannot legally publish it. Iron Man and Superman are owned. Generating them for yourself is one thing; distributing it is infringement.
• Generation is not the hard part. A two-hour film is roughly 130,000 frames that must stay consistent in character, lighting, geography and emotional logic. Prompting is easy. Coherence is not.

And film is a shared ritual. People watch the same thing so they can talk about it. A film only you have seen has no conversation attached to it.

Verdict: the industry changes shape, it does not fall. Budgets drop, independent creators gain enormously, and the studios that own characters become more powerful, because owning the IP is suddenly the whole moat.

AI will replace doctors

The claim: AI has more data than any doctor, so doctors become assistants.

What supports it. The benchmark results are genuinely striking. Stanford's 2026 AI Index reports a multi-agent diagnostic system scoring 85.5% on complex published cases against 20% for unaided physicians.

The FDA has now authorised over 1,300 AI-enabled medical devices. An AMA survey found 66% of physicians already use AI in practice.

What argues against it. This is where the headline number and the reality separate, and it matters.

A systematic review of 83 studies found generative AI's overall diagnostic accuracy was 52.1%. No better than physicians overall — and significantly worse than expert physicians.

Worse, of 1,357 FDA-cleared AI devices, an analysis found only 3 were evaluated against patient outcomes like mortality or readmission. Around 6% were linked to recall events.

That is a field clearing products faster than it can prove they help.

Verdict: assistance yes, replacement no — and not soon. Diagnosis is one slice of medicine. Examination, consent, judgement under uncertainty, and legal responsibility for being wrong all sit with a licensed human.

This section is general information, not medical advice. Decisions about your health belong with a qualified clinician.

Robots will fight the wars

The claim: future wars are fought by machines.

What supports it. Drones already dominate modern battlefields. Loitering munitions, autonomous targeting aids and uncrewed naval and air systems are in service now, not in development.

The trend line here is not subtle.

What argues against it. Wars are fought over territory, and territory is held by people standing on it. Machines can clear a position. They cannot govern one.

There is also no international agreement on autonomous lethal force, and the debate is active rather than settled.

Verdict: the most likely of all twelve, with a caveat. Machines will do more of the killing. Humans will still do the occupying, the negotiating and the dying.

If you want to track this as it moves, we maintain a live view: the AI risk score at World Tension Meter.

Artificial food becomes the default

The claim: lab-grown and synthetic food becomes normal, real food becomes a luxury.

What supports it. Precision fermentation and cultivated meat have regulatory approval in several markets. Plant-based protein is mainstream. Vertical farming runs commercially.

What argues against it. Cost curves, mostly. Cultivated meat remains far more expensive per kilo than conventional, and scaling bioreactors is a hard engineering problem, not a software one.

The "only the rich eat real food" inversion also gets the economics backwards. Processed food is cheap; fresh food is expensive. That is already true today and synthetic production pushes further in the same direction.

Verdict: right that synthetic food grows, inverted on who eats what. If anything becomes the luxury good, it is the real tomato — as it already is.

Transport, restaurants and farming go fully automated

The claim: no drivers, no waiters, no farmhands.

What supports it. Driverless taxis operate commercially in multiple cities. Autonomous tractors, robotic milking and drone crop-spraying are all in production agriculture now. Kitchen automation is real for high-volume, low-variation food.

What argues against it. Each of these is winning in narrow, structured conditions — mapped city centres, flat fields, one dish repeated a thousand times.

Difficulty rises sharply outside them. Unmapped roads, uneven terrain, an unfamiliar menu, a customer changing their mind.

SHRM's numbers show it plainly: food preparation and serving sits at 10.8% highly automated. Computer and mathematical occupations sit at 51.2%.

Verdict: right about the direction, wrong about the order. Desk work is being automated faster than physical work. The office is falling before the kitchen.

Gold and currency will be worth nothing

The claim: in 20 years, gold and currency values go to zero.

What supports it. Genuine monetary change is underway — digital currencies, central bank digital currencies, and AI-driven trading now moving a large share of volume.

What argues against it. Almost everything else.

Gold has held value across every technological revolution in recorded history — printing, steam, electricity, computing, the internet. It is valuable because it is scarce and hard to produce, and AI changes neither property.

Currency has value because a state enforces it and people accept it. Both conditions are political, not technological.

Verdict: this one is not supported. Expect volatility, new instruments and fierce argument about what money is. Expect zero only if you also expect states to stop existing.

Not financial advice. Nothing here is a recommendation to buy or sell anything.

Privacy will go to zero

The claim: technology erases privacy entirely.

What supports it. Face recognition, gait recognition, data brokerage, and AI that reassembles identity from fragments that used to be anonymous. Re-identification is now cheap.

This is the prediction with the clearest evidence behind it.

What argues against it. Regulation is moving in the opposite direction — GDPR, the EU AI Act, CCPA and their successors. On-device processing and end-to-end encryption are becoming default rather than niche.

Verdict: largely correct, unevenly distributed. Default privacy is collapsing. Deliberate privacy is becoming a skill and, increasingly, a product people pay for.

Human talent will not be needed

The claim: no cause left for human ability.

What supports it. Skills that were valuable because they were rare — basic copywriting, stock illustration, simple code, straightforward translation — have collapsed in price. That is real and it happened fast.

What argues against it. Value moves, it does not evaporate.

The same research showing heavy automation exposure also shows a 56% wage premium for roles requiring AI skills. Photography did not end painting. It ended portrait painting and created an industry.

Verdict: wrong as stated, right underneath. Talent is not becoming unnecessary. It is being repriced — and "I can do this thing" is worth much less than "I know which thing is worth doing".

What this means for your business, practically

Forecasts are interesting. Decisions are useful. Here is the practical read.

• Automate the routine, keep the judgement. The 20% that is already half-automated is admin, data entry and routine drafting. Move it now.
• Own something that cannot be generated. A customer relationship, proprietary data, a physical operation, a real reputation. Generated output is no longer a moat.
• Make your website readable by machines. Answer engines now sit between you and your customer. If a model cannot parse your page, you are invisible to the layer that increasingly decides.
• Do not fire people to buy tools. The research showing displacement also shows firms that automate with staff outperform those that automate instead of them.
• Verify before you trust. The medical device numbers are the clearest warning in this article — certification is not the same thing as evidence.

Want to know whether AI answers can read and quote your own business? That is exactly what our SEO & Rank in AI work checks — and if you would rather put AI to work inside your business, start with AI agents.

Frequently asked questions