Latest / Taylor Tailored / How AI Will Reshape Society Faster Than Most People Realise
Transcript
- 0:00How AI will reshape society faster than most people realize.
- 0:05AI will not replace society, it will reorganize it.
- 0:09Artificial intelligence is no longer a niche technology story.
- 0:14It is becoming a force that will alter how people work, learn,
- 0:17govern, trust, create, compete and live with the biggest
- 0:21changes likely to come not from robots replacing humanity, but
- 0:24from institutions quietly reorganising around machine
- 0:28intelligence. The biggest AI story is not the
- 0:31technology, it is the social rewiring.
- 0:34Most people still talk about AI as though it is a product story.
- 0:38Better chat bots, smarter search, faster coding, weird
- 0:43image generators. A helpful assistant here, an
- 0:46automation there. That framing is too small.
- 0:50The deepest story is that AI is starting to become
- 0:53infrastructure. It is moving from novelty to
- 0:56system layer, embedded in offices, schools, hospitals,
- 1:00call centres, software, logistics, media and public
- 1:03administration. Stanford's 2025 AI Index
- 1:07captured the shift clearly. Business use is accelerating.
- 1:11Private investment remains enormous and generative.
- 1:14AI is no longer sitting at the edge of the economy.
- 1:17It is moving into the middle of it.
- 1:19That matters because infrastructure changes societies
- 1:22differently than gadgets do. A gadget adds convenience.
- 1:26Infrastructure reorganizes behaviour.
- 1:29When electricity spread, it changed factories, cities,
- 1:33leisure and domestic life. When the Internet spread, it
- 1:37changed commerce, news, politics and attention.
- 1:41AI looks increasingly like the next layer in that sequence, not
- 1:45just a new tool but a new operating logic.
- 1:48The defining question, then, is not whether AI will be
- 1:51impressive. It already is.
- 1:54The defining question is what happens when societies begin to
- 1:57depend on systems that can generate language, classify
- 2:00risk, predict behaviour, recommend action and automate
- 2:04parts of thought itself. That is where the real upheaval
- 2:08begins. Work will change first, but not
- 2:11in the simple way people imagine.
- 2:13The laziest version of the AI debate is the one that asks
- 2:17whether machines will replace workers.
- 2:19Some jobs will be heavily automated, some tasks will
- 2:22disappear. But the more immediate change is
- 2:25subtler and in many ways more disruptive.
- 2:28Work will be broken apart, measured differently, and
- 2:32redesigned around human machine collaboration.
- 2:35The evidence already points in that direction.
- 2:37The IMF has argued that AI exposure does not produce one
- 2:42single labour market outcome. In some occupations, AI can
- 2:46substitute for human effort. In others it complements workers
- 2:50and boosts productivity. The distinction matters.
- 2:54It means the future is unlikely to be a simple mass extinction
- 2:57of jobs. It is more likely to be a
- 2:59sorting process in which some roles are enhanced, some
- 3:03degraded, and some quietly hollowed out.
- 3:06The Ilo's updated work on generative AI makes the pattern
- 3:10even more concrete. Clerical work remains among the
- 3:13most exposed categories, while more digitized professional and
- 3:17technical work is also becoming more vulnerable as models
- 3:21improve. That should kill off the
- 3:23comforting fantasy that only routine, low status work is at
- 3:27risk. AI does not just come for
- 3:29repetitive factory motions, it also comes for paperwork,
- 3:33drafting, summarising, support, formatting, analysis and
- 3:38coordination. The invisible white collar
- 3:40tissue that holds modern organisations together.
- 3:43This is why the first big social effect of AI may not be mass
- 3:47unemployment, but class instability within white collar
- 3:50life. Junior analysts, paralegals,
- 3:53coordinators, copywriters, customer support staff, and
- 3:57administrative teams may find that the ladder they expected to
- 4:01climb is thinner than it used to be.
- 4:03If AI does the entry level work, where do people gain experience?
- 4:07If firms can get more output from fewer people, what happens
- 4:11to progression? If one strong worker using AI
- 4:14can do the work of several average workers, what happens to
- 4:18the middle? Those are not abstract
- 4:20questions. They go to the structure of
- 4:22modern careers. There will also be gains.
- 4:25New skills are already appearing in job ads and the IMFS 2026
- 4:30work suggests AI related capabilities are increasing wage
- 4:34and employment prospects in advanced economies.
- 4:37But that is exactly the point. Gains will not be distributed
- 4:40evenly. Workers with digital fluency,
- 4:43domain expertise and adaptability are better
- 4:47positioned to benefit. Those without them risk being
- 4:50pushed into weaker bargaining positions.
- 4:52So AI will not simply eliminate work, it will reorder the status
- 4:57inside the workforce. Some people will become
- 5:00dramatically more productive and valuable.
- 5:02Some will become easier to monitor, benchmark and replace.
- 5:07Some will discover that the hardest thing to automate is not
- 5:10intelligence, but trusted judgement.
- 5:12In messy, high stakes human settings, that last category may
- 5:17become more valuable than many currently realise.
- 5:21Education is about to face a brutal test of purpose.
- 5:24Schools and universities are among the institutions least
- 5:28prepared for what AI means. For years, education systems
- 5:32treated digital technology as an add on.
- 5:35AI makes that impossible. If students can generate essays,
- 5:39summaries, study guides, code images and seemingly competent
- 5:44explanations in seconds, then a huge amount of conventional
- 5:47assessment becomes unstable. UNESCO has repeatedly pushed a
- 5:51human centred approach for exactly this reason, warning
- 5:54that policy and pedagogy are lagging behind the technology.
- 5:58The challenge is bigger than cheating.
- 6:00Cheating is only the shallow end of the problem.
- 6:03The deeper issue is cognitive outsourcing.
- 6:06If young people grow up using AI to draft, condense, brainstorm,
- 6:10explain an answer before they have built confidence in their
- 6:13own thinking, education may drift toward a dangerous
- 6:17illusion, polished output without durable understanding.
- 6:20A student can submit better looking work while learning
- 6:23less. That is a serious social risk
- 6:26because societies do not just need credentials, they need
- 6:30adults who can reason under pressure, detect nonsense, and
- 6:34form judgements without instant machine assistance.
- 6:37At the same time, it would be foolish to pretend AI offers no
- 6:41educational upside. Used well, it can personalize
- 6:44explanations, provide feedback at scale, widen access to
- 6:49tutoring, support teachers with planning, and help learners move
- 6:52faster through material. OECD and UNESCO works both point
- 6:56toward the same conclusion. AI in education should not be
- 7:00treated as a ban or embrace question.
- 7:03The real challenge is how to redesign learning so that AI
- 7:06supports understanding rather than replacing it.
- 7:09That likely means schools will have to shift away from over
- 7:12reliance on take home output and toward more in person reasoning,
- 7:17oral defence, iterative drafts and demonstrated process.
- 7:21In other words, education may have to become more visibly
- 7:24human just as AI becomes more capable.
- 7:27That would be an irony worth noting.
- 7:30The age of artificial intelligence may force
- 7:32institutions to rediscover the value of authentic human
- 7:36thought. Healthcare could improve
- 7:38sharply, but only if trust survives.
- 7:41Health is one of the clearest examples of a is double edge.
- 7:45The potential upside is obvious. Better pattern recognition,
- 7:49faster triage, smarter decision support, improved administrative
- 7:54efficiency, more accessible information, and new ways of
- 7:58handling complex data across imaging records, language and
- 8:02diagnostics. Whose recent guidance on large
- 8:05multimodal models in health reflects how real this shift has
- 8:09become. AI is no longer a futuristic
- 8:12talking point in medicine. It is entering operational
- 8:15reality. But Healthcare is also one of
- 8:18the worst places to be naive. A wrong answer in a chat bot is
- 8:23embarrassing. A wrong answer in healthcare can
- 8:26be dangerous. Bias, hallucination, weak data
- 8:29governance, poor oversight, overconfident users and opaque
- 8:33model behaviour are not minor bugs in this context.
- 8:37They are systemic risks whose health governance work keeps
- 8:41returning to ethics, accountability and equity.
- 8:44Because the promise of AI in medicine can quickly collapse if
- 8:48trust breaks, this is where a broader social lesson appears.
- 8:53AI will work best in domains where it augments professionals
- 8:56rather than displacing them. Doctors, nurses, pharmacists,
- 9:01and clinicians do more than process information.
- 9:04They interpret, contextualize, reassure, escalate, and take
- 9:09responsibility. Those are social functions as
- 9:12much as technical ones. So the future of AI in health is
- 9:16unlikely to be machine medicine replacing human care.
- 9:19It is more likely to be a constant negotiation over where
- 9:23automation stops and human accountability begins.
- 9:26That negotiation will not stay inside hospitals.
- 9:30It will spread across law, finance, education, social care,
- 9:34insurance and public services too.
- 9:37Politics, truth and trust may become the hardest problem of
- 9:41all. The most unsettling long term
- 9:43effect of AI may be what it does to shared reality.
- 9:46Modern societies run on more than roads, contracts and
- 9:49markets. They run on trust in evidence.
- 9:53Trust that a voice recording is real.
- 9:55Trust that a video is not fabricated.
- 9:57Trust that a scale is hard to fake.
- 10:00Trust that information friction prevents total pollution of the
- 10:04public. Square AI weakens all of that.
- 10:07The World Economic Forum has repeatedly ranked misinformation
- 10:11and disinformation among the top short term global risks, and
- 10:15it's more recent work shows why. Even when deep fakes do not
- 10:19single handedly decide elections, they still intensify
- 10:22confusion, lower confidence, and make truth easier to contest.
- 10:27Once citizens know convincing synthetic content exists, every
- 10:31real clip becomes deniable and every fake clip becomes usable.
- 10:35That is a deeper civilizational problem than a single viral
- 10:39hoax. It creates what might be called
- 10:41ambient epistemic instability, a background condition in which
- 10:46seeing is no longer believing, hearing is no longer proof, and
- 10:49institutional trust erodes further because people can no
- 10:52longer agree on what is authentic.
- 10:55This matters for democracy, It matters for courts, it matters
- 10:59for financial markets, it matters for journalism, it
- 11:02matters for ordinary relationships.
- 11:05A society saturated with synthetic media does not merely
- 11:09become more deceptive, it becomes more suspicious.
- 11:12And suspicion, once normalized, is corrosive.
- 11:16What media misses The most important thing many people
- 11:19still miss about AI is that it's biggest danger is not a sudden
- 11:24robot takeover. It is a slower institutional
- 11:27drift in which more and more decisions become shaped by
- 11:30systems that are cheap, scalable, and difficult to
- 11:33challenge. That drift can look efficient A
- 11:35hiring screen here, a fraud score there, A risk flag.
- 11:40A predictive model. A triage tool, A generated
- 11:44performance summary, A recommended sentence, A school
- 11:48intervention alert, a customer vulnerability rating, a policing
- 11:53prioritisation system. None of these has to be
- 11:56apocalyptic on its own. The danger is cumulative.
- 11:59Once AI enters decision chains, humans often start adapting to
- 12:04the system rather than the other way around.
- 12:06Workers optimize for what the model tracks.
- 12:09Managers defer to outputs they do not fully understand.
- 12:13Citizens are judged by data patterns they cannot inspect.
- 12:17Institutions become more legible to machines and less legible to
- 12:21the people inside them. That is the real societal turn.
- 12:25AI does not need to become conscious to become powerful.
- 12:28It only needs to become administratively normal.
- 12:31The AI economy could deepen inequality if power concentrates
- 12:35further. Every major technological wave
- 12:38creates winners, but AI may produce unusually concentrated
- 12:42ones. Training frontier models
- 12:44requires capital, compute data, infrastructure, talent and
- 12:48energy on a scale few actors control.
- 12:51Stanford's 2025 data on private investment shows how large the
- 12:55gap already is, with the United States massively ahead of rivals
- 12:59in private AI spending. That is not just a business
- 13:02statistic, it is a power map. If a small number of firms own
- 13:06the best models, the deepest compute stacks, the most
- 13:09valuable distribution channels, and the enterprise relationships
- 13:13that turn AI into daily workflow infrastructure, they do not just
- 13:17sell software. They help shape the future
- 13:19bargaining power of labour, the cost base of business and the
- 13:23information architecture of public life.
- 13:25The World Bank has warned that without proactive policy, AI
- 13:29could widen divides between large and small firms, high and
- 13:33low skilled workers, and advanced and emerging economies.
- 13:37That matters because societies do not experience technological
- 13:41progress evenly. Some places have the capital and
- 13:44institutions to absorb it. Others become dependent
- 13:47consumers of systems designed elsewhere.
- 13:50This is where the phrase AI divide starts to matter.
- 13:53The old digital divide was partly about access who had
- 13:57Internet, devices and connectivity.
- 14:00The AI divide is also about capability and leverage.
- 14:03Who can build models, fine tune them, govern them, and capture
- 14:07the gains? That is a more strategic
- 14:10inequality. A country, company, or class
- 14:13that uses AI is not in the same position as one that shapes AI.
- 14:18Those who set standards, own infrastructure, and train talent
- 14:22will enjoy compounding advantages.
- 14:25Those who do not may find themselves permanently
- 14:28downstream. Energy, resources and the
- 14:31physical world still matter. AI is often discussed as though
- 14:35it were weightless code floating in the cloud.
- 14:38It is not. It runs on chips, data centres,
- 14:41cooling systems, electricity grids and supply chains.
- 14:45The International Energy Agency's 2025 analysis makes the
- 14:49point starkly. Global data centre electricity
- 14:52consumption is projected to rise sharply this decade, with the
- 14:56base case seeing it roughly double by 20-30 to around 945.
- 15:01Two WA. That has two big implications.
- 15:04First, AI is not just a software story, it is an industrial and
- 15:09energy story. The future of AI will be shaped
- 15:12partly by who can build power hungry infrastructure and feed
- 15:16it reliably. That ties AI to electricity
- 15:19markets, grid investment, energy security, and industrial policy.
- 15:24The next phase of the AI race may be as much about substations
- 15:28and generation as it is about model releases.
- 15:31Second, the social cost of AI cannot be discussed only in
- 15:35terms of convenience and productivity.
- 15:37It must also include the physical burden of scale.
- 15:40Any serious conversation about how AI reshape society has to
- 15:45include the question of what kind of infrastructure society
- 15:47is willing to build to support it, and who pays for that build
- 15:51out. That is another reason the
- 15:53future of AI will be political, not merely technical.
- 15:57Governments are moving, but governance is still behind the
- 16:01curve. Regulation is no longer
- 16:03hypothetical. The EU AI Act is moving through
- 16:07phased implementation, with prohibited practices and
- 16:10obligations around general purpose AI becoming real
- 16:13compliance territory rather than abstract policy talk.
- 16:16The law matters not because it solves every problem, but
- 16:19because it signals a new era. Governments are no longer
- 16:22treating AI as a field that can be left entirely to voluntary
- 16:26self restraint. At the same time, governance
- 16:29remains fragmented. NIST generative AI risk profile
- 16:33is influential, but voluntary. UN work is pushing for broader
- 16:37international coordination. But global governance moves
- 16:40slowly, especially in a field advancing this quickly.
- 16:43This creates a familiar modern gap.
- 16:46Technology scales globally, while oversight remains partial,
- 16:49national and delayed. That gap matters because AI
- 16:54risks are not all the same. Some concern safety, some
- 16:58concern labour, some concern competition, some concern
- 17:02children, Some concern surveillance, some concern
- 17:06speech, culture, or sovereignty. There will not be 1 master rule
- 17:10that handles all of this neatly. The future will likely involve a
- 17:14growing web of sector specific controls, procurement standards,
- 17:18liability questions, audit rules, and disclosure
- 17:22expectations. In plain English, society is
- 17:26heading toward a world in which using AI will no longer be the
- 17:30interesting question. The interesting question will be
- 17:33under what rules, with what accountability, and in whose
- 17:36interests. Human value will not disappear,
- 17:39but it will be redefined. Whenever a powerful technology
- 17:43appears, people jump to extremes.
- 17:45Either humans become obsolete or nothing really changes.
- 17:50Both instincts are wrong. AI will make some human
- 17:53strengths more valuable, not less.
- 17:56Judgement under uncertainty, Trustworthiness, taste,
- 18:00leadership, social intelligence, ethical reasoning, context
- 18:05sensitivity, responsibility, calm under pressure, the ability
- 18:10to decide when the model is wrong.
- 18:12These are not decorative traits. In a society filled with machine
- 18:16generated output, they may become premium traits.
- 18:20But there is a catch. Human value will not
- 18:24automatically be protected. It will have to be organized,
- 18:27defended and designed into institutions.
- 18:30If businesses use AI primarily to cut headcount, squeeze labour
- 18:34and accelerate output without reinvesting in people, the
- 18:37result could be a harsher, thinner social order.
- 18:40More efficient, less secure, more productive, less trusted.
- 18:44If instead, AI is used to remove drudgery while raising the value
- 18:49of genuinely human contribution, the result could be different.
- 18:53That is why this is ultimately a political and moral question as
- 18:57much as an economic 1 Technology opens possibilities.
- 19:01Societies decide which possibilities become normal.
- 19:05What happens next? The most likely next phase is
- 19:09not a dramatic single event. It is continued diffusion.
- 19:13More companies will embed AI into standard workflows.
- 19:16More workers will be expected to use it.
- 19:19More students will rely on it. More public bodies will
- 19:23experiment with it. More citizens will encounter it
- 19:26without always knowing they are doing so.
- 19:28The most dangerous next phase is silent dependency.
- 19:32Institutions may adopt AI faster than they build safeguards using
- 19:36systems they do not fully audit because the performance gains
- 19:39are too tempting. That is how trust problems
- 19:42become systemic, not through 1 spectacular failure, but through
- 19:46normalization before governance catches up.
- 19:48The most underestimated next phase is social sorting.
- 19:52People, companies, and countries that learn to work with AI
- 19:56effectively will pull ahead. Those that do not may not merely
- 20:00fall behind. They may become structurally
- 20:03dependent on other systems, standards and terms.
- 20:06That could reshape class, regional inequality, and
- 20:10geopolitical power more than many current headlines suggest.
- 20:14The real future of AI is a fight over social design.
- 20:18The future of AI will not be decided by benchmark charts
- 20:22alone. It will be decided by how
- 20:24workplaces are redesigned, how schools respond, how governments
- 20:28regulate, how media rebuilds trust, how health systems set
- 20:32limits, and how societies answer a simple but enormous question.
- 20:37When machines can simulate more and more of intelligence, what
- 20:41do we still insist must remain deeply human?
- 20:44That is the real pressure point. AI will almost certainly make
- 20:48parts of life faster, cheaper, and more scalable.
- 20:51It may also make societies more unequal, more monitored, more
- 20:55suspicious, and more dependent unless those tendencies are
- 20:59actively resisted. The point is not to be blindly
- 21:02pro AI or reflexively anti AI. The point is to understand that
- 21:06the stakes are now much larger than product features.
- 21:10This is no longer just a story about smarter tools, it is a
- 21:13story about what kind of society Smart Tools will help build.