Generative AI is not only a new tool in the digital toolbox. Generative AI is a turning point in how societies think, create, and organise knowledge. This technology opens an extraordinary opportunity to amplify human potential, to free time and attention for higher-order thinking, and to explore creative territories that were previously out of reach for most people. At the same time, it introduces a very concrete risk of cultural and cognitive impoverishment, a drift towards intellectual laziness and copy-paste thinking.
The future is not written in the code of these systems. The future will depend on how people choose to use them: as detonators of creativity or as anaesthetics of thought; as engines for a more informed public sphere or as accelerators of noise and manipulation.
This post explores that tension along three axes: creativity, information, and the institutions that shape our collective intelligence.
1. Augmented creativity: detonator or sedative?
Generative AI has already lowered the barriers to entry in many creative domains. People who cannot draw can now produce illustrations. People who have never coded can generate working prototypes. People who struggle with writing can draft essays and emails in fluent language.
This shift does something subtle but powerful. The centre of gravity moves away from technical execution and towards vision:
- The question is less “Can I technically do this?”
- The question becomes “What is worth doing? What is worth saying? What is worth showing?”
When machines handle much of the execution, ideas, concepts, and taste start to matter more. Human value migrates upstream, towards the ability to imagine, to select, and to judge.
However, the same tools can also produce the opposite effect. When people accept the first output, when they use AI as an answer machine instead of a thinking partner, creativity starts to atrophy. A “lazy path” emerges:
- Type a vague prompt.
- Accept whatever comes out.
- Repeat.
At scale, this behaviour leads to what many designers already see: blending. Logos begin to look the same. Websites feel interchangeable. Marketing copy melts into a single, generic voice.
The danger is not only aesthetic. The deeper danger is cognitive. When people adapt their thinking to what the system is good at producing, instead of forcing the system to adapt to their thinking, they start to “think like a machine”: convergent, predictable, optimised for the average.
Generative AI can therefore either explode our creative range or compress it into a narrow band. The difference depends less on the model’s architecture and more on the discipline and ambition of its users.
2. The synthetic information crisis
The information ecosystem was already in trouble before generative AI: disinformation, clickbait, outrage-driven feeds, polarisation. Generative AI does not invent these problems. Generative AI scales them.
The cost of producing plausible but false content now approaches zero. Text, images, audio, and video can be fabricated in minutes. Fact-checking remains slow, expensive, and cognitively demanding. This asymmetry reshapes the information environment.
2.1 A market full of lemons
Economists describe a situation where low-quality products flood a market and drive out the good ones as a “market for lemons”. Something similar is happening with information.
When timelines and search results fill with low-credibility content, the average person faces a harsh trade-off:
- Either invest huge amounts of time verifying everything,
- Or withdraw, become cynical, and stop engaging in depth.
In such an environment, many people simply give up on serious information. The risk of being misled feels higher than the perceived value of being well-informed.
Generative AI, with its ability to create endless streams of convincing text and imagery, acts as a powerful engine for this “lemonisation” of the public sphere.
2.2 Bespoke realities and information incest
Algorithms already personalise what each person sees. Generative AI adds a new layer: it can assemble entire bespoke realities around an individual, a group, or a nation.
- Each person can be surrounded by content that confirms their assumptions.
- Each community can inhabit its own narrative, with little friction from opposing views.
This is not just a filter bubble. This is a custom-built epistemic environment.
A second, longer-term risk sits in the background. As more content online is generated by AI, future AI systems will inevitably train on that synthetic content. Models will then start learning from other models, not from human experience or empirical reality.
This feedback loop—sometimes called “information incest”—threatens to degrade the quality of our collective memory. Knowledge slowly drifts away from the world and folds back on its own outputs, like a photocopy of a photocopy.
2.3 Who controls the narrative?
Control of large-scale generative models means control over a powerful narrative machine.
Design choices (what data to include, what to filter, how to align responses) are not neutral. They reflect values, priorities, and sometimes political or commercial interests. When billions of queries pass through a small number of systems, even small biases can have massive effects on how people see the world.
Some countries and blocs are already moving towards their own national or regional models. An “aligned” model can be tuned to fit local political narratives, cultural norms, or strategic interests.
Society may therefore not end up with one shared internet, but with several partially incompatible informational universes, each mediated by its own AI stack.
Technology can help detect deepfakes, authenticate sources, and flag synthetic content. Yet the central response cannot be purely technical. The central response must be human: better judgement, better institutions, and better incentives for truth.
3. Redefining Human Value in a Machine Age
In this new landscape, the professional value of many skills is changing quickly.
Tasks such as basic coding, standard report writing, technical illustration, literal translation, or surface-level research are highly automatable. Generative AI is already competitive in these areas and improves rapidly.
The scarce resources shift elsewhere. They shift towards capacities that guide, question, and correct the machine:
- Critical thinking: The ability to pose sharp questions, detect inconsistencies, and evaluate sources.
- Taste: The ability to recognise what is coherent, meaningful, and original in a given context.
- Initiative: The willingness to invent new use cases, to try unconventional prompts, to combine tools in novel ways.
- Curiosity: The drive to explore unfamiliar domains and connect distant ideas.
- Patience: The discipline to iterate, refine, and let ideas mature beyond the first acceptable output.
A useful image is that of “distance authorship”. The human is no longer the primary typist or drafter. The human becomes the architect of the text, the designer of the experiment, the director of the visual concept. The machine handles much of the construction work, but the intention, structure, and final judgement remain deeply human.
If people stop at that point, if they accept default outputs and surrender their taste, this advantage dissolves. If they embrace that role with seriousness, they can gain leverage that previous generations could not imagine.
4. Information by subtraction: becoming your own gatekeeper
In a world saturated with content, the old instinct is to seek more sources, more feeds, more apps. Generative AI multiplies this instinct by making content instantaneous and infinite.
However, the most rational strategy today is often the opposite. The rational strategy is subtraction.
Information diets now need editors, not just consumers. Every person must learn to become the gatekeeper of their own attention:
- Selecting a small number of trusted outlets instead of skimming hundreds of headlines.
- Following individual writers and experts rather than anonymous streams.
- Paying attention to how something is produced, not just to how it is packaged.
- Valuing depth over volume and consistency over novelty.
This approach is slower. This approach is less entertaining. This approach is also one of the few ways to remain intellectually sovereign in an environment designed to capture and monetise attention.
Generative AI makes this discipline more important, not less. When machines can fabricate any style, humans have to protect the integrity of their filters.
5. From deterministic to generative mindsets
For decades, dominant software tools have trained users to expect deterministic behaviour. Spreadsheets, databases, and traditional programs take an input and return a single, predictable output. If the result is not what the user expects, the user assumes there is an error.
Generative AI works differently. The same prompt can produce different answers. Small changes in wording may unlock radically new directions. The “right” result is often not unique.
To use these systems well, people must adopt a generative mindset:
- People must treat the model less as a calculator and more as a collaborator.
- People must see surprising outputs not only as bugs, but as invitations to explore.
- People must learn to navigate possibility spaces, not just to retrieve single answers.
This attitude feels uncomfortable at first. The deterministic mindset promises control and certainty. The generative mindset demands curiosity and tolerance for ambiguity.
Yet the second mindset is the one that unlocks genuinely new forms of creativity and problem-solving. The first mindset turns a powerful tool into a slightly faster version of yesterday’s software.
6. Universities and research at a crossroads
Education systems sit right in the middle of this transition. Many of them are still optimised for a world in which information is scarce, exams test memorisation, and success depends on reproducing the “right” answer.
This model produces excellent human calculators, diligent rule-followers, and careful summarizers. It also produces exactly the kind of competence that machines can now imitate or outperform.
The skills that matter most in the age of generative AI—divergent thinking, problem framing, creative synthesis, ethical judgement—often remain peripheral in formal curricula.
At the same time, generative AI offers immense opportunities for education and research:
- Accelerated research: Literature reviews, data exploration, and hypothesis generation can move much faster, allowing researchers to focus more on design, interpretation, and critical debate.
- Personalised tutoring: Students can receive instant explanations, examples, and practice tailored to their level and pace.
- Relief from cognitive bureaucracy: Administrative writing, basic summarisation, and repetitive tasks can be offloaded, freeing time for genuine inquiry.
To turn these possibilities into reality, universities and schools will need to move beyond ad-hoc bans or naive enthusiasm. They will need to redesign learning around a few strategic pillars:
- Prioritising divergent and critical thinking: Assessments should reward the ability to ask good questions, to challenge assumptions, and to generate multiple plausible solutions.
- Teaching “conversation with the machine”: Students should learn prompt design, output evaluation, and the limitations and biases of AI as core literacies, not as side topics.
- Embedding interdisciplinarity: Curricula should encourage crossing boundaries between fields, so that students learn to connect technical, social, economic, and ethical dimensions of problems.
The long-term mission is simple and demanding: to educate people who can decide what machines should do, not people who wait to be told what to do by machines.
7. Delegation or elevation?
Generative AI forces a choice that is less technological than cultural.
Societies can treat these systems as tools to elevate human capacities. They can automate drudgery, expand access to knowledge, and invest the saved time in deeper thinking, richer art, and more ambitious research.
Societies can also drift into pervasive delegation. They can let algorithms choose what to read, what to watch, what to believe, and, eventually, what to do. In that scenario, creativity flattens, information fragments, and the space for independent judgement shrinks.
The unsettling element is that the second path does not require a clear decision. The second path is the default if education remains unchanged, if institutions reward speed over understanding, and if individuals treat AI as a shortcut rather than an amplifier.
Human beings still have time to steer towards the first path. The steering will not happen through slogans about “trusting” or “fearing” AI, but through concrete investments in human capital: critical thinking, curiosity, ethical courage, and the willingness to remain intellectually responsible in an age of intelligent machines.
The societies that make those investments will be the ones that keep the authority to instruct machines. The societies that do not may wake up one day to discover that they have trained themselves to follow instructions instead.