AI and the New Renaissance

I have a theory: AI will make a new Renaissance possible.

Not the naive version.

Not "one person now knows everything." Not "models replace expertise." Not the fantasy of the lonely genius casually outperforming every specialist.

Something more interesting:

AI lowers the cost of entering a new field.

That sounds small. I think it is enormous.

The Renaissance polymath was a bridge

Leonardo da Vinci is the obvious caricature of the Renaissance polymath, but the caricature exists for a reason.

The Metropolitan Museum describes him as a painter, sculptor, architect, engineer, scientist and inventor, and notes that his notebooks cover anatomy, animal and plant life, water, birds in flight, machines and engineering problems [Met Museum]. The National Gallery of Ireland, in an exhibition of Leonardo drawings from the Royal Collection, describes the same range: painting, sculpture, engineering, zoology, botany, cartography and anatomy [National Gallery of Ireland / Royal Collection Trust].

The important part is not that Leonardo had many hobbies.

It is that the fields connected.

Anatomy fed drawing. Geometry fed painting. Observation of water fed mechanics. Birds fed flying machines. He was not switching tabs between unrelated curiosities. He was building a model of the world from multiple angles.

We still had people like this later, although in more constrained forms. Georges Lemaitre is not a Renaissance polymath, but he is a useful reminder that important ideas sometimes come from people who do not fit one clean administrative category. He was a Belgian Catholic priest, mathematician, physicist and cosmologist. Britannica credits him with formulating the modern Big Bang theory and applying general relativity to cosmology [Britannica]. In 2018, the International Astronomical Union recommended renaming Hubble's law as the Hubble-Lemaitre law to recognize Lemaitre's role in discovering the expansion of the universe [IAU].

That kind of crossing became harder.

Not because people became less curious.

Because knowledge became deeper.

Modern science killed the Renaissance man for rational reasons

Benjamin Jones gave this problem a brutal name: the burden of knowledge.

In "The Burden of Knowledge and the Death of the Renaissance Man", he argues that as knowledge accumulates, innovators must train longer, specialize more, and increasingly work in teams [Review of Economic Studies].

The intuition is simple: the more humanity knows, the longer it takes to reach the frontier.

This is not just a poetic complaint about specialization. Wuchty, Jones and Uzzi studied 19.9 million papers and 2.1 million patents and found that knowledge production became increasingly dominated by teams; teams also produced more highly cited work than solo authors [Science / Kellogg].

So we split the work.

One person knows the instrument. Another knows the statistical method. Another knows the biological literature. Another knows the physical model. Another knows the code. Another knows the publication norms.

Specialization is not a failure.

It is a rational adaptation to a world that knows too much.

But it has a cost.

It breaks bridges.

You get excellent people in narrow corridors, and fewer people able to look at two distant fields and say:

Wait. This might be the same problem.

AI changes the cost of the detour

Most discussions about AI are stuck on productivity.

Can it write faster? Can it code faster? Can it replace this job?

I think the more interesting effect is different:

AI reduces the cost of the intellectual detour.

Before, if I wanted to seriously explore a field far away from mine, the entry cost was huge.

I had to find the right books, learn the vocabulary, identify the central concepts, separate serious debates from peripheral noise, understand the methods, know where the traps are, and produce a first artifact without embarrassing myself.

That cost still exists.

But it is lower.

A good AI system can help build a rough map of a new field. It can explain vocabulary, compare conceptual frameworks, point to primary sources, write exploratory code, generate visualizations, translate an idea from one domain into another, prepare questions for an expert, or turn an intuition into a prototype.

That is not expertise.

It is scaffolding.

And scaffolding matters.

Scientific journals are already describing this tension. Nature Reviews Physics noted in 2023 that large language models were creating both excitement and concern in scientific work [Nature Reviews Physics]. In chemistry, researchers built an LLM-based system that could plan and execute certain research tasks with tools and laboratory automation [Nature].

But the caveat matters. Models can be confidently wrong. A 2024 Nature paper found that larger and more instruction-tuned models do not automatically become more reliable on simple problems that humans can verify [Nature].

So the conclusion is not:

AI makes everyone an expert.

The conclusion is:

AI helps a curious person reach the level where they can ask better questions in more fields.

Sometimes that is the rare part.

The new Leonardo does not replace experts

I do not believe in the fantasy of the lone genius replacing modern science.

Science is too complex. Instruments, protocols, datasets, peer review, domain-specific failure modes and institutional memory all matter. In fields where reality pushes back hard, AI does not remove experts.

It makes it easier to talk to them intelligently.

The modern polymath will not be someone who knows everything.

It will be someone who can cross.

They will enter a field, learn its grammar quickly, build a provisional mental model, notice an analogy, make a prototype, then go back to sources, experts and experiments to check whether the bridge is real.

The new Leonardo is not one person replacing ten specialists.

It is one person holding ten partial maps in their head long enough to notice a connection.

That distinction matters.

It also matches another pattern in research. A Nature study of more than 65 million papers, patents and software products found that small teams tend to produce more disruptive work, while large teams tend to develop existing ideas [Nature].

Both are necessary.

But if AI lowers the cost of crossing domains, it may make individuals and small teams much more powerful at the exploratory stage.

Not to replace institutions.

To explore before them.

I see a tiny version of this in my own life

I am obviously not comparing myself to Leonardo da Vinci.

That would be absurd.

But I recognize the motion at a small scale.

Lately I find myself working across domains that used to feel like separate lives:

• apps and products
• video games
• music
• research
• AI agents for paperwork
• courses
• open-source tools

Before AI, each of these domains had a heavier entry cost. You had to choose earlier. Curiosity was expensive.

Now AI acts as a translation layer between my curiosities.

It does not give me taste.

It does not give me truth.

It does not give me validation.

But it helps me stay longer in the most valuable state:

I do not understand this yet, but I can start.

That state is underrated.

It is where a lot of real work begins.

The scarce thing is not information

If this theory is right, access to information will not be the bottleneck.

The bottleneck will be:

• curiosity
• taste
• rigor
• verification
• finishing
• not getting lost

AI gives you more surface area.

It does not give you a center.

So the new polymath will need more discipline, not less. They will need to know when to explore, when to stop, when to ask an expert, when to publish, when to be silent, and when to go back to the primary source.

But for the first time in a long time, broad curiosity feels operational again.

Not just a personality trait.

A mode of production.

If that is true, the next Leonardos may not look like court artists or tenured scientists.

They may look like strange, obsessive people surrounded by agents, simulators, notebooks, models, primary sources, failed prototypes, and human experts they keep asking:

Does this intuition survive contact with reality?

Less romantic than the Renaissance.

Much more scalable.