top of page
  • Writer's pictureJakob Nielsen

The 3 Ages of Authorship and Creation

Summary: Oral tradition, writing and printing, and newly developing human-AI co-created media differ along multiple dimensions. Each type of authorship creates very different works and experiences for both authors and readers. The change to AI-supported creation may or may not be as big as the change from recitals to books, but it will have a dramatic impact.


We are about to experience a third age of creation and authorship, driven by AI. This new age differs from the previous age of the written tradition along all parameters, even if the differences may not be as big as that between oral tradition and writing.

The three ages are:

  • Oral tradition: Before the advent of writing, information was purely transmitted in oral performance. People would talk to each other, and whatever information they could remember was the only data available to them. The oral tradition probably started with the first modern humans, and was the only form of authorship until writing became popular around 750 BC in Europe and earlier in China.

  • Writing: Information would be fixed in a written format, which would then be replicated. At first, handwritten copies were produced one at a time, and later by mass production in the form of printing. These few or many copies would then be distributed and could be retained over time. Initially, only text was replicated, but the introduction of audio and visual recording technology expanded the number of media types that were distributed.

  • AI-Human Synergy: Information is produced in collaboration between a computer and a human, which vastly expands the range of possible creations and the creative prowess of individuals.

The following table summarizes the differences between these three forms of authorship:

Media Characteristics

In the oral tradition, there was no writing and no fixed representation of any authored work. The work only existed while it was being performed, and the author would recreate it from memory each time. Since nobody can remember every single word of a long piece, every performance would be different.

Ethnographic studies of performers in modern times have found that even famous epic poems gradually drift over the years and differ more and more from the initial performance. Also, the only people who can experience any given version of the work are those who assemble within hearing range of the author when he or she performs the work. The need for the labor of a performance for each small audience made the work very expensive per recipient.

Because the work only lives in the author’s brain it is impossible to engage in true editing, the way we can do in a word processor, or even on a piece of paper with handwriting.  

In oral tradition, a different variation of the work is created every time it’s performed. The reach of that version is limited to the number of people who assemble within hearing range of the creator-performer. (Leonardo)

In contrast, writing fixes the work in a physical form (later digital), which can be saved and read many years later. It can also be moved through the world and accessed far away from the author’s location. In the beginning, copies were hand-made and thus very expensive, but after the invention of printing, copies became cheaper. However, the initial version became more expensive than in the oral tradition, because the expectation arose that authors would polish their work through extensive editing before committing it to publication. This made sense, since the published version was fixed and could not be changed, except by publishing a new edition which would require even more editing.

Handwritten copies of manuscripts were expensive for more than a thousand years of the age of written media. (Midjourney)

Even though I use the term “written tradition” for what came after the oral tradition, the works don’t have to be comprised of words. Many more media forms are now mass produced, including music, film, and video. Also, in recent years, reproduction has been through digital media rather than physical media.

Now that we are turning to co-creating media with AI, the cost of authorship can drop to almost zero: it becomes cheaper, the more responsibility we leave to the AI.

The human contribution is still fixed when the author decides to publish the work, but the AI contribution can in principle be recreated every time the work is accessed, even though this is usually not done yet. My image to represent the making of handwritten copies could be generated afresh for each individual user, such that, for example, readers in Asia would see a Buddhist or Hindu monk instead of a Christian monk. Female readers might see a nun. In general, there would be no end to the potential personalization possible for each image or other AI contribution to the work. We could recapture an important quality of the oral tradition, where the author would likely have adapted the work to the specific audience when performing it.

Having the AI vary its contribution is not limited to images. The words could be translated into the reader’s preferred language, and the text rewritten at a reading level suited for that reader. (This article is written at a 12th-grade reading level, which is the level of difficulty I usually target. The text of this article is human-written by yours truly, but if it were AI-generated, you could be served the article at your preferred reading level: maybe you appreciate more highbrow intellectualization, or maybe you’d rather have a bit easier reading.)

Business Model and Social Impact

In the oral tradition, the work itself could not be sold, since it did not exist independently from the author. One could sell tickets to a performance of the work, and in many cultures, the nobility or the community would fund skalds to keep them around for entertainment and glorification purposes.

Today, almost no authors from the oral tradition are known by name. The exceptions are authors who were lucky enough to live around the transition from oral to written tradition so that they would have the work transcribed by a literate friend or somebody who could still remember their work a few generations later.

For example, the Viking Egil Skallagrimsson who lived from around 904 to 995 was a famous bloodthirsty warrior, but also a renowned poet, as were many Vikings. Living in a mostly illiterate society, they became adept at composing skaldic poetry on the fly, as did Egil when he was facing execution at the hands of King Erik Bloodaxe. (If you’re captured by a guy nicknamed Bloodaxe, you should not expect things to go well for you.) Egil composed a poem in praise of Erik that so impressed the King that he let Egil go. This poem became so famous that it was recited through several generations until it was finally written down around 1220. But it would have been very rare for a work to survive for almost 300 years in an oral tradition.

Thus, in the oral tradition, the business model was tickets or patronage. (Or saving your life by impressing the King.) The author had to be personally present to benefit from his or her work. This again made creation totally decentralized, at the hands of each individual creator, with no support system.

Since patronage and local community pride were important for each locally-based author, much of the work in the oral tradition was in the form of epic poetry glorifying heroes and warriors capable of feats of bravery in individual combat.

In the written tradition, the work is replicated in a physical form that can be sold. Usually, the name of the author is prominently featured on the work, especially in the case of famous creators, such as pop stars. This means that thousands of creators have gained fame, and millions are known, if only to specialists.

Printing allowed for mass production, leading to mass sales and high income for best-selling authors, but also incentivized a centralized system that preferred to cater to mass audiences instead of niches. (Midjourney)

Because you make more money, the more copies you sell, the written tradition rewards scale. If you can write a bestseller or a hit rock song, more credit(s) to you. And the more support you receive from a vast publishing and promotion machine. This incentivizes somewhat bland works with mass appeal. The system tends to be centralized, with a few big publishers publishing bit hits. A few major newspapers used to have print runs in the millions, even though their business model has been undermined by the web, which famously turned print dollars into digital pennies.

Writing also encourages centralization in other aspects of society. Record-keeping and the transmission of written orders facilitate the growth of political entities far beyond the city-state, where all citizens would meet in the agora to discuss the matters of the day. Kings and presidents could collect taxes and order people around across long distances, and executives could manage ever-bigger corporations.

The new AI-human symbiants allow people to create work even if it’s not a best-seller. We have already seen this trend in pre-AI digital media. For example, YouTube and TikTok support highly niche video creators who can shoot and produce videos with no other tools than a smartphone. While the production quality of these videos isn’t up to the product of the major Hollywood or Bollywood studios, niche videos published on the Internet still collectively rack up more viewing hours than any chart-topping movie.

We can estimate that the total viewing time across YouTube and TikTok is slightly less than one trillion hours of video watching per year. (This doesn’t even account for other platforms publishing user-contributed video, such as Tencent or Bilibili in China.)

In comparison, the highest-grossing movie of 2023 was Barbie, which sold about 140 M tickets worldwide. Since this film ran for almost two hours, it accounted for 280 M viewing hours, or only slightly more than a quarter billion. User-created video was 3,500 times bigger than Barbie, even when only counting the two biggest platforms.

We’re definitely moving to a decentralized publication model, and AI will accelerate this trend, because authoring becomes even cheaper and even more niche, though appealing even more strongly to specific audiences. As discussed in the next section, media forms like video and songwriting that traditionally were expensive can now be had at a simple prompt to specialized AI services, which again means that individual authors and small companies can add these previously-elite media forms to their repertoire.

Even though I can only speculate at this early stage, even more decentralized authoring might support other forms of decentralization as smaller companies gain the same abilities as huge conglomerates for reaching customers with high-quality highly-target media.

Media Mastery

In the oral tradition, it was insufficient for an author to master the media form itself. He or she also needed to be a skilled performer, with stage charisma, a strong voice that could carry, and a prodigious memory to remember a long poem or other work well enough for a smooth presentation.

Since much oral performance happens by singing, creators in the oral tradition would preferably master both words and music.

In contrast, the written tradition rewards specialization. It’s possible to be rich and famous by only mastering a single media form, but mastering it very well. Write best-selling books or compose the music for a series of hits, and you have it made. The music business has many examples of pairs of specialists who created major hits by having one person write the lyrics and another compose the music. Of course, there are also many examples of songwriters who did both, and often added the third dimension of also performing the song.

Mostly, though, the written tradition is one of media specialization. Not just for creators, but also for publishers.

In contrast, with AI, each individual creator can integrate as many different media as he or she wishes. Taking my personal example, I am now including extensive illustrations in my articles. All I need to do is to imagine a visual, and I can have AI make it for me. We’re on the cusp of gaining similar mastery-by-everybody of music, songs, animations, video, and interactive computer games. I made a small song about my 10 usability heuristics. Nothing good yet, but next year, I can well imagine becoming a more full-fledged multimedia creator.

The Brain vs. the World

In the oral tradition, all information is in the brain, and nothing is offloaded to the world. The text of a poem is in the author’s head, as was the structuring of the plotline that came before the detailed composition of the work.

In the written tradition, the content moves out of the brain and into the world: maybe onto paper, maybe into a word processor. All the thinking about the work, from initial ideation, over outlining and structuring, to the picking of individual words, still happens in the author’s brain.

With AI-human co-creation, potentially everything moves into the world. The processing can be by computer, which ideates plot options for the human. After the human picks the general direction, the AI can generate multiple alternate outlines, again for the human’s choice. And all of the words can be generated by the AI, only to be edited or refined by the human.

Different authors have different workflows. I currently still create most of the direction and structure of my works in my brain and only use AI for individual pinpoints of ideation.

The balance between the brain and the world may create much tension as we move to combined AI-human authorship.

The more of the process we allow to reside externally to the human brain, the more we empower larger segments of the population. Most people simply don’t have sufficiently powerful brains that they can retain and process copious amounts of information, for example to create complex plotlines purely by memory. Taking notes is already a way of offloading some cognitive burden to the world, but having the AI perform more processing and manipulation of rough ideas and preliminary structures will take this further and allow people with less elite brainpower to become authors.

Pride in Authorship

Worrying about pride of authorship may seem rather egotistical and elitist. However, I’m predicting an explosion in authorship with billions of people co-creating multimedia works with AI support. So it becomes an issue of broad interest to what degree creators will take pride in their works.

In the two old traditions, there was no doubt. The oral creators were a small elite held in high regard in the community. They did something nobody else could do. And since their creations were intensely personal and only existed while the creator performed the work, it’s extremely likely that creators took high pride in their skills and their creations.

Similarly, in the written tradition, the best creators would be highly rewarded with both money and awards, which would tend to give them a (maybe sometimes inflated) sense of worth. And most creators would take pride in their work and only release it once it had been polished.

Who holds the (metaphorical) pen during human-AI co-authorship? I still tend to prefer a prevalence of human initiative, but the best balance for different scenarios remains to be determined. (Midjourney)

I am less certain in the degree of pride when it comes to human-AI co-created works. I have not studied this formally, but speaking for myself, I know that the illustrations I create through AI are not as good as the work of the top human artists. (But it’s unrealistic to expect top artists to illustrate a newsletter with a circulation of 10,000.) I still take pride in my work and think that my articles are worth publishing. I also believe that my mixed word-image format is better than either media alone.

Quite likely, the degree of pride in authorship for co-created work will depend on the degree of human involvement and mastery. If somebody just asks AI to make a multimedia production with minimal prompting and guidance on the part of the human, that human may not feel much pride in the result.

On the other hand, somebody who follows my workflow of intense human control will likely take much pride in the work, even if many details are contributed by AI.

As an analogy, I don’t know if Shakespeare was proud of his penmanship, or only took pride in the actual plays, no matter how pretty or ugly the words looked in the manuscript. But I doubt that modern writers have less pride in their manuscripts, even if they are hammered out on a typewriter or entered into a word processor. Calligraphy is an art form, as is typography. And I know from personal experience that I did take pride in those of my books that had exceptional book design, showcasing typography far beyond my meager skills. It was still my book, even when the book designer (deservedly) won an award for the book design of that publication.


Top Past Articles
bottom of page