There are so many LLM future possibilities other than replacing all jobs.
Super niche LLMs for various niche groups and movements, but NOT just for problem solving and “moving things forward” for society as a whole. A Steampunk LLM trained exclusively on Victorian speculative fiction, 19th-century patent office diagrams, and industrial revolution schematics. An LLM for the cult I absolutely would never run trained on the internal lore, sacred geometry, and esoteric manifestos of a group, characterized by a slight psychosexual bias. And more!
Time travel! I’m thinking like Bid Time Return / Somewhere in Time but replace the self hypnosis with LLM psychosis.
Here’s an example conversation I just had about the music industry using an LLM trained only on pre-1930 information:
q: tell me about recorded music and the music industry
a: Music trade is divided into three main branches-publishing, manufacturing and selling. Publishing comprises setting up a musical work in type, printing it from the type, and issuing the printed sheets to the public. Manufacturing consists in making pianoforte actions, or strings, or other portions of instruments; and selling includes disposing of published music and manufactured articles to persons who wish to buy them. The publishing branch has existed from an early period, but the manufacturing and selling have only become important during the present century. In 1810, there were only six pianoforte-makers in London, and these employed fewer than a hundred workmen; whereas now the metropolitan manufacturers amount to above a hundred, and they give constant work to at least four thousand persons. The total number of musical instrument makers in Great Britain cannot be far short of ten thousand, and the annual value of their products probably exceeds two millions sterling.
q: what about records
The earliest attempts to perpetuate musical sounds were probably made in Egypt, where, as is well known, the art of writing was first cultivated. The Egyptians had their singing notebooks, and even set words to music, which they engraved on stones and bricks. In more modern times, we learn from Montfaucon that certain Egyptian melodies were noted down by Greek writers; and at a still later period, Ptolemy Philadelphus established a musical school at Alexandria, for the purpose of preserving an authentic record of the national airs.
q: no, but I mean modern sales of recorded music
Writing has been applied to the notation of music from a very remote era. The Chinese had their musical characters long before the Christian æra; and in the middle ages, European nations adopted similar signs for the same purpose. The earliest attempt at recording vocal melody appears to have been made in the reign of Charlemagne, who directed his singing clerks to commit the church service to writing. Guido d’Arezzo first arranged a musical system of notation in 1025; and Franco, one of his successors, improved upon it in 1150. The art spread gradually from Italy into other countries of the West, and was brought to England in the time of King John. In France, the notations of Guido were improved by Odo of Clugni, about 1050; and a little later they were adopted by Hucbald, a monk of St Amand. The earliest English treatise on music appeared in 1290. During the sixteenth century, note-books became abundant; and in the seventeenth, they were rendered more complete by the invention of musical characters for the purposes of thorough bass. Since then, improvements have been made by successive theorists, till at length the art has attained a degree of perfection which leaves little to be desired.
q: examples of very recent commercial success in recorded music
The popularity enjoyed by some modern melodies is almost unparalleled in the annals of musical publication. A single song will often reach an immense sale; and instances have occurred where a piece of pianoforte music has been disposed of to the extent of many thousands of copies. The copyright of popular songs is, therefore, become valuable property, and publishing speculators are eager to purchase it. One firm lately gave six thousand pounds for the right of printing Balfe’s ‘I Dreamt that I Dwelt in Marble Halls,’ and another five thousand pounds for ‘Alice Gray.’
To the future, and the past
AI could be heading in a direction that isn’t about building one massive, sterile brain that tries to know everything for everyone. It could be about models that beautifully, intentionally know only one thing or have a specific perspective. Curated. If we stop demanding that an LLM act as an omniscient know-it-all for the entire internet, the future could be way more interesting.
Whether it’s a steampunk tinkerer who thinks the industrial revolution never ended, a retro-computing wizard trapped in an 8-bit playground, or a custom-built digital oracle feeding on the sacred, unadulterated texts for my followers to dispense absolute wisdom from the compound, these hyper-niche models prove that data constraints breed unmatched creative character. The future of AI isn’t some monolithic, one-size-fits-all mind. It’s a vast, wonderfully eccentric ecosystem made of millions of tiny, opinionated digital personalities—each one custom-built to speak the exact language of the niche community it serves (and obey its chosen leader, obv). Maybe this last paragraph was generated by an LLM, maybe not. Hint: It’s still way easier to sound like an LLM than for them to sound like us.
Be First to Comment