The Picture of AI: Crafting Beauty in Large Language Models
Oscar Wilde-Inspired Reflections on the Art, Science, & Beauty in the Innovation of AI.
The Large Language Models (LLMs) of today are impressively advanced, capable of generating human-like responses and perspectives that were unimaginable just a few years ago.
I’ve been fortunate to have had a front row seat to the evolution of this new age in AI. From empathetic conversational AI’s at Maslo and Inflection, to the handful of startups I advise (chatbuds like Spuddie, Blue, & empathy experts of mpathic) along with my day job of developing advanced concepts within Mattel Future Lab. I often get the same question: Should we build and train our own LLM? My response is usually along the lines of “its expensive, probably not, perhaps use embeddings and/or fine-tuning”.
Finances aside, there’s a question hidden within the question — will LLM features & capability be a common thing for all businesses, or will they be a competitive advantage?
The Art in “State Of The Art”
Drawing inspiration from Oscar Wilde’s “The Picture of Dorian Gray,” (my favorite book) there is a scene where Wilde, told through the perspective of the artist Basil Hallward (a creative, sensitive, respected, and principled painter), shares his views on art:
“there are only two eras of any importance in the world’s history. The first is the appearance of a new medium for art, and the second is the appearance of a new personality for art.”
Two significant eras: a new medium, and a new personality.
Basil went on to give examples and compare oil-painting techniques that revolutionized the Venetians’ art scene, to the face of Antinous and its impact of late Greek sculpture.
I’m no historian, but I love the sentiment. Wilde, through the lens of Basil, made the point that art is not static. Instead, it evolves and grows alongside cultural phenomena, and can sometimes define a new-category-of-things. Painting and sculpture being the medium; the Venetians style & Antinous being the defining personality.
Much in the same way AI and large language models are evolving today, the advent of this new age of AI could be considered that new “medium” for the art of software development, communication, and even creativity. As for the “new personality”, that could arguably be reflected in how each AI or LLM is trained, the unique output it generates, and ultimately how it interfaces with its human counterpart (IMO: an area where there is much to be discovered).
As we develop increasingly sophisticated models, there’s a significant push towards fostering these ‘personalities’ within the AI. (note: I mean “ personality” both in the literal and figurative sense.) We have AI models trained to anticipate behavior, to engage in specific areas or topics, and even models that can mimic the writing style of influential figures, imbuing them with a ‘personality’ of sorts. Thus, like art, these AI models start to bear the individuality, spirit, and essence of their creators.
I want to call out: the development of a personality for an AI is a subjective, conflicting, and personal task. We all know that if a company is planning an ice cream party, there will most definitely be vanilla and chocolate options. But who decides to bring the matcha flavored ice cream? Or the lavender honey? While the ice cream example is mundane, it does take conviction and a unique POV to make that call.
Perhaps in the future, a unique element or feature of a company’s AI might be its specific style of ‘unique interaction’. Maybe AI will be able to capture the ethos of a company, embody its culture, and communicate/interact authentically to its audience. This could mean a shift in emphasis for AI designers and developers, from purely technical refinement to a blend of artistic and technical prowess.
Back to Wilde’s book — it is the perspective and value system of the painter; inspired by the sitter; and critiqued by the observer, that defines what a thing is, in our society.
“Those who find ugly meanings in beautiful things are corrupt without being charming. This is a fault. Those who find beautiful meanings in beautiful things are the cultivated. For these there is hope. They are the elect to whom beautiful things mean only Beauty. There is no such thing as a moral or an immoral book. Books are well written, or badly written. That is all.”
The value of a thing… a book, painting, sculpture, or an AI model, lies not in the materials that go into it, but in how well it is crafted to relate to an audience and the beauty it can reveal. Yes, beautifully crafted LLMs have the potential to redefine our understanding of technology — opening up new avenues of experiences, interactions, and creations — with the halo effect of us-humans obtaining a better understanding our own purpose (its woowoo, I know, but isn’t all of life a little mystical?)
To return to the original question — will the capabilities of Large Language Models become a commodity, universally available as sliced bread? Or will they instead be a competitive differentiator, where the allure lies in the artistry of its crafting? Both things will be true.
In the age of AI, language models will likely become more accessible, adopted by countless businesses as a fundamental component of operations. At the same time, the ability to fine-tune and implement these models in a beautiful, potent, and distinctive manner will remain a competitive advantage. Just as a painter might utilize common tools but wield them in a unique, inspired way, so too might a company employ AI. Ultimately, AI technology will become a commodity (where it will be a race to the bottom), but the beauty and sophistication of its utilization will not. As Wilde’s artist Basil Hallward demonstrates, it’s not simply about the paint or canvas, but rather how these tools are harnessed to evoke creativity and capture the essence of the subject.
In conclusion, the future of AI will likely see a mixture of both universal accessibility and distinctive artistry. Companies should not just ask, “Should we use AI?” but more importantly, “How can we use AI, in our own unique way?”. Now more than ever, the appreciation of beauty and the application of AI technology are intertwined. Both the medium and the personality — the tools and the artistry, that shape our world.
As Oscar Wilde said, “No great artist ever sees things as they really are. If they did, they would cease to be an artist.”
Practical Advice For Those Building AI:
- Seek to hire teams that “see beauty in beautiful things”, as their value system will have a direct impact on the production of a model.
- Invest in data sources that allow you to uncover different dimensions and modalities than your competitors. Applying LLMs to diverse, unique datasets can uncover insights that are likely not accessible to those only using common, popular data.
- Take advantage of the dualistic nature of LLMs in this moment — the convergence of capabilities, and divergence in applications. Use it to your benefit by focusing on the niches where your LLMs can truly shine.
- Always remember that technology — AI models included — are tools. They’re not intrinsically good or bad, but reflect our intentions, behavior, and biases. Use these powerful tools wisely, responsibly and with character your Grandma would appreciate.
- Understand that while the technical capabilities of LLMs might become a commodity, the applications will vary starkly. Differentiation will then lie in how these capabilities are integrated into the market, driven by unique data and user interactions.
- In a market where the technology might be replicated, it is the value system that sustains consumer trust and loyalty. Leverage this to your advantage.