This article is an update on our ongoing effort to better describe AI and LLMs to our audience.  Many people in the Arts view large language models (LLMs) and Artificial intelligence (AI) as readily available technology. Thanks to LLMs, there has been tangible transformations in creative content and marketing strategies, audience engagement, and administrative operations. Initially developed for experimentation, AI tools now serve essential roles in artistic work and administrative practices through their influence on creative development and operational workflows.

Large Language Models in the Creative Process

AI's role in artistic creation continues to expand across multiple disciplines:

AI-generated content enhances the creative process for artists by serving as a collaborative tool that generates novel ideas and enriches their work.

AI in the Arts (AI art)

AI-generated art combines human creativity and large language models, creating exciting new opportunities for artists. With the latest AI tools, artists can develop fresh ideas, try different styles, and handle the more repetitive tasks in their creative process. This collaboration expands the way art can be made and enjoyed. Artists can enhance their creativity by simplifying repetitive processes and developing new means to express themselves. This can make making art more enjoyable and bring fresh ideas to the art world.

AI images

AI image generator platforms DALL-E, Midjourney, and Stable Diffusion generate artworks based on user-provided text instructions. AI-generated images achieved popularity after winning competitions in 2022, such as "Théâtre D'Opera Spatial," yet artists primarily use AI to supplement their work instead of replacing it. AI-generated images generate novel creative potential, which artists can develop into exceptional works through their artistic skills. Human editors play an important role in shaping and guiding artistic work and are essential to the creative process. The rise of images created by artificial intelligence (AI) raises crucial questions about copyright laws and how art is made. This situation calls for new rules and a better understanding of what AI-generated art means for artists and the art world.

AI Music

Early in the evolution of LLM, AI systems such as OpenAI's MuseNet could generate multi-instrument compositions in various styles, while Google's MusicLM created music from text descriptions. The primary purpose of these digital tools is to help musicians produce musical ideas for subsequent editing and orchestrating operations. The viral "Heart on My Sleeve" track that mimicked Drake and The Weeknd using AI demonstrated the technology's capability for voice synthesis and style imitation, though it also raised legal and ethical questions.

More recently, in May 2024, country music singer Randy Travis released "Where That Came From," his first new song since a 2013 stroke left him unable to sing. This recording utilized AI technology to re-create Travis's singing voice by compositing over 40 existing vocal recordings alongside those of vocalist James Dupré. This development demonstrates AI's potential in restoring or enhancing musical performances and human art in general for artists facing physical limitations.

AI and Content

Large language models assist writers, playwrights, and content creators by generating drafts, overcoming creative blocks, and suggesting narrative directions. Human oversight ensures quality, coherence, and artistic vision, as with other creative fields.

Large Language Models in Audience Engagement

Cultural and arts institutions are leveraging AI to create more interactive and personalized audience experiences:

Chatbot Guides and Virtual Docents

Museums like the Museum of Tomorrow in Rio de Janeiro use AI chatbots to engage visitors in conversations about exhibits. The São Paulo Art Gallery implemented an IBM Watson-powered system that allows visitors to "converse" with artworks, asking questions about a painting's history or techniques.

Personalized Recommendations

Arts organizations employ AI to analyze visitor preferences and suggest relevant content, exhibits, or performances—similar to how streaming services recommend movies. This personalization enhances the visitor experience and increases engagement.

Interactive Storytelling

AI-driven interactive installations adapt content based on visitor choices or even recognize visitor emotions via sensors. This creates engaging stories that uniquely involve each person, making art more enjoyable and easier to connect with.

AI in Marketing and Outreach

AI-enhanced Customer Relationship Management systems analyze audience data to segment audiences and develop targeted marketing campaigns. Generative AI helps draft newsletters, social media posts, and press releases, which staff can edit for accuracy and tone.

AI in Customer Service and Ticketing

AI chatbots handle routine inquiries about show times, ticket prices, and directions, freeing staff to address more complex issues. Predictive analytics help forecast which shows might sell out and which might need additional marketing.

AI and Internal Operations

AI tools assist with transcribing meeting notes, drafting grant proposals, and scheduling. Museums use image recognition algorithms for provenance research and collection management, helping identify artworks or flag possible forgeries.

The Business of Art in the AI Era

The growth of AI-generated art is changing the art world in many ways, bringing both challenges and new possibilities. As more AI-generated art appears, it raises important questions about who owns the work and what it means for human creativity. The art community is actively exploring ways to value and market AI-generated pieces while ensuring that traditional artists remain at the forefront. Artists now have fascinating chances because AI-generated art shows artists fresh possibilities to turn their work into revenue and build international connections with audience members. The blend of AI technology allows artists to reach their largest possible market while finding new avenues to showcase and sell their artwork creations, which presents opportunities to enhance the commercial side of the art community through collaborative relationships between traditional and AI-based approaches.

Large Language Models in Immersive Experiences

AI enables new forms of artistic expression that transform how audiences interact with art. The distinction between handmade and AI-generated art created in immersive experiences can significantly impact the perceived value and exclusivity of the artwork.

Responsive Installations

Collectives like teamLab create digital art exhibitions where projections react to visitor movements or sounds. These installations use machine learning to allow visitors to "paint" with gestures or to evolve the exhibit over time based on audience interaction.

AI-Generated Environments

Media artists like Refik Anadol use AI to analyze massive datasets (weather patterns, astronomical data) and generate stunning visuals for immersive environments. His work includes 360-degree rooms where AI "dreams" of data are projected, creating entirely new art forms.

Interactive Performances

Experimental theater groups incorporate AI chatbots that interact with live actors or audience members, allowing the narrative to change based on these interactions. While still largely experimental, these approaches demonstrate AI's potential to make performances more immersive and personalized.

AR/VR Enhancement

AI enhances Augmented and Virtual Reality experiences by generating dynamic content. AR art applications might use AI vision to animate static murals in real-time, while VR experiences use AI to create unique dialogue for virtual characters.

Understanding Large Language Models

Large Language Models (LLMs) are artificial intelligence designed to process and generate human-like language. Trained on vast amounts of text data, these models can perform tasks including language translation, summarization, sentiment analysis, and developing coherent, context-specific text.

How LLMs Work

LLMs utilize transformer model architecture—neural networks designed for sequence-to-sequence tasks. The encoder processes input text and generates a continuous representation, while the decoder takes this representation and generates output. Training involves exposing the model to large text datasets and fine-tuning for specific applications.

Leading Language Models: March 2025

GPT (OpenAI)

The latest ChatGPT-4o models offer faster processing and enhanced capabilities across text, voice, and vision, with a 128,000 token context window. These proprietary models excel in conversational dialogue, reasoning, and real-time interactions.

Claude (Anthropic)

Claude 3.7 Sonnet, released in February 2025, represents Anthropic's most advanced model and a significant evolution in the Claude family. Building upon the foundation of Claude 3.5 Sonnet, this new iteration features enhanced reasoning capabilities and an expanded 200,000 token context window (approximately 150,000 words or 300 pages of text).

Claude 3.7 Sonnet has a unique feature called "reasoning" mode" that allows it to think through problems more deeply before giving answers. This feature is only available to people who subscribe to the Pro account, and it helps the model tackle complex questions, math problems, and situations that require multiple steps to solve. It also does a great job of understanding language, picking up on subtle details, and keeping conversations clear and consistent, even over a long period.

DeepSeek

DeepSeek-R1 is a 671B parameter Mixture-of-Experts model with 37B activated parameters per token. It demonstrates superior mathematics and code generation performance while offering greater cost-efficiency than many competitors.

Mistral

Mistral Small 3, a 24-billion-parameter model, processes approximately 150 tokens per second—over three times faster than some competitors. It's ideal for low-laIt'sy applications and can be deployed on devices with limited computational resources.

Other Notable Models

LG AI's EXAONE 3.0, Meta's laMA 3.3, Google's Gemini 2.0 Flash, and Cohere's Command R's unique capabilities are suited to different applications, from multilingual processing to retrieval-augmented generation.

The Artist's Role in the Artificial Intelligence

The popular use of LLMs is excitingly changing the way artists work. With new AI tools, artists can spend less time on repetitive tasks and more time on their creative ideas. LLMs have changed artist work processes so they can discover novel ideas while expanding their creative boundaries beyond previous limits.  

This technology directs the world to question human creative involvement while addressing vital aspects of human creativity's future. Future artists must develop creative methods for AI tools, so their original artistic qualities remain unaltered. Artists who accept AI tools for their creative work can explore new artistic ground and achieve uncharted artistic outcomes.

Collaboration and Co-Creation

Artists use co-creation techniques with AI to drive their artistic mission in the artist's AI era. Through collaborations with AI tools, artists gain access to new ideas that collaborate with multiple artistic styles as they implement automation features for art production. A combination of human and machine creativity emerges from partnerships, resulting in novel artistic compositions.

The synergy between artists and AI tools can push the boundaries of artistic expression, resulting in previously unimaginable new art forms. Artists and AI systems work together to generate art pieces that demonstrate superior qualities from human and computer skills, enhancing the artistic frontier.

Transparency and Disclosure

Openness about artificial intelligence tools and public disclosure remains essential for all operations related to AI-produced artwork. Artists, designers, and creatives must reveal their AI tool usage so human creative input remains appropriately acknowledged. The use of AI technologies should be indicated during creation alongside measures to safeguard human artistic rights.

Through openness and transparency, artists maintain audience trust, which leads to proper recognition of their work. Honesty about using AI tools fosters trust and highlights the art's innovative nature, celebrating the collaboration between human creativity and artificial intelligence.

Challenges and Future Directions

Despite their impressive capabilities, large language models face several challenges:

• Bias and Fairness: LLMs can inherit biases from training data, potentially leading to discriminatory outcomes
• Explainability: The complexity of LLMs makes it challenging to understand how they reach conclusions
• Efficiency: Training and deploying LLMs require significant computational resources
• Security: LLMs can be vulnerable to adversarial attacks

Researchers address these challenges through multitask learning, transfer learning, explainability techniques, and more efficient architectures.

Conclusion

The transformation of artistic fields through AI results from increased creative capabilities and optimized operational performance of innovative activities rather than from replacing human artistic talent. These evolving technologies will equip creative sectors to transform their boundaries and improve operational tasks through their arts sector integration. It becomes vital to differentiate AI-generated artwork from human-made pieces because the growth of AI art could lower the value of human artistry while facing obstacles for artists. Future AI developments will succeed by combining conscious creative ventures with AI capabilities to provide new artistic possibilities to audiences, artists, and art institutions.