InnovationManagementMarketingEconomics and Finance
Sign In Subscribe

Innovation

Transposing Creativity: The New Capabilities Unlocked by AI Tools

By Jaroslav Urbanovič | January 30, 2026
Business Insights | Innovation

AI In Programming for Beginners

A 2024 study demonstrated that novice programmers using ChatGPT produced higher-quality code; they violated coding conventions less frequently and generated solutions of lower complexity compared with the control group. This suggests that large language model (LLM) assistants (e.g., GPT, Claude) help beginners write cleaner, clearer code while adhering to established best practices. Beyond accelerating execution, AI functions as a form of scaffolding: it guides learners toward better structural decisions and supports the acquisition of foundational programming skills.
This aligns with a broader pattern observed in practice, where individuals with no formal technical background successfully develop simple scripts, prototypes, and automation workflows by iteratively refining AI-generated code. In these contexts, AI serves less as a replacement for learning and more as an interactive tutor, offering reasoning steps, explanations, and refinements that help users grasp core programming concepts.

The Shift Toward AI-Assisted Creation

While AI-supported programming enhances the learning trajectory for beginners, generative models introduce a broader transformation: they increasingly abstract away the need for traditional programming skills altogether. Users without technical expertise can now produce functional applications within hours by relying on AI code generators that translate natural-language descriptions into executable HTML, CSS, and JavaScript (and others). These emerging “vibe-coding” practices, where individuals co-construct software with an AI model, illustrate how natural-language reasoning is gradually replacing the demand for syntactic precision.


A notable real-world example of this “vibe-coding” approach was the “AI-Only Hackathon” organized by Tesonet, a prominent Lithuanian venture builder and startup investor. During this event, teams competed by building functional products (MVPs) using AI agents instead of traditional coding, relying almost entirely on natural-language instructions. The hackathon demonstrated how AI-assisted creation is swiftly becoming a viable development paradigm, even at the scale of competitive innovation environments.
Importantly, this trend differs from beginner-oriented AI assistance. Here, the primary goal is not to acquire programming skills but to accelerate production, rapidly test ideas, and iterate prototypes. AI becomes an accessible digital manufacturing layer, allowing creators to focus on conceptual design rather than technical implementation. This abstraction extends beyond software development: modern generative tools streamline the creation of brand assets, marketing materials, music compositions, short clips and films, and interactive media, significantly expanding what individuals can build with limited technical training.

A Contradictory Perspective: Limitations in Complex Tasks

However, recent research provides a necessary counterbalance to this optimistic view. A 2025 study evaluating the performance of ChatGPT-4o across Python and Java found that while the model performs well on simpler programming tasks, its effectiveness declines sharply as complexity increases. The authors observed several limitations:


• Runtime inefficiency, especially for Python solutions.
• Weak exception handling and insufficient error management.
• Limited documentation quality, reducing code maintainability.
• Functional but non-optimal solutions, unsuitable for real-world deployment.


Moreover, Java-generated code exhibited better runtime performance, whereas Python outputs were more memory-efficient, but neither language consistently produced high-quality solutions at higher complexity levels. These findings suggest that although AI can substantially accelerate prototyping, it has not yet reached the reliability required for complex, production-grade systems.

Scaling Capabilities

A crucial contextual note is that this contradictory study evaluated ChatGPT-4o, whereas the current frontier of AI-assisted coding from OpenAI is GPT-5.1-Codex-Max, which is being promoted as much more capable in complex long tasks. Public reports and performance benchmarks for 5.1-Codex-Max remain limited, and comprehensive academic evaluations have not yet been published. Therefore, future research comparing ChatGPT-4o, GPT-4.1, GPT-5.1-Codepublicor models from other companies will be essential for understanding whether the limitations documented in beginning of 2025 persist or whether newer models meaningfully improve robustness, optimization, and real-world applicability.
In sum, AI-driven making is advancing rapidly, but the evidence shows a nuanced picture: AI expands who can create software and digital content, yet its reliability for complex development remains uneven and requires scrutiny as newer models evolve.

Generative AI In the Arts

A large-scale analysis of 4 million digital artworks created by more than 50,000 authors revealed that text-to-image generative models (such as Midjourney or Stable Diffusion) significantly increase creative productivity by approximately 25% and even raise the “value” of created works by as much as 50%. While peak Content Novelty (focal subject matter and relations) increased, demonstrating exploration at the creative frontier, the average Content Novelty and all Visual Novelty (pixel-level stylistic elements) measures declined, because it produces a considerable amount of stylistically similar content. Nevertheless, the study confirms that AI tools can meaningfully enhance the creative potential of amateurs and reshape the creative process.


Generative AI promotes a more inclusive creative domain by making value capture (favorites earned) less concentrated among adopters. The ability to produce more valuable works is linked to the artist’s capacity to successfully explore novel ideas and filter model outputs for coherence, rather than prior originality.


Authors of the study refer to this process as “generative synesthesia”, a hybrid creative mode where humans steer ideation while models expand the search space. This points to a future of augmented creativity, in which the creator’s intent becomes the primary driver and AI handles execution.

The Debate: Empowerment Vs. Authenticity

Technology commentators emphasize that AI’s greatest benefit in creative work is the democratization of idea realization. Without specialized training, more people can experiment and create content, from film scripts to artworks, because AI enables the “home creator” to execute ideas and share them with the world, potentially triggering an explosion of new perspectives in the creative market.


However, this empowerment also brings debate. Critics argue that easily generated content is not equivalent to traditional creative work. Generative models are trained on existing works, often without direct permission from their authors, raising ethical and compensation questions related to original creators’ rights. It is also stressed that AI does not eliminate the need for human creativity, such as idea generation, problem-solving, and the development of a unique artistic voice. Thus, AI should be understood as a tool that reduces the technical threshold, while genuine creative work still requires human contribution.

Afterthoughts

Artificial intelligence is transposing the power of creative production by lowering technical barriers and enabling individuals with little or no specialized training to participate meaningfully in fields such as programming, design, and digital content creation. Tools powered by generative models allow users to transform abstract ideas into fully realized outputs, whether code, visuals, or written material, through natural-language interaction rather than technical mastery.


This surge in creative capability expands who can contribute to innovation and cultural production. However, it also surfaces questions about authorship, authenticity, and ethical data use, particularly as AI-generated content blends seamlessly with human work. The challenge ahead lies in balancing empowerment with responsible adoption, ensuring that AI amplifies human creativity rather than diluting its value or obscuring its origins.

References