Custom Applications with AI - Are they required?

Is there any scope for custom programming of AI Models?

It's undeniable that modern AI platforms like Gemini and ChatGPT have become incredibly adept, handling a vast array of tasks with remarkable ease. This widespread capability, however, prompts a deeper inquiry: What is the true scope for custom programming of AI models? I've posed this very question to several AI systems myself, only to be met with a frustrating lack of definitive answers, leaving the path forward feeling murky.

For instance, I remain unconvinced that large language models (LLMs) are poised to replace fundamental graphical user interface (GUI) elements such as buttons or dropdown menus. Their strength lies in language and pattern recognition, not in directly redesigning our digital interfaces.

When we consider practical applications, many processes are already a well-oiled machine of automation. Banking transactions are digitized, airline tickets are booked online — these aren't new territories crying out for AI intervention. The question then becomes: where does this advanced, custom AI truly fit?

Even the concept of using AI to orchestrate a complex chain of processes isn't entirely novel. Traditional programming has long managed such sequences with sequential modules, patiently waiting for a 'flag' or trigger in a database to signal the next step. It's like a finely tuned domino run, where each piece falls precisely when cued.

Similarly, while generative AI is undoubtedly impressive, its current frontier appears largely confined to creating images or videos from textual prompts. So, amidst all these established capabilities and existing solutions, where exactly lies the genuine opportunity for bespoke, customized AI development? What unique problems are still waiting to be solved by tailor-made AI, rather than off-the-shelf solutions or traditional code?

Conclusion: Well, I am myself looking for answers to these questions!