A wave of scepticism is now sweeping over generative AI. The initial excitement is giving way to questions, and while growth figures continue to soar, doubts are surfacing: is the hype of this still-young technology overblown? Can the promises be fulfilled?
AI is not just generative
It is often too easily overlooked that the story of generative AI is merely one chapter in the much broader history of artificial intelligence.
This journey began in the 1950s with pioneering experiments, notably by Warren McCulloch and Walter Pitts, followed by Alan Turing. Since then, AI has made spectacular advances, particularly in the last decade.
The introduction of Generative Adversarial Networks (GANs) in 2014, followed by the Transformer architecture in 2017, marked significant milestones.
The arrival of ChatGPT in 2022 propelled generative AI into the spotlight, demonstrating capabilities far beyond simple content creation, encompassing advancements in image generation, audio synthesis, code generation, and even complex decision-making models.
As a direct result of these major advances, the AI market has quadrupled over the last four years, reaching €490 billion in 2023. This, however, is just the beginning: the latest study by Sopra Steria Next projects that the AI market will grow three times faster than the global technology sector by 2028.
It is therefore unsurprising that the generative AI market is experiencing a similar trajectory, with forecasts predicting an expansion from €10.6 billion in 2023 to €208.8 billion by 2030. However, this growth comes with formidable challenges.
The flip side of the coin
The costs associated with computing power and model training are enormous. The figures are staggering: each conversation with ChatGPT costs several cents, potentially amounting to millions in operational expenses.
Training GPT-3 alone required an investment of €4.6 million for a single session while the costs for GPT-4, given its increased complexity and scale, are estimated to have soared into the tens of millions, potentially reaching as high as €50-100 million.
Scaling remains a significant challenge: fewer than a third of projects reach the production stage (or only 1 in 7 algorithms make it to production – according to the Sopra Steria Next study). With over 500 generative AI companies launched in 2023, market saturation is becoming increasingly apparent.
Yet, despite these hurdles, generative AI continues to demonstrate tremendous potential across a wide range of sectors. To fully appreciate the scale of the transformation underway, consider the following examples:
In healthcare, generative AI accelerates drug development and enhances the precision of medical imaging. In accessibility, it restores voices to those who have lost them and helps the visually impaired interpret the world around them.
In environmental science, it enables more accurate climate forecasts and aids in designing sustainable products. In scientific research, it revolutionises protein studies and speeds up the discovery of new materials.
Striking the right balance
The promises of generative AI are undeniable, but caution is still warranted. "LLM washing" — the tendency to portray generative AI as a miracle cure for all challenges — would not only be a critical strategic error, but could also cause us to overlook other equally powerful, or even superior, AI tools.
To move beyond the initial wave of excitement, it is essential to foster a deep-rooted AI culture while preserving subject-matter expertise. A well-defined AI strategy, aligned with the company’s economic and societal objectives, should guide its adoption. High-impact pilot initiatives will demonstrate its real value, well beyond the early fascination.
These initiatives must also embed ethical considerations and responsible AI practices into every step. It is crucial to remember that, like any technology, generative AI is, first and foremost, a tool — undoubtedly powerful, but one that demands careful and informed application.
Avoid burning out what we once adored
Generative AI has made remarkable progress in recent years. It has evolved from simple content generation to demonstrating emerging reasoning capabilities, still in development, but continuously pushing the boundaries of what AI can achieve.
Multi-agent systems, for instance, are bringing it closer to human-like problem-solving and decision-making processes.
I am convinced that the future of generative AI, as part of a broader technological ecosystem, holds great promise. Its success will rely on our ability to balance groundbreaking innovation with pragmatism, while realigning our expectations with what the technology can truly deliver.