The reality behind AI hype

As governments and tech leaders gather at global forums such as the AI Impact Summit in New Delhi, one assumption dominates discussion: the more computing power poured into AI, the better it will become. In his blog ‘‘The elephant in the AI room’: Does more computing power really bring more useful AI?’, Jovan Kurbalija questions whether that belief is as solid as it seems.

For years, the AI race has been driven by the idea that ever-larger models and vast GPU farms are the key to progress. That logic has justified enormous energy consumption and multi-billion-dollar investments in data centres. But Kurbalija argues that bigger is not always better, especially when everyday tasks often require far less computational firepower than frontier models provide.

He points out that most people rely on a limited vocabulary and a small set of reasoning tools in their daily work. Smaller, specialised AI systems can already draft emails, summarise meetings, or classify documents effectively. The push for trillion-parameter models, he suggests, may reflect ambition more than necessity.

There are also technical limits to consider. Adding more computing power can lead to diminishing returns, and some prominent researchers doubt that simply scaling up large language models will lead to human-level intelligence. More hardware, Kurbalija notes, does not automatically solve deeper conceptual challenges in AI design.

The economic picture is equally complex. Training cutting-edge proprietary models can cost hundreds of millions of dollars, while newer open-source systems have been developed at a fraction of that price. If cheaper models can deliver similar performance, questions arise about the sustainability of current spending and whether investors are backing efficiency or hype.

Beyond cost and performance lies a broader ethical issue. Even if massive computing power could eventually produce superintelligent systems, the key question is whether society truly needs them. Kurbalija warns that technological possibilities should not be confused with social desirability, and that innovation without a clear purpose can create new risks.

Rather than escalating an arms race for ever-larger models, the blog calls for a shift toward needs-driven design. Right-sized tools, viable business models, and ethical clarity about AI’s role in society may prove more valuable than raw computing muscle.

In challenging the prevailing narrative, Kurbalija urges policymakers and industry leaders to rethink whether the future of AI depends on scale alone or on smarter priorities.

Would you like to learn more about AI, tech and digital diplomacy? If so, ask our Diplo chatbot!