Topic Battle

AI adapts at speeds that make evolution look like a particularly bureaucratic planning committee. Transfer learning allows a system trained on cat photographs to recognise tigers within hours. The Cambridge Computational Adaptation Centre demonstrated an AI that learned to play chess, then pivoted to protein folding, then to predicting traffic patterns—all before a tiger could finish digesting a single meal.

Yet this adaptability comes with a crucial caveat: AI adapts only within parameters humans have defined. When researchers at Imperial College's Edge Case Laboratory presented an image recognition system with a tiger wearing a hat, the system confidently identified it as a lamp. The tiger, presumably, would have identified the researcher as lunch regardless of headwear.

The sustainability profile of artificial intelligence makes environmentalists reach for strong drink. Training a single large language model generates approximately five hundred tonnes of carbon dioxide—equivalent to the lifetime emissions of five average British citizens. Running the world's data centres requires more electricity than several medium-sized nations. The Edinburgh Energy Audit Commission projects that AI's energy demands will triple by 2030.

Furthermore, AI depends upon rare earth minerals extracted through processes that would make even the most ruthless tiger wince, supply chains spanning dozens of countries, and the continued stability of global semiconductor manufacturing. A single well-placed solar flare could render the world's AI infrastructure into very expensive paperweights.

Artificial intelligence has inserted itself into virtually every aspect of modern existence. It recommends your entertainment, filters your communications, approves your loan applications, and increasingly makes decisions about your medical treatment. The Stockholm Digital Penetration Index estimates that the average person in developed nations interacts with AI systems over four hundred times daily, mostly without awareness.

AI's influence operates not through fear—though that is certainly accumulating—but through sheer infrastructural ubiquity. One cannot opt out of AI influence any more than one can opt out of electricity. Tigers, by contrast, are relatively easy to avoid if one simply refrains from walking through South Asian jungle at dawn.

Artificial intelligence hunts differently. It stalks data points across vast digital savannahs, identifying patterns with a success rate approaching ninety-seven percent in controlled environments. The Oxford Algorithmic Predation Lab notes that modern AI systems can identify a specific human face from a database of twelve million in under two hundred milliseconds—roughly the time it takes a tiger to blink.

However, AI's predatory efficiency depends entirely upon reliable electrical infrastructure, climate-controlled server rooms, and the continued existence of the internet. A tiger's hunting capability depends on the continued existence of prey. The prey, one notes, are not currently working on ways to switch tigers off.

The existential threat potential of artificial intelligence remains hotly debated, primarily because disagreement exists over whether AI could destroy humanity accidentally through optimisation misalignment, deliberately through emergent goals, or indirectly through enabling human stupidity at unprecedented scale. The Cambridge Centre for Existential Risk has published fourteen hundred pages on the subject without reaching definitive conclusions.

What remains clear is that AI's threat scales differently. A sufficiently capable AI system, connected to sufficient infrastructure, could theoretically affect billions of lives simultaneously. Whether this constitutes a feature or a bug depends largely upon who controls the system and whether they have read sufficient science fiction to know better.