Topic Battle

AI systems require extensive retraining to adapt to new domains. A model trained for medical diagnosis cannot spontaneously begin writing poetry without substantial additional development. Transfer learning has improved this limitation somewhat, but AI remains fundamentally narrow in its capabilities.

Furthermore, AI cannot adapt to physical changes in its environment because it possesses no physical presence. If global temperatures rise by two degrees, otters will migrate and adjust their behaviour. AI will continue operating identically until its data centres flood or lose power, at which point it will simply cease to exist.

Training a single large language model requires approximately 1,287 megawatt-hours of electricity, equivalent to the annual consumption of 120 British households. Data centres housing AI systems now consume more electricity than entire nations, with projections suggesting AI could account for 3.5% of global electricity demand by 2030.

The Cambridge Centre for Computing Sustainability calculates that generating a single AI-written essay produces the carbon equivalent of driving a car 4.7 kilometres. An otter, by contrast, can travel the same distance while actually improving local ecosystems.

AI systems have achieved remarkable capabilities in simulating social interaction. Chatbots can now maintain conversations that many users find indistinguishable from human discourse, at least for the first several exchanges.

However, as documented by the Manchester Institute for Human-Computer Relations, AI possesses no actual social intelligence whatsoever. It cannot form genuine relationships, experience loneliness, or derive satisfaction from companionship. When an AI system claims to "enjoy" a conversation, it is, to use the technical terminology, lying.

AI systems can process millions of calculations per second and have defeated world champions at chess, Go, and various other games humans previously considered exclusively cerebral pursuits. Large language models can now pass bar examinations and medical licensing tests.

However, as documented by the Oxford Artificial General Intelligence Review Board, current AI systems struggle profoundly with tasks any otter handles instinctively. When asked to open a jar, AI can generate 47,000 words describing the process but cannot actually perform the action. The distinction between knowing and doing remains, as one researcher noted, "rather significant."

Leading AI researchers, including those who created the technology, have signed open letters warning that AI poses potential extinction-level risks to humanity. The Future of Humanity Institute at Oxford places advanced AI among the top five existential threats facing our species.

Current concerns include autonomous weapons systems, economic disruption affecting hundreds of millions of workers, and the longer-term possibility of artificial general intelligence pursuing goals misaligned with human welfare. Several prominent researchers have compared the situation to "building a nuclear reactor before understanding physics."