Author Warns AI Could Fail Under Software Bloat, Citing Wirth's 1995 Law

1 articles · Updated · CleanTechnica · Jun 20

Neural-network AI may be headed for failure because software inefficiency is outpacing hardware gains, the author argues, warning that the risk extends beyond power shortages.
A 1995 essay by Niklaus Wirth underpins that case: software gets slower faster than hardware gets faster, making ever-larger AI systems more bloated, sluggish and costly to run.
A recent CMS upgrade and decades of personal computing experience are used as evidence that added features often make software harder to use without improving results.
AI's heavy infrastructure needs sharpen the concern: data centers are expanding rapidly, some backers are eyeing nuclear power, and the author notes a plant is only about 35% efficient.
Large AI investment could become an economic shock if the technology proves overbuilt or underperforms, the author says, calling for oversight before more money is committed.