In Other News: Human Harms - October, 2025

Oct 6, 2025

A World Beyond LLMs?

A new Apple study confirms that LLMs fail at true reasoning, relying on pattern matching rather than logic. Experts warn that current evaluation methods overstate LLM capabilities, and some researchers argue that advancing AI will require moving beyond LLMs toward other approaches that better mimic human understanding.

Oct 1, 2025

Data Centers: The Canary in the AI Hype Coalmine?

AI infrastructure is now a major driver of U.S. economic growth, but it's being financed through increasingly complex and risky financial strategies that echo the subprime mortgage crisis. As tech giants and startups rack up debt to fund data centers and cloud capacity far ahead of actual demand, experts warn of mounting systemic risk, unsustainable valuations, and potential fallout for both investors and local communities.

Oct 1, 2025

AI Bubble Watch

Despite massive VC investment, many AI startups are struggling to deliver returns, with valuations skyrocketing even as most projects fail to generate meaningful revenue and the costs of model training and inference rise. Experts are cautioning that the market may be more overvalued than in the dot-com era.

Oct 1, 2025

All Eyes on Agents

AI agents are generating major interest, with both Big Tech and VC rapidly investing in the space. But many so-called "agents" are just rebranded chatbots or automation tools. Despite their promise, real-world performance remains inconsistent, and concerns over reliability, misuse, and integration challenges are leading some analysts to warn of high failure rates.

Oct 1, 2025