I'm drowning in AI features I never asked for and I hate it (www.makeuseof.com)

A growing backlash from a long-time user argues that AI has moved from useful augmentation to intrusive checkbox features that often degrade products. The author cites real-world regressions—Google’s Gemini replacing Assistant but introducing LLM inference delays and failing simple tasks like turning on lights, Apple Intelligence worsening Siri, Microsoft’s Copilot shoved into Windows UI, and browsers like Arc’s Dia becoming indistinguishable from Comet/Atlas. Beyond UX harm, the piece describes the “dead internet” effect: forums, social media, and search results flooded with recycled, AI-generated content that amplifies spam, rewards engagement farming, and erodes authentic human voices. For the AI/ML community this is a wake-up call about deployment trade-offs. Key technical issues raised include inference latency and reliability (making assistants slower and less dependable), hallucination-prone summarization in search, opaque training data provenance (models built on scraped creative work without consent), and persistent telemetry—every prompt or voice query becoming training data that enables powerful profiling. The implications point to priorities: optimize latency and accuracy, improve provenance/transparency and consent, offer on-device options where feasible, and design AI as supporting features rather than product-defining gimmicks. Ignoring those will harm user trust and the long-term adoption of generative AI.