7-year CV engineer on the genAI hype: where are the hard non-LLM problems?
SchwingBesen · reddit · 2026-11-09
A machine learning / computer vision engineer with 7+ years of experience vented on Reddit about the current generative AI hype: after years training custom vision models, working with time-series data, and architecting multimodal pipelines from scratch, he finds the industry's equation of AI with prompt engineering and LLM wrappers reductive.
He wants to build software with genuine real-world impact — environmental science, specialized B2B workflows, complex physical monitoring — where off-the-shelf generative APIs don't cut it, and asks peers still doing core ML/CV: which high-impact non-LLM problems desperately need custom models and domain expertise, and how do they stay motivated for foundational ML craft in an industry obsessed with generative text and images?
More from AGI Musings
- Copy the memory, not the permission: a test case for AI agent obligations — tallmetommy · 2026-11-09
- Hinton: don't dominate superintelligence, design it to care like a mother — danfaggella · 2026-11-09
- Are standalone AI agent chatbots the wrong interface for real team work? — notinvariant · 2026-11-09
- AI capability has increasing returns: 100 points worthless, 170 earns $100bn/yr, 240 could 'take over the world' — ben_j_todd · 2026-11-09
- 'The slop will continue until morale improves': BBC mocked for cutting foreign correspondents while churning out low-quality content — sebkrier · 2026-11-09
- AGI leader's 2010s plan — build a lead, then unilaterally slow down for safety — never worked — S_OhEigeartaigh · 2026-11-09