Andrew Ng's letter to an 18 y.o. worried about AI (www.deeplearning.ai)

Andrew Ng responded to an 18-year-old worried that AI will make it “too late” to contribute to the field, reassuring them that meaningful work will remain for decades while warning against hype. He argues that although modern LLMs are impressively general compared to prior AI, they’re still “incredibly dumb” for many practical business tasks (e.g., prioritizing calendars, running reliable résumé screens) unless engineers invest significant customization and context engineering. He highlights concrete technical gaps: models are strongest with text, weaker across other modalities, and we lack efficient tools for systems to learn from feedback or repeated exposure to a specific task. For the AI/ML community this is both a caution and a call to action. Ng emphasizes that AGI—systems that can do any intellectual task humans can—remains decades away, so career and startup opportunities at the application layer are far from exhausted. The implication is that engineers who learn to build with and around current models (custom pipelines, fine-tuning, feedback loops, domain-specific engineering) will be highly valuable, while simplistic wrappers around LLMs are the ones most at risk from model improvements. His message encourages precise, realistic understanding of capabilities and continued investment in practical systems work.