Is it too late to contribute to AI? (Andrew Ng) (www.deeplearning.ai)

Andrew Ng responds to an 18-year-old worried that it’s “too late” to contribute to AI, arguing the opposite: despite rapid progress and sensational headlines, modern AI is powerful but still limited and far from human-level general intelligence. He stresses that frontier LLMs are excellent at text tasks but remain brittle, require substantial customization (he cites his team’s labor-intensive AI résumé screener), and are poor at many real-world duties businesses give juniors—calendar triage, hiring decisions, or autonomous prioritization—without careful engineering and supervision. Technically, Ng highlights that current models are modality-biased (text-centric), need extensive context engineering, and lack efficient mechanisms to learn from repeated exposure or feedback for narrowly defined tasks. AGI — a system that can perform any intellectual task a human can — is likely decades away, so many application-layer opportunities remain safe from immediate displacement. His takeaway for the AI/ML community: don’t be deterred by hype; there’s plentiful, impactful work in building, customizing, aligning, and deploying AI systems today, and now is an excellent time to learn to build with AI rather than opt out.