The biggest problems in using AI
The article discusses key challenges in using AI, including issues of reliability, bias, security vulnerabilities, and the difficulty of aligning AI systems with human values. It highlights that current AI models can produce incorrect or harmful outputs, and that ensuring safety and fairness remains a significant technical and societal problem.
Background
- **Alex Shearer** is an AI strategy consultant and newsletter writer. This piece targets managers and executives deploying large language models (like GPT-4) in real organizations.
- "Biggest problems" here means *practical deployment failures* — hallucinations, unreliability, latency, cost, and difficulty evaluating outputs — not existential risk or AI alignment.
- Key distinction: **AI capability vs. AI reliability.** Shearer argues the harder problem is making models trustworthy enough for business workflows, where even a 1% error rate breaks the product.
- The article reflects a 2024–25 industry reckoning: prompt engineering and RAG (retrieval-augmented generation) are not silver bullets. Teams find frontier models still fail on simple factual tasks, and evaluation tooling remains immature.