Why Does the Empowerment of AI Seem Ineffective?

Why might you feel that the empowering effect of AI isn’t obvious, and that you still haven’t grasped this new wave of productivity?

  1. Ability to connect to AI tools: IP restrictions and the need for stable accounts have stumped countless people.
  2. Ability to subscribe to Plus versions: The willingness to pay, the financial means, and the payment channels have stumped countless more.
  3. Ability to figure out model selection and features: Faced with a dazzling array of models (currently 8 model options and 6 feature options) and features like memories, Projects, reference all chats, and connectors, countless people are bewildered.
  4. Ability to do prompt engineering: Customizing models, setting project instructions, and managing the context of each conversation have stumped countless people.
  5. Ability to form a habit of using AI: Making it a first instinct to ask AI, using AI to solve problems in life, academia, or engineering, and accumulating a long history of AI conversations have stumped countless people.
  6. Ability to ask good questions: The required academic taste, engineering intuition, and one’s own cognitive abilities have stumped countless people.
  7. Ability to build one’s own AI applications: Local deployment, API calls and hyperparameters, building custom applications, Retrieval-Augmented Generation (RAG), and framework selection have stumped countless people.

All I can say is, this has already become a technology with significant barriers to entry, far from being a mere toy or something for students. When used well, it’s no exaggeration to say it can make one unparalleled below the PhD level. A “PhD-level AI” might be a bit of an overstatement, but a person amplified by the super-leverage of AI? That’s a completely different story.