AI development path
The current approach to AI training (based on existing data) is efficient now but will become limiting if the industry doesn't shift towards more autonomous learning, despite the commercial pressures.
- It's possible for AI to improve recursively and learn more than what it's given, similar to how more primitive AIs have been trained.
- Current AI training methods, based on existing information, are the most efficient right now but severely limit AI's potential capabilities.
- The industry is stuck in a cycle driven by competition and the pursuit of money/power, preventing a shift away from the statistical machine approach, which will eventually limit AI's development.


