I hate the pop sci ill-defined notion of AGI. As soon as a task is defined, an AI is developed which completes the task from real world data with superhuman suc

I hate the pop sci ill-defined notion of AGI. As soon as a task is defined, an AI is developed which completes the task from real world data with superhuman success. The work of making the superhuman model isn't even conceptual usually, it's a matter of dispatching training jobs. It's quite clear that if your definition of AGI is superhuman perf at arbitrary tasks, there's no conceptual barriers right now. Everything is mere scale, efficiency and orchestration.