[starchild3001]

I pretty much agree with the perspective in this paper. That is, depending on your definition, either

1) AGI has already happened (in a narrow domain of question answering, machines can answer any question & converse, nearly as well as an average human)

2) AGI requires biological embodiment (such as real time learning, growth, long term memory, various real life motives and behaviors) therefore it will never happen with current devices.

[singularity2001]

3) AGI has almost happened, the current integration of real time visual, spatial and audio inputs with logical reasoning and different layers of learning (beyond context window and 'memory') just doesn't satisfy even the most benevolent definitions yet