The problem is the assumption that 'AGI' has anything to do with a bipedal physical body that resembles ours. We evolved the way we did due to evolutionary con

The problem is the assumption that 'AGI' has anything to do with a bipedal physical body that resembles ours. We evolved the way we did due to evolutionary constraints. The constraints of the evolution of our silicon computing devices are completely different. Why would you hook all that gear up to a single automaton, with a head, two eyes, two arms and two arms. Those are completely arbitrary. Why not hook it up to 100's or millions of cameras? Thousands of 'limbs'? i.e. ever camera and operable apparatus connected to the internet? Our 'silicon' will have a recall advantage that our pure-NN brains do not, which is consistent lookup of data and 'perfect recall' of every bit of human data every recorded, which means that it's evolution will be quite different as well. We'll have distributed systems that go way past the 'Turing Test' decades before anything remotely approaching AGI. 'Siri' is almost there, in 10 years I'll bet it becomes difficult to tell if Siri is a person or not. And Siri isn't even 'one entity' - it's a generic interface to other, distributed systems. 'Siri' in 10 years will start to resemble AGI but with other, more fantastic abilities (information access and synthesis). Imagine in 50 years your Tesla is deeply integrated with satellite GPS location data, weather data, it's 'talking' to the 'local road controller' for control data with ms latency, and also 'talking' to every other car within 500m to get relevant positioning and other sensor data. Maybe some of the relevant AI processing requires far too much power, and it's done in real-time at the closest data centre via networking. At that point you have to say the Tesla isn't really the vehicle in front of you, it's an entire system of technology. Sun Microsystem's motto was 'The Network Is The Computer' implying to the extent that all computers are comprehensively connected ... they form at least some kind of coherent entity. The concept of AGI is so abstract that it almost doesn't make any sense. By the time we are able to assemble a discrete group of computing components, to a discrete bipedal robot (i.e. a Metropolis-like automaton, i.e. Data from Star Trek) we will already be many decades ahead of that in terms of understanding how all of that maps to consciousness. We'll already have intelligent systems doing things way beyond what any human could do and probably what we could imagine right now.