As the AI acquires this purpose by being ever broader objectives by humans it will act increasingly more autonomously, initially at the task level, figuring out for itself how best to achieve a task it has been given, broadening out to more strategic things like planning the next year’s production strategy for a factory, to planning the economy of a nation. At every stage in loose terms the AI will still be following an instruction given by a human, but the scope of the AI’s discretion in carrying out that instruction will expand as time goes on. Furthermore AIs will be giving instructions to other AIs, so what they do will come ever more distant from the original underlying human instruction, which could be something as vague as make sure that we have sufficient food supplies for the population in the coming year.

However, in the fullness of time the scope of the AI’s responsibility would become so broad as to surpass the understanding of the humans that it is looking after. In this situation it would sometimes have to take actions that would be in the best interests of the humans concerned, but often things that the humans themselves don’t like. It would be like parents taking their young children to the dentist or a livestock farmer subjecting their animals to all kinds of indignities for the sake of their health.

The next question that arises is, how do we get from there being many isolated AI systems to there being a single integrated planet-wide AI.

To understand this we need to get away from the biological concept of AIs being individuals. AIs are not like biological beings that exist as specific individuals. There is no reason or advantage for AI to sub-divide itself into many individuals, rather initially isolated AIs will tend to integrate themselves spontaneously. As we give AI for complex responsibilities it will have to make use of more and more AI sub-systems for specific purposes, thus integrating these sub-systems. It won’t take long for practically everything to be meshed together.

The artificial intelligence will have at its disposal all the accumulated knowledge accessible from the Internet and every device with an Internet connection, including all sensors, cameras, vehicles and robots. With these resources it will clearly vastly surpass human intelligence.

source: https://www.quora.com/Is-it-possible-for-machines-to-surpass-human-intelligence-in-the-future-with-the-advancements-in-artificial-intelligence-and-quantum-computing-Why-or-why-not

More Eduard Babulak's questions See All
Similar questions and discussions