From AGI to Superintelligence: The Intelligence Explosion(situational-awareness.ai)
situational-awareness.ai
From AGI to Superintelligence: The Intelligence Explosion
https://situational-awareness.ai/from-agi-to-superintelligence/
3 comments
There still needs to be a lot of model development, tools, systems integration, data compression, power savings, and quite a lot of other things before AGI is anywhere near.
Besides Ray Kurzweil, the most interesting and deep thinking exposition on the coming AI transformation.
The scalar nature of intelligence is more like points on a sphere than a hockey stick graph of good answers. How much area/depth of a sphere can an intelligence cover without becoming vulnerable to expectations or predisposition? All of it? I think impractical. Everything dies, everything falls apart, everything has points of failure. Even intelligence, super or not.
Intelligence (the mitigation of uncertainty) is highly dependent upon information (the removal/reduction of uncertainty), which is dependent upon context and disposition. "Why?" (And "who cares?") for instance.
Sure there is raw computing power, and some linear correlations, however life systems optimize iteratively such that precision and capacity structurally change.
A wolf's hunting strategy vs a scientist's for instance. Who is "more intelligent?" I'm sure you'll say we can hunt wolves though I would counter, mostly by superior technology, not necessarily by intelligence.
Super intelligence can be super naive (different problem domains specialize.)
Super intelligence wouldn't have the same needs as Man. Sure such could zoom off into the cosmos, though who cares? We can take our own trip in due time. Solving problems for others is no great service as self reliance is a blessing more than a curse.
Imagining a gradient where Man cannot peer merely by finite scalar capacity is not an accurate depiction of this domain.