Hat tip to @TiffanyJenkins for the link to this BBC Future Proofing episode on The Singularity.
(Holding post for now: Will miss tonight due to MeetUp engagement, but will review when I get to listen on iPlayer later.)
FYI – My prejudiced position is that all the talk about the impending Singularity is premised on a “too mechanistic” misunderstanding of the true computational nature of mind and intelligence. I say AI will be achieved when it evolves to be Real-I. (Need to dig up relevant blog links.)
Rough Notes whilst listening:
Computing power = smartness ? Hmmm.
Exponential human development – quartering time-base since homo-sapiens – OK. Predicts 2030-ish Omega point. If (that is IF) computers overtake human intelligence at that point, then by 2045 they will be billions of times more intelligent. Not sure computational power is the main driver of that cycle, but …
(Personally – I still believe in the three human generations timescale of technology development, ~80 year cycles in their actual effect on how humans live. Kondratiev / Kuhn etc. Interestingly 2030 is about the cusp of the next wave.)
Ray Kurzweil (now Eng Dir @ Google) – it’s about Language, the Turing Test. Zzzzz.
IBM “Watson” – “more” natural, sure. Anthropomorphic interactions (saying thanks), sure.
Humans do have advantages over robots – creativity, problem solving, emotional intelligence, dexterity, flexibility, etc. Doh! that’s what makes us humans as opposed to Accounting PhDs. Sure technology will automate the boring bits. Accounting however is a very complex human game – a million miles from arithmetic computation. Can’t see any robot playing “Tabletop” – the level-slipping analogies of creative problem solving.
Still (@24:10) only talking about “digital” computers and algorithms. Fast and accurate is not the point of human intelligence. Additive is the key – human intelligence will exploit ever more powerful tools.
(Interesting both presenters wrong in detecting human vs machine composed music. Second was clearly “formulaic” – metronomic – time-base-wise. Hard to prove now, proves nothing anyway, but I made the right choice. Not sure whether we were dealing with recordings or if both compositions machine-played?)
The Meta-aspects of humans, standing back to reflect on humanity. (Meta is the level-slipping in Tabletop.)
Ultimate conclusion is humanity – and human intelligence – is more than “machine” AI.
Agree. This is why I say AI can and will evolve, but it will only approach the intelligence of an intelligent life-form like humans, when it evolves to be an intelligent life-form itself. Real-I. (On evolutionary timescales.) Nothing special about humans here – any similarly intelligent life-form will do, we just happen to be the encumbents around here.
So long as AI workers have digital computers as their “computation” model of human brains (let alone minds) they’ll be severely hampered in getting AI to develop that way. Human intelligence does of course exploit “algorithms” to save time and brain power from the boring bits, free to do the creative “slipping”.