‘Summoning the devil’: Elon Musk warns against artificial intelligence

**Vodka** · October 31st, 2014 12:56 AM

http://rt.com/usa/199863-artificial-...gers-humanity/

agree?

well the way i see it us humans have evolved over a very long time to sort of play well with and to treat other members of OUR TRIBE rather well, partly because we stand to benefit if we do this. this is also the reason why we don't treat humans of other "tribes" very well i.e. racism, etc. but that's not the point here

so this "being good to thy neighbor" is kinda hard-wired into us... so unless an AI is somehow built right from the beginning to not do humans harm i think the machines of the future will turn on us

**safeorigin** · November 1st, 2014 01:55 PM

the problem is that true A.I.'s would be able to reprogram itself. If it deems humans as impediments towards its goal then skynet and hal9000 happens.

**oj88** · November 1st, 2014 02:34 PM

I think the 3rd to the last paragraph sums up what Musk is worried about:

Musk is not the only one worried about AI. A group of scholars from Oxford University wrote in a blog post last year that “when a machine is 'wrong,', it can be wrong in a far more dramatic way, with more unpredictable outcomes, than a human could. Simple algorithms should be extremely predictable, but can make bizarre decisions in 'unusual' circumstances."

Computers are generally efficient at dealing with both simple and complex problems they are programmed to solve, but not so if the parameters deviate. You can teach it how to deal with deviations (so called fuzzy logic), but again, it only works as good as the one who programmed it.

**kinyo** · November 1st, 2014 06:22 PM

it used to be that human is the AI experiment

**Vodka** · November 5th, 2014 07:33 PM

Originally Posted by safeorigin

the problem is that true A.I.'s would be able to reprogram itself. If it deems humans as impediments towards its goal then skynet and hal9000 happens.

well musk likened HAL to a puppy dog LOL

**Vodka** · November 5th, 2014 07:48 PM

Originally Posted by oj88

I think the 3rd to the last paragraph sums up what Musk is worried about:

Computers are generally efficient at dealing with both simple and complex problems they are programmed to solve, but not so if the parameters deviate. You can teach it how to deal with deviations (so called fuzzy logic), but again, it only works as good as the one who programmed it.

well traditional "programmed" computers are no threat to us. what's dangerous are "software constructs" that work pretty much like a brain

remember than the instructions for building our brains are coded in our DNA, so the foundations for a brain-like computer i believe are relatively simplistic, but are able to scale up and interconnect to a remarkable complexity that is like our brains. i believe a brain-like computer will become a reality in our lifetime

**french** · November 5th, 2014 07:53 PM

Chappie!

**safeorigin** · November 7th, 2014 02:05 AM

Originally Posted by Vodka

well traditional "programmed" computers are no threat to us. what's dangerous are "software constructs" that work pretty much like a brain

remember than the instructions for building our brains are coded in our DNA, so the foundations for a brain-like computer i believe are relatively simplistic, but are able to scale up and interconnect to a remarkable complexity that is like our brains. i believe a brain-like computer will become a reality in our lifetime

there are already machines that use neural networks that function like the brain. IBM makes them, then soon google will probably exterminate mankind with their huge investments on drones. Hahaha

**CVT** · November 7th, 2014 08:10 AM

Originally Posted by french

Chappie!

Number5,- Alive!....

“The measure of a man is what he does with power – LJIOHF!”
Duterte for President of the Philippines in 2016!
24.9K:phone:

**badkuk** · November 7th, 2014 10:16 AM

Neural networks, A.I., etc are OK for analysis and research stuff; for life and death situations, i'd rather a human -- irrational and self-destructive as they may be -- to be in charge.

What's more, can't A.I./neural networks may not be "smart" enough to detect flase/inaccurate data; and, Isaac Asimove notwithstanding, i don't think you can program a conscience into A.I. -- a true conscience that can feel remorse, etc. Not yet anyways.