Why smart machines are a threat to the way humanity currently exists
September 5, 2017 at 2:36 am
(This post was last modified: September 5, 2017 at 3:26 am by ThoughtCurvature.)
A user here recently directed me to the article: "Smart Machines Are Not a Threat to Humanity".
Of course, like many other articles of the same form, the article fails to consider the following:
(1) The functioning machine intelligence at the onset of the industrial revolution, are very different models than the brain inspired beasts that function well today. The old models didn't function well, while "learning" to do tasks as humans do.
By 2020, we will likely see conventional machines with human level brain power. (See Wikipedia "Exascale Computing")
(2) In the distant past, the top three firms in some region of the world earned billions, and hired millions.
Recently, the top 3 in a similar region earned trillions, and hired thousands.
(See article "Fourth Industrial Revolution brings promise and peril for humanity")
FOOTNOTE:
In other words, the initial trend (where more jobs naturally arose, as more machines came) has already changed, and changes are occurring faster still, opposite to the initial trend.
EDIT:
Here is a useful edit for everybody to see, which answers user Mathilda's first response to this thread:
Unsupervised learning models, which are things needed for Artificial General Intelligence (or significantly smarter ai) already exist, and get better and better, as time goes by.
Here are some examples:
(1) Manifold learning or Deepmind's "Early Visual Concept Learning with Unsupervised Deep Learning"?
(2) Generative Adversarial Networks that uses unsupervised learning? (See Wikipedia "Generative_adversarial_networks")
Of course, like many other articles of the same form, the article fails to consider the following:
(1) The functioning machine intelligence at the onset of the industrial revolution, are very different models than the brain inspired beasts that function well today. The old models didn't function well, while "learning" to do tasks as humans do.
By 2020, we will likely see conventional machines with human level brain power. (See Wikipedia "Exascale Computing")
(2) In the distant past, the top three firms in some region of the world earned billions, and hired millions.
Recently, the top 3 in a similar region earned trillions, and hired thousands.
(See article "Fourth Industrial Revolution brings promise and peril for humanity")
FOOTNOTE:
In other words, the initial trend (where more jobs naturally arose, as more machines came) has already changed, and changes are occurring faster still, opposite to the initial trend.
EDIT:
Here is a useful edit for everybody to see, which answers user Mathilda's first response to this thread:
Unsupervised learning models, which are things needed for Artificial General Intelligence (or significantly smarter ai) already exist, and get better and better, as time goes by.
Here are some examples:
(1) Manifold learning or Deepmind's "Early Visual Concept Learning with Unsupervised Deep Learning"?
(2) Generative Adversarial Networks that uses unsupervised learning? (See Wikipedia "Generative_adversarial_networks")
