When Machines Outsmart Humans
By: Nick Bostrom
Within this article Nick Bostrom states that half a century after the first electric computer, we still have nothing that resembles an intelligent machine (one that possess the kind of general-pur-pose smartness that we humans have). For multiple reason this process has been delayed such as the complex human brain so It is therefore important to consider the possibility, that intelligent machines will not be built within fifty years.
This article will outline the reasons for this delay and the consequences of human-level artificial intelligence. One limiting factor is speed rather than memory; the human brain typical estimated range from 100 million MIPS (Million Instructions Per Second) to 100 billion MIPS, and the most powerful supercomputer to date performs at about 10 million MIPS. So it is easy to see that in the sixties, it was just not possible due to hardware limitations.
But what is hardware without software, which is the next leading problem. One way to build the requisite software is to figure out how the human brain works and copy natural solution. This will require continuing rapid progress in neuroscience, for simulating a whole brain will take an enormous amount of computing power, which such a capacity should be available in a couple of decades.
Even if we succeed in achieving artificial intelligence, we should consider the ramifications of the following:
- Artificial minds can be easily copied since they are software and therefore quickly come to exist in great numbers
- Human-level artificial intelligence leads quickly to greater-than-human-level artificial intelligence, which could cause machines that think more than hundred times more rapid than humans do,
- Technological progress in other fields will be accelerated by the arrival of artificial intelligence, by means of technological research that will be done more effectively by machines than humans. This could cause machine intelligence to devote their abilities to designing the next generation of machine intelligence, which could all lead into the “singularity”.
- Unlike other technologies, artificial intelligence are not merely tools. They are potentially independent agents, which could cause these machines to be technically “slaves” meaning there is an ethical and political debates that could occur.
In conclusion, two themes can be drawn from all of this. First that there is currently no warrant for dismissing the possibility that machines with greater-than-human intelligence will be built in fifty yearsand second that the creation of such artificial intellect will have wide ranging consequences for all the social, political, economic, commercial, technological, scientific, and environmental issues that humanity will confront in this century.
I find that is article purposes many concerns we should have for artificial intelligence, but all seem to be inspired for Sci-Fi movies, books, and media. I feel that if we did come to the stage of having successfully developed artificial intelligence these issues would have addressed before any ethical matter became of it. This article is dated in 2003, and almost another decade later, we are still struggling with the progress on this front. Many say that the date is even sooner, than 2050, but it is hard to say if true human intelligence can be understood, let alone copied.