Tuesday, May 15, 2007

Half way to Vernor Vinge's Singularity

In 1993 Vernor Vinge wrote:
Within thirty years, we will have the technological means to create superhuman intelligence. Shortly after, the human era will be ended.
He didn't write this in a science fiction novel. This was not hyperbole. Vernor Vinge, a faculty member of the San Diego State University department of Mathematical Sciences is also a science fiction author. But his goal, in the non-fiction essay "The Coming Technological Singularity: How to Survive in the Post-Human Era," was to outline what he calls "The Singularity" -- defined as a course of events that would bring the human race to "...a point where our old models must be discarded and a new reality rules."

We are now almost to the halfway point of his 30 year prediction. With 15 years behind us and 15 years to go, are we still enroute to the Singularity?


It is easy to think, living in the early 21st century, that human life goes on much as it has in the past. As Ray Kurzweil details in his book on this subject, The Singularity is Near, human beings are ill-equipped to evaluate the pace of change that they are experiencing. Our perceptions of the now are mired in our own personal memories of the past. We judge our surroundings according to the relatively limited knowledge that we each contain. Furthermore we have no historical or personal experience that prepares us to comprehend the speed of change which is now occurring.

Just the changes that have occurred in the short 41 years I have been alive are staggering. As Vinge points out in his essay, the core technical innovation necessary to bring about the Singularity is computational power. In 1966 nothing on the planet existed that we would think of as a computer. In 1993, when Vinge wrote his essay, there were only a handful of computers. Last week I walked into the co-location facility where a portion of Technorati's server farm lives -- the room (one room on one floor of an immense building) throbbed with power -- heat and light came from every rack. There was more computational power in that one room than has previously existed in the history of mankind... multiply that room by the dozens of such rooms in that one building and then the thousands of such buildings around the planet and the size and scope of the transformation begins to come into focus.

In reviewing his own essay, in a set of thoughts 10 years after his original prediction, Vinge writes (in 2003) of his prediction of sufficient technical progress to bring about the Singularity within 30 years:
Now in 2003, I still think this time range statement is reasonable.
The Intel website details the hypothesis proposed by Gordon Moore commonly known as Moore's Law:
In 1965, Intel co-founder Gordon Moore saw the future. His prediction, now popularly known as Moore's Law, states that the number of transistors on a chip doubles about every two years. This observation about silicon integration, made a reality by Intel, the world's largest silicon supplier, has fueled the worldwide technology revolution.
Amazingly, Moore's 1965 vision has continued to hold true to this day. So following this graph out to the 2023 date in Vinge's Singularity prediction, we can expect that the 1 trillion transistors available in 2006 will double roughly 8 times to 128 trillion transistors. The computing power that will fit into that Technorati server room will be roughly 128 times more powerful than it is today in just 15 years.

In Ray Kurzweil's 2001 essay The Law of Accelerating Returns, he provides an analysis of the computational power necessary to represent a functioning brain of a variety of different species -- insects, mice, humans... and when that computational power is likely to be available (again according to Moore's law). The date by which he has proposed modeling a mouse brain is right about now. And is if on cue, just last month a team was about to run a "simulated mouse brain at 1/10 time." From the team's write-up:
We deployed the simulator on a 4096-processor BlueGene/L supercomputer with 256 MB per CPU. We were able to represent 8,000,000 neurons (80% excitatory) and 6,300 synapses per neuron in the 1 TB main memory of the system. Using a synthetic pattern of neuronal interconnections, at a 1 ms resolution and an average firing rate of 1 Hz, we were able to run 1s of model time in 10s of real time!
The scientific team working on this research noted that there were numerous problems that they encountered in trying to provide a realistic simulation of a mouse brain in this test. But the news bulletin for the rest of us is simple -- the supposedly radical suggestion that Vinge made way back in 1993 is now coming to pass. 15 years into his 30 year time horizon, the milestones are being achieved, on schedule.

What does this mean for all of us alive today who are still likely to be around in 15 years? Stay tuned for part 2 of this post...

2 comments:

NextGen Designer said...

So am I to believe that all of the genius, capability, and ideas of hard working hard thinking people who develop silicon chips and design all of the processes all comes down to Moore's law and that a self fulfilling prophecy will eventually bring instantaneous change? I am also to believe that progress is unlimited, transistors don't have a size limit and that the software for this is just suppose to fall out of the sky. For any of those people who design out there and come up with "real ideas" that fuel progress i am sorry for the above disrespect. The only thing truly inventive here is the fact that so many will believe such a large lie and the only way they can consciously do so is if they are avoiding a fact of reality. That fact of reality which you are evading is that progress is not to be taken for granted and neither are those that make progress possible. The designer, the inventor, the investor, the engineer, and the innovator are real. Their genius is the fuel of progress.

Thanaporn said...

On schedule? Kurzweil said the computational equivalent of a mouse brain would be available at $1,000. The million-plus dollar blue gene computer running at 1/10th speed is off by a factor of 10,000. Double your computational capabilities every year, and you're still of by 13 years. Exactly how is being off by 13 years over 15 years "on schedule"?