With Scale Intuition Fails
Hi! Happy Monday!
With Scale Intuition Fails
I can't be right on both accounts, can I? How can I recommend both a rapid pace to innovate and slow speed ahead? I must be wrong.
If you believe I'm wrong then you would be wrong. That's because your intuition fails you. With scale intuition fails: latency is not throughput.
Let us consider scale as it is taught in computer science. Yes, we are going to apply principles of computer science to political science. We are going to learn to scale. I'm teaching you.
In computer science when network bandwidth is considered one has to consider the trip time, or latency, of the first packet and then again the rate of packets that can be sent. The first network packet of a request may take an entire second to reach google.com, for example, but all the subsequent packets are right behind the first packet being sent. The same applies on the response. The trip time of the response may take one entire second to return but the subsequent packets will be right behind the first.
For this exercise we are going to simplify network bandwidth to a unit called a network packet. The bandwidth will be in units of packets-per-second (pps).
In my Google search example above I supposed a two-second round trip time: one second for the first packet to arrive and another second for the first return packet. That is a total latency time of two seconds.
Now, if I send one packet and receive one packet then my bandwidth is very poor at 1pps. Assuming Google can produce one-thousand packets in one-tenth of a second and further assuming the Google search response was one-thousand packets then my bandwidth is now 1000p/2.1s or approximately 500pps. Notice how 500pps is approaching the rate of 1000pps, which is the rate at which Google is assumed to be able to produce packets.
In my post entitled "Slow Speed Ahead" I'm demanding we take the time to do experiments and get things correct. That is recommending a slow latency. In fact, I'm thinking that the latency will be on the order of one-hundred years. It is going to take a century of building social scale infrastructure before we start seeing results. Slow speed ahead if we want to do this right. We need to treat social experiments as science and quantify the hell out of them.
In my post entitled "Sixty-Three Quad Zero" I was not suggesting the number of experiments that be undertaken, but I am now. I am suggesting that we initiate millions of experiments: right here, right now. That's called innovation. Innovation requires tons of experiments that through a combination of science along with trial-and-error produces winners.
Over the course of one-hundred years if the human race embarks on one-million social experiments and nine out of ten of those fail, the same as the Internet start-up failure rate, then there will be one-hundred thousand successful experiments over the course of one-hundred years. That's a throughput of one-thousand social program innovations per year. That rate of innovation is not possible to scale with politics as we have today. Not even close.
We need to start thinking in scale.
For example, rushing to put body cameras on policemen is bound to fail because we are not experimenting over many years socially. We need the time for the feedback of social factors to kick in and influence the process. We need to innovate.
We also need to consider alternate ideas. For example, perhaps an alternate idea to cops-with-cameras is camera operators. For all 630,000 cops on the street today then we hire 630,000 operators to do nothing except what citizens are doing today: take video of incidents. This will create 630,000 jobs overnight. Big plus. Second, cops don't have to be under surveillance with body cameras when they go to the bathroom, they make a private call, etc. If you are not disturbed by the notion of mandatory 24/7 surveillance then let us mandate that for you. Now.
We need to innovate millions of social experiments, but do so with science. The science of experimenting with social programs needs to be voluminous, ergo the Data Center on the Moon where all experiments feed results into a database owned by the public for the public good. Part of the experiment of running millions of simultaneous experiments is learning how to experiment from experience.
If we can produce a rate of one-thousand successfully innovative social changes-per-year over the next century then we will have left behind a better world for our children. After all, a generation is no longer twenty years when we are living upwards of one-hundred years.
Cheers!
Be sexy people!
The future is coming!
Innovate at a rapid pace!
Slow speed ahead!
Well come! and well met!
With Scale Intuition Fails
"Technology has put pressure on us to move beyond the very slow pace of evolving our social systems and to a rapid pace to innovate."
-Mybrid Wonderful, Sixy-Three Quad-Zero
"Yes I'm recommending a lot of slow, hard, long work with lots of candles and moaning."
-Mybrid Wonderful, Slow Speed Ahead
I can't be right on both accounts, can I? How can I recommend both a rapid pace to innovate and slow speed ahead? I must be wrong.
If you believe I'm wrong then you would be wrong. That's because your intuition fails you. With scale intuition fails: latency is not throughput.
Let us consider scale as it is taught in computer science. Yes, we are going to apply principles of computer science to political science. We are going to learn to scale. I'm teaching you.
In computer science when network bandwidth is considered one has to consider the trip time, or latency, of the first packet and then again the rate of packets that can be sent. The first network packet of a request may take an entire second to reach google.com, for example, but all the subsequent packets are right behind the first packet being sent. The same applies on the response. The trip time of the response may take one entire second to return but the subsequent packets will be right behind the first.
For this exercise we are going to simplify network bandwidth to a unit called a network packet. The bandwidth will be in units of packets-per-second (pps).
In my Google search example above I supposed a two-second round trip time: one second for the first packet to arrive and another second for the first return packet. That is a total latency time of two seconds.
Now, if I send one packet and receive one packet then my bandwidth is very poor at 1pps. Assuming Google can produce one-thousand packets in one-tenth of a second and further assuming the Google search response was one-thousand packets then my bandwidth is now 1000p/2.1s or approximately 500pps. Notice how 500pps is approaching the rate of 1000pps, which is the rate at which Google is assumed to be able to produce packets.
In my post entitled "Slow Speed Ahead" I'm demanding we take the time to do experiments and get things correct. That is recommending a slow latency. In fact, I'm thinking that the latency will be on the order of one-hundred years. It is going to take a century of building social scale infrastructure before we start seeing results. Slow speed ahead if we want to do this right. We need to treat social experiments as science and quantify the hell out of them.
In my post entitled "Sixty-Three Quad Zero" I was not suggesting the number of experiments that be undertaken, but I am now. I am suggesting that we initiate millions of experiments: right here, right now. That's called innovation. Innovation requires tons of experiments that through a combination of science along with trial-and-error produces winners.
Over the course of one-hundred years if the human race embarks on one-million social experiments and nine out of ten of those fail, the same as the Internet start-up failure rate, then there will be one-hundred thousand successful experiments over the course of one-hundred years. That's a throughput of one-thousand social program innovations per year. That rate of innovation is not possible to scale with politics as we have today. Not even close.
We need to start thinking in scale.
For example, rushing to put body cameras on policemen is bound to fail because we are not experimenting over many years socially. We need the time for the feedback of social factors to kick in and influence the process. We need to innovate.
We also need to consider alternate ideas. For example, perhaps an alternate idea to cops-with-cameras is camera operators. For all 630,000 cops on the street today then we hire 630,000 operators to do nothing except what citizens are doing today: take video of incidents. This will create 630,000 jobs overnight. Big plus. Second, cops don't have to be under surveillance with body cameras when they go to the bathroom, they make a private call, etc. If you are not disturbed by the notion of mandatory 24/7 surveillance then let us mandate that for you. Now.
We need to innovate millions of social experiments, but do so with science. The science of experimenting with social programs needs to be voluminous, ergo the Data Center on the Moon where all experiments feed results into a database owned by the public for the public good. Part of the experiment of running millions of simultaneous experiments is learning how to experiment from experience.
If we can produce a rate of one-thousand successfully innovative social changes-per-year over the next century then we will have left behind a better world for our children. After all, a generation is no longer twenty years when we are living upwards of one-hundred years.
Cheers!
Be sexy people!
The future is coming!
Innovate at a rapid pace!
Slow speed ahead!
Well come! and well met!
Comments
Post a Comment