John Robb's Blog, page 10

Deep Maneuver (Autonomous Robotic Weapons)

Many cyber weapons are designed for deep maneuver.  These virtual weapons drift across the Internet, jumping from computer to computer to computer, potentially travelling for years until they find the target they were designed to destroy.  

Deep maneuver is also possible with autonomous robotic weapons in the real, physical world.  I'm not talking about the minimal performance improvements  achieved by removing the weight of a pilot or crew from a manned system.  Instead, I'm talking about autonomous robotic systems that can undertake missions that last for years and traverse tens of thousands of miles. 

Let's dig into this idea a bit.

The earliest example of robotic deep maneuver I've found is an operation from WW2 called Fu-Go. Fu-Go was the Japanese attempt to bomb the continental US using balloon bombs. Although Fu-Go was a complete failure, I find it useful as a way to think productively about how robotic intelligence can be used to surmount physical challenges (distance, time, etc.).

Where the Fu-Go balloons landed in the US

Here are some details about Fu-Go:

The operation began in late 1944 in November, as US B-29s began the bombing mainland Japan and a couple of months after the Germans began launching V2 rockets. Operation Fu-Go was commanded by Major General Sueyoshi Kusaba of the Imperial Japanese Army and carried out by 2,800 soldiers. These soldiers launched 9,300 balloons made with mulberry paper and held together with potato paste, by hand.


The balloons were built to carry hundreds of pounds of explosives across the Pacific in about three days, using the winter jet stream as propulsion. In order to access the jet stream the balloons were outfitted with systems (releasing ballast and venting gas) that kept them between 30,000 and 38,000 thousand feet. Some of the balloons were outfitted with radio transmitters, so their progress could be mapped by Japanese facilities on island bases across the Pacific.


The operation was a complete failure.  The level of robotic autonomy used by the Japanese wasn't advanced enough to overcome the challenges of the task.  For example, the Japanese predicted that 10% of the balloons would reach the US (about 900 balloons), but only 300 balloons made it.  On top of that, no major damage was done by the balloons that completed the journey.   

I find that this example provides me with some insight into how robotic weapons can make deep maneuvers like cyber weapons.   As we know, cyber weapons are already experts at using the environment for propulsion.  They use everything from open network connection to the stochastic motion of personal gadgets (cell phones, etc.) to maneuver themselves to their target.  

Autonomous robots can do the same in the physical world by substituting intelligence for mechanical performance.  This intelligence would allow them to leverage a wide variety of environmental factors to extend mission duration and range, from using wind/ocean currents to hitchhiking on vehicles (ships, trucks, aircraft, etc.) to slow self-propulsion using solar energy (or buoyancy).  Deep maneuver makes it possible to:

Traverse an ocean.  Hide in the muck of an opponent's harbor or in the coral reef near a disputed island. Engage kinetically with PGMs when required.
Infiltrate a remote region and set up a sensor network to monitor enemy activity and look for targets of opportunity.  Persist for a decade, permanently denying the area to opponents.  
Fly by night.  Hide and/or recharge by day.   Tap into the opponent's electrical grid or fuel systems.  Do so until target is found/neutralized.  

Have fun

John Robb

I spent last year working for the Chairman of the Joint Chiefs of Staff on a vision for how advanced robots will transform warfare over the next twenty years.  This year I'll share my thinking with you .  Tag along if you are interested.

Zero Day Warfare

The winner of the next big conflict will be the side with the best understanding of how to use bots in warfare. Bots aren't just an iterative improvement in warfare, like stealth or PGMs, it's a revolution in the making. The US military, to its credit, is working on this. So far, the US military has identified three (out of nearly a dozen) of the foundational ideas needed to successfully employ bots in warfare:

Teaming
Autonomy
Swarming

Learning from Nitro Zeus

However, these early ideas are a long way from the operational thinking required to win wars using bots. That type of thinking requires a synthesis of the foundational ideas into new operational concepts. Here's a good example operational concept I'm calling zero day warfare. It builds off the thinking already demonstrated in recent US cyber operations:

The US recently leaked plans for Nitro Zeus, a sweeping cyber attack on Iran to be used only if the nuclear negotiations with the country broke down.


Nitro Zeus, building on the earlier success of the Stuxnet/"Olympic Games" (the earlier cyber attack that set back Iranian nuclear activities by destroying 1,000 centrifuges), was designed to seize control or knockout Iran's air defense system, communications grid, transportation system, and energy grid on the first day of the conflict.


The rapid onset of chaos caused by Nitro Zeus would have then made it possible for immediate kinetic attacks on the real objective of the operation: the Iranian nuclear facility at Fordo.   

Zero Day Warfare

The goal of zero day warfare is to win the war before it starts (a very zen concept) by deeply penetrating the opponent's territory years before the conflict begins. Like all maneuver warfare, it is focused on shattering the opponent's physical and logical cohesiveness. Here's a quick summary of the highlights:

Autonomous robots and software bots (collectively "bots") deeply penetrate the opponent's territory both physically (territory) and logically (their computer systems). Most would be hidden and remain dormant until activation. Some would actively or passively map opponent networks, analyze them for vulnerabilities, and take advantage of opportunities for stealthy exploitation.


When activated, these forward bots conduct a coordinated attack from inside the opponent's territory and systems. Damaging, degrading, or taking control of computer systems and physical infrastructure. Advanced robots would emerge from stealth to kinetically engage with opponent forces or physically seize points (airports, ports, etc.) to enable the rapid entry of conventional forces. 


External forces, both bots and conventional, would utilize the disruption of the Zero Day attack to rapidly enter the territory and seize control of key facilities and capture remaining leadership.  

Have fun,

John Robb

PS: A zero day warfare that includes deeply deployed autonomous robots will be possible within the next decade. Almost all of the tech needed to pull it off is almost here.

____________

I spent the last year working on the Chairman's (of the JCS) vision for how advanced robots will change the face of conventional and unconventional warfare. This year I'll share my thinking with you.  Tag along if you are interested.

Why Apple is on the side of National Security and not the FBI

The FBI wants access to the encrypted data on an iphone owned by one of the the San Bernardino terrorists.  The FBI has already gotten access to the data this iphone uploaded to Apple's cloud. However, the FBI thinks there is more data on the phone that wasn't uploaded. Here's a quick recap:

The FBI can't access information on the iphone because the brute force technique the agency uses to "guess" user passwords doesn't work with the iphone. Brute force password guessing would cause the phone to permanently block access to the data after 10 attempts if that option is turned on (and they think it is). Even with the option off, it could take up 5.5 years to crack the password, because the iphone inserts a 10 ms delay between password attempts.  


The FBI went to court to force Apple to provide them with a way to turn off the features that prevent them from using their simple password guessing technique in the future.


Apple refused, because it is currently engaged in a struggle with the Chinese government over the same issue. Apple knows that if it complies with the FBI on this, it will become the rope in a violent tug of war between the two governments over who gets access to encrypted iphone data. Since 25% of Apple's sales and nearly all of its growth is coming from China, it's clear they would like to avoid making a choice between the US and China.

Of course, Apple's worry doesn't make it right in this fight.  I couldn't care less if they make money or not.  They are right because, with the threats we will soon face, backdoors like this are no longer good for national security. They harm it.

Backdoors that allow access to encrypted data, like the backdoor the FBI is demanding, can be useful in fighting blood and guts terrorism. It makes it easier to unwind terrorist cells and financing (usually after the attack though).


However, blood and guts terrorism isn't the threat it once was. It is the last war. The growing threat will be from nations and groups armed with cognitive bots.  Bots capable of accomplishing a great number of human scale tasks very, very quickly.  In this new type of war, anything that slows or delays an attacker is worth preserving.  In this case, the difficulty of breaking the encryption on the widely used iphone is a good thing.


In wars where bots do most of the fighting, all backdoors become a zero day vulnerabilities -- unpatched vulnerabilities that allow attackers rapid, unopposed access to the system.  Inserting vulnerabilities into the iphone would almost certainly make it useful as a major avenue of attack by the bot armed threats we'll face in the near future. 

Let's not prepare for the last war by making ourselves vulnerable to the next one.  

Why Trump Wins

"The fault line in American politics is no longer Republican vs. Democrat nor conservative vs. liberal but establishment vs. anti-establishment. This is an inevitable result of serial failure in establishment policies." Failure as a Way of Life  Bill Lind

Donald Trump is crushing the establishment candidates of the Republican party.  Sanders is doing the same with the Dems.  With that in mind, here are a few insights into Trump that I found useful:

Donald Trump isn't a fool.  Here's a recent interview that shows Donald Trump is much more interesting than the performance art of his stump and debate.  BTW:  he's a non-interventionist when it comes to foreign policy.


"From the start, Trump targeted the (mostly) white working class, which happens to be 40 percent of the country. And he’s done it not just with issues, but with how he talks — the ball-busting, the “bragging,” the over-the-top promises...

But it speaks volumes — whole encyclopedias — about the ignorance of our political and media elites that they’re only now realizing that much of what Trump’s been doing is just busting balls.  It’s a blue-collar ritual, with clear rules — overtly insulting, sure, but with infinite subtleties. It can be a test of manliness, a sign of respect, a way of bonding and much more.

  America hasn’t been great for the working class for decades — which is why “Make America Great Again” is a great slogan for a guy who’s talking tough on the problems that blue-collar Americans (and more than a few middle-class folks) see as killing them."  And it is killing them:  "The rise in mortality from 1999 to 2014 was 22 percent: Up 134 deaths per 100,000 for whites aged 45 to 54 whose education ended in high school....  (due to) jumps in suicides and in deaths from drug abuse." Donald Trump has Invented a New Way to Win Mark Cunningham


Nils Parker:  "He is the first true social media/internet candidate. He is not only harnessing its tools like no one before, but he is both speaking to and responding from the culture social media has created. It is social media more than any other influence that has created the cowed culture that the 40% white male demo he is going after feels alienated by and furious at.  

The trolling, the bombast, the bragging, the conspiracies--it's quintessential Internet.  also he understands the temporariness with which all normal people treat social media engagement on Twitter, Snapchat, Instagram. It's a bottomless scroll that effectively disappears since no sane person goes scrolling back more than a week's time in order to find something to get pissed off or offended about. So he pops off and then lets it go--which his supporters fundamentally understand.   When opponents and media pull up a tweet from 2012, literally no one gives two shits and it only ends up reflecting (poorly) on the person pulling up the tweet."

 

A Robotic 9/11

Last year I was hired to craft a long term vision for the Chairman of the Joint Chiefs of Staff on how autonomous robots should be used to radically improve how the US military fights wars.  

This year I will be sharing my thinking with you.

_____________

The current revolution in robotics is due to rapid advances in the ability of robots to think.  

This means that most of the big improvements we'll see in the use of autonomous robots in warfare will be due to finding new uses of this attribute more than any other.  Let's explore this a bit.  

It's now possible to turn a simple low performance drone into a weapon that is nearly as effective as a precision guided missile (PGM) that costs hundreds of thousands of dollars.  This is accomplished through the creative substitute inexpensive and sophisticated machine thinking for expensive mechanical performance.  

In other words, the smarter the drone is, the better it can mimic the performance of the much more expensive PGM.

This is already possible today with inexpensive, commercially available drones.  Low cost drones are now smart enough to approximate the performance of an expensive surface to surface missile system with a little creativity.   Let's dive into this a bit. 

From a mechanical perspective, consumer drones aren't that impressive:

~1-2 pound payload
~20 min flight time 
20-40 miles per hour flight speed

However, these drones are already very smart:

They can fly themselves.  They can take-off, fly enroute, and land autonomously.
They can precisely navigate a course based on the GPS waypoints you designate.
They can now (a recent development) use digital cameras to find, track, and follow objects.  Some can even land on objects they find based on a description of that object.  

Even this basic capability is more than enough to turn a basic drone into an extremely dangerous first strike weapon. Here's a scenario that pits ten drones against a major airport:  

Ten drones would take off autonomously in 1 minute intervals.
Each would follow a GPS flightpath to a preselected portion of an airport.
Upon arrival, a digital camera would identify the nearest wing of an aircraft.
The drone would autonomous land in the middle of that wing.
A pound of thermite in the payload would ignite upon landing.  
The thermite would burn through the wing, igniting the fuel inside...
Most of the airport and nearly all of the planes on the tarmac are destroyed.

Here are the takeaways:  

Even the simple robotic platforms of today can be extremely effective as weapons.  At current rates of improvement in machine intelligence, the situation will get much more interesting very, very soon. 
It's possible to creatively trade inexpensive machine smarts for expensive mechanical performance.  
We need to figure this out before the bad guys do.  However, figuring this out requires a deep insight into the dynamics driving this forward.  

 

What I've Been Doing: Working for the CJCS on the future of Warfare

I spent 2015 thinking and writing about the future of warfare while working for the Chairman of the Joint Chiefs of Staff.  Specifically, I was tasked with figuring out how autonomous robotics could/would change warfare over the next twenty years, and how the US military could exploit that change.  

Up until this effort, the US military didn't have a cohesive vision for how it would use autonomous robotics.  Now, it does (at least the first iteration of one).  The cool part, for me, is that I now have a very strong grasp for how this tech is going to roll out and how to exploit it to achieve decisive results both tactically and strategically.  

Needless to say, the change on the way due to autonomous robotics is going to be very, very disruptive.  So, absent another contract that would pay me to think/write for them, I'm going to write some of my thinking up for GG and put the rest into a book so the world can put it to use.

PS:  I'm interviewed on Episode 236 of the Robot Overlordz radio/podcast.   A general talk.

 

Germany Just Screwed Europe

Germany needs young people.  Its population is dwindling.  It doesn't have enough of them to support a rapidly aging German population.

A smart solution to this shortfall would have been to actively recruit millions of young people from around the world -- from India to Tanzania to Argentina to Malaysia to the US.  
Instead, for some reason, the people running Germany thought it would be a good idea to use the Syrian crisis as a way to bring millions of young people to their country. 
It worked.  Germany imported 1.1 m migrants last year.  

However, if you dive into the details, it's clear this policy is very dangerous.

Almost all (~800,000) of these migrants were young, single men (Canada, in contrast, refuses entry to young single men) from Syria.  Insurgencies and terrorism run on a fuel of young men.
These young men are culturally incompatible with EU/US standards re: women, homesexuality, free speech, etc.  With a group this large and cohesive, cultural integration will be nearly impossible over any meaningful time period.
Most of these young men (~500,000-700,000) have recent combat experience earned on the killing fields of a fractured Syria, fighting for a variety of bad causes.  This makes them potentially dangerous.

Here's how I believe this will play out:

Social disruption will rise.  We are already seeing this with recent attacks by roving gangs of immigrants in Cologne.  Further, Schengen will disintegrate as transborder attacks by radicals ramp up.  We saw this with an attack by a "German" migrant on a police station in Paris and the flow of weapons/attackers from Brussels during the Paris attacks last year.  
Over the medium term?  Terrorist violence.  This population, and those that soon follow, will soon become the main conduit for extremism in Europe.  Its large size, antagonism and cohesiveness will make it impossible to police.
Over the long term?   As the demographics of these countries rapidly shift in favor of the new arrivals: open source insurgency.  An insurgency that will spread far from the borders of Germany.  An insurgency I'm not sure Germany, nor the EU, can win.

PS:  What a missed opportunity for Germany and the EU.  Instead of recruiting millions of economically and socially beneficial migrants from countries across the world...

Culture is Everything

What is culture? In the broad sense, it's a way of life.  More specifically, it's a basket of shared behaviors that determine how we solve problems, define success, and treat each other.

Culture is important. It has been proven critical to socioeconomic success, at every level, from the extremely large group to the individual (although at the individual scale, we call it character). For example, in the corporate world, most successful CEOs will tell you the same thing: culture is everything.

So, if it's so important, why don't we talk about culture more?

It isn't easy to quantify. It's not easy for bureaucrats to dictate or markets to measure.

Fortunately, there is a way to understand it a bit better. Culture is important because it plays a critical role in personal and group decision making. More specifically, it drives the "orientation" step of John Boyd's decision making model, the OODA (observe, orient, decide, act) loop.

Orientation is different than the other steps in decision making.  It's a gut check.  A check of core values.  It is a synthesis of everything you've learned as it applies to the problem you face.  This makes it squishy and holistic.  It's the step that Einstein so elegantly referred to in this quote:

If I had only one hour to save the world, I would spend fifty-five minutes defining the problem, and only five minutes finding the solution.

Unlike orientation, the other steps (observe, decide, act) used in decision making are largely mechanistic, analytic and quantitative. To improve these other steps, you speed them up (i.e. computers), increase their fidelity (accuracy without error), and widen their scope (more data).

In contrast, culture is how human beings have learned to speed up orientation in a dependable way.

Culture can provide any individual, organization, or country with the outlook needed to successfully orient problems repeatedly and without hesitation.

Here's an example:

Some business cultures place a high value on treating the counterparty in a transaction with respect and dignity.  In those cultures, it's important that every business transaction is a win-win, where both sides are better off for doing business together, regardless of the contractual details.  

In other business cultures, business transactions are highly competitive.  In those cultures, it's important to win every business transaction and contractual details are used as a weapon to bludgeon the counterparty into submission.

See the difference in approach due to culture?

What should also be obvious from this example is that cultures differ.  They can be wildly different.

They aren't equally effective, the usually don't mix well, and some can be toxic.

Sincerely

John Robb

PS: John Boyd developed the OODA loop to figure out how to win conflicts.  He postulated, correctly, that successful decision making is the most important factor in survival -- from simple organisms surviving evolutionary pressures in primordial pools to winning wars on modern battlefields.   Therefore, success in any conflict was largely due to faster, better decision making.  The faster you can make good decisions, the faster you can iterate to success.

The Changing Face of Illiteracy

Estimate:  2030.  About 15 years for this to become a truism.  Amazingly fast given the magnitude of the change.    

In the 19th and 20th century, illiteracy was seen as an inability to read or write.  Illiteracy damaged your economic prospects and your ability to function in society as a whole.  

With the arrival of the Internet (and the world of knowledge it connects you to), being illiterate is increasingly being seen as an inability to learn on your own.  

Learning is now a continuous lifelong activity and with the Internet available, it's largely accomplished using a self-service approach.  When you want to find something out you can usually find it on the Internet, or if you need to learn a new skill, you can usually find a course on the Internet to learn it and experts to help you out.

With the arrival of cognitive machines (using "deep learning" and other model free methods), our definition of illiteracy will change again.  Based on what I see, illiteracy will increasingly be seen as an inability to teach, coach, and train machines that learn.  

From teaching a car to drive itself to training to evaluate an insurance claim to coaching a bot to spot a person with a concealed weapon, you will find yourself in charge of helping bots learn about the world.  

AI for the Masses

All of the big tech companies are now open sourcing their AI (deep learning) software.  They are also open sourcing the hardware designs of the machines needed to train the software.

That's a big deal.  This new approach to AI is proving so powerful and useful, its use is growing exponentially (see image below).

At this rate of adoption and with the barriers to participation dropping daily, I'm confident this technological revolution will upend the world in less than a decade.  Most of the innovation we will see will occur at the grassroots level.

Here's an example of what is possible when a small company open sources a relatively "dumb" machine cognition software.  This software is called "openaplr" from a company of the same name, and it's available for download on Github.  

What does it do? It can turn any IP enabled camera into a license plate reader capable of reading 60 license plates a second.  

Very cool and very powerful and completely legal.

With this amount of capability it wouldn't be hard to crowdsource an effort to track the movement of nearly every car in the country.

A next generation deep learning program (like the one they are using at Facebook) would allow you to identify and track people, rather than cars, using a small body camera or a smartphone to do it.

Open source little brother anyone?

Have fun,

John Robb