Global Guerrillas

The Revolution in Robotics

There are two parts to the revolution in robotics we are seeing.   

The first revolution in robotics is based on tech trends already underway.  

The inflection point for this revolution occurred in 2001, when the standard computer chip exceeded the intelligence of an insect. 

With chips like this, robots quickly became inexpensive, accessible, and powerful.  For example, an autopilot system that cost ten thousands of dollars a decade ago is now available for $30 and can run on a small amount of power.

Here's an example of what is possible with this revolution:

While this revolution is fairly dramatic, don't expect too much.  

The capabilities of these robots won't advance any faster than the ability of human beings to write code, design and build hardware, and build successful businesses to support than activity.  

This reliance on human design and development also means that progress in navigating, interacting with, and making sense of complex, dynamic environments will be slow and hardwon.   

The second revolution in robotics is different.  It will be much more dramatic in its impact.  It is based on exponential improvements in machine learning.

These advances make it possible for machines to learn behaviors that make it possible for robots accomplish tasks that only humans can do today -- like driving cars safely in urban traffic to providing physical assistance and medical support to homebound elderly.    

Further, these cognitive machines will learn in days what it takes human developers months to accomplish (if they can do it all).  

In contrast to the previous revolution, this one will be amazing and traumatic at the same time.   

For example, this revolution in synthetic cognition has the potential to remake the modern economy as completely as industrial machinery and computation changed the agrarian economy of the 1700's.  

I've spent the last year thinking about how this machine learning revolution will change the way we fight wars and provide security.  

Happy Thanksgiving!

John Robb

Supersoldiers and Autonomous Weapons

Here's some thinking on how warfare will change over the next twenty years.   

Fast forward 20 years (about the age of the WWW).  An aging, schlerotic EU has become the destination for over a hundred million refugees and migrants fleeing the densely populated killing fields of Africa and SW Asia.  

The rapidity of influx has led the EU to take extreme measures.   Tens of millions of these migrants/refugees are roughly housed in relocation camps all across Europe.  

Violence within these camps has risen steadily, leading to an EU-wide Islamic insurgency.

The soldiers sent to counter this insurgency are outfitted with autonomous weapons.  These weapons combine deep learning (making them very smart) and cloud robotics (allowing the military to rapidly share advances in training and technique) to provide these soldiers with capabilities far beyond what we've seen in previous wars.    

Here's an idealized example so you can get the idea.  A human/robot team advances down a street in an urban environment.  

• Big Data:  The autonomous weapons used by the team continuously scans the street in all directions.  These weapons can visually ID everyone on the street from a database of 3.5 billion people in under a second.  It also continuously analyzes the people, windows, etc. down the street looking for the visual signatures of concealed weapons and IEDs.  i.e. A car at the end of the street is resting a bit too heavily on its springs, indicating there may be explosives in it.  These weapons learned to do this based on billions of hours of combat and police training images/footage (aka Big Data).  
  
  
• Customized Training:  The human members of the team have trained the weapons to alert the team when it sees any electric vehicles demonstrating even the slightest bit of irregular behavior -- the rapid acceleration possible with autonomously driven electric vehicles can make them dangerous kinetic threats in three seconds.
  
  
• Cloud training:  The autonomous weapons with the soldiers with connections to military's cloud.  Fortunately, this connection to the cloud gave these weapons access to the certified methodologies for identifying and neutralizing a new DIED (drone IED) used by Islamic insurgents only yesterday.  This paid off.  The new DIED entered the street behind the team, and the systems new how to ID it, engage it, and neutralize its countermeasures flawlessly.  During the engagement, the human team member noticed a slight change in the behavior of the DIED -- it released its homemade cluster bomblets earlier than anticipated.  The data/footage of the engagement is tagged with a note to this effect and it is uploaded to the cloud in order to add to the approved methods for countering it.  

Of course, much of this capability might become open source and available to anyone smart enough to employ it.

John Robb

The Gulf's Reality Distortion Field

Scott Atran:  I was just with the Saudi royal family and they were telling me that 700 Saudis have just joined ISIS in Libya.

Is ISIS the real source of Islamic terrorism?

No.  ISIS is just a symptom.  

The real source of Islamic terrorism is found in the the Gulf Kingdoms.  

To stay in power, the ruling families in these Kingdoms have turned their countries into feudal theme parks replete with aristocratic governance, social backwardness, and draconian punishments.  

The "reality distortion field" required to accomplish this perversion is maintained by the constant counter pressure of extreme fundamentalism.  A pressure that has steadily increased due to decades of torrential global change.

However, this fundamentalism isn't costless.  

Fundamentalism, as a force, is never satisfied.  It will strive to constantly turn back the clock to an ever more idealistic (and unreal) golden age.  

To avoid bloody fundamentalist revolts, the Kingdom exports excess fundamentalism abroad.

Simply, it funds movements like al Qaeda and ISIS and sends thousands of overly zealous young men to die for them.

Sincerely,

John Robb

PS:  When there isn't a fundamentalist revolt underway, the KSA is always ready to provide Angel funding (as Prince Bandar did with ISIS in Syria).  

Some notes on how ISIS is advancing Open Source Warfare

Here's how ISIS is using open source warfare (wikipedia) to become more effective in conducting their global Jihad:

• ISIS is using open source networks to radically improve the quality and the security of its global insurgency.  Aaron Brantly (at West Point's CTC) says that the ISIS "help desk" is "a fairly large, robust community that is anchored by at least five or six core members who are technical experts with at least collegiate or masters level training in information technology. There are layers of other associates, living all around the world, who allow the service to operate -- and respond to questions -- at any time of the day or night. CTC researchers have spent a year or so monitoring the help desk -- and its senior operatives -- via online forums, social media and other means.
  
  
• ISIS is using the unmonitored person to person communications of the Playstation Network (PSN) to coordinate attacks across the world.  Wrote about something like this over a half a decade ago.  Further >>  Call of Duty and other games are also better training simulators for attacks than the simulators used by Western governments.  
  
  
• Here's an article called "The Coming Urban Terror" I wrote back in 2007 for the City Journal.  Development on this front is moving forward.

John Robb

Urban Terror will Usher in Next Gen Surveillance State

The effectiveness of blood and guts terrorism isn't found in the physical and psychological damage it does. 

It's found in the reaction it provokes.  

With this in mind, will the Paris attacks provoke a reaction that makes them effective?

I think so.

Over the long run, the Paris attacks of 2015 (and those that follow) will be seen as the start of a shift to the next generation surveillance state in the US, China, and Europe. 

Due to a revolutionary technological change currently underway, it will be possible to add pervasive physical surveillance to the proven systems of electronic surveillance (voice, email, chat, etc.) already in use.  

This new capability will make it possible to deploy a cognitive sensor network that will:

1. "know" who everyone in an urban area is (if you ever had a picture taken, it will likely "know" who you are),
2. simultaneously track where everyone in a city is (and has been), and
3. understand what everyone in the city is doing (from voice to behavior capture/analysis).  

Further, this network will learn.  It will get continuously better with experience.  

It will also be proactive.  For example, if it finds a gap in its coverage, it can actively move cognitive sensors to cover the gap (even penetrating structures to gather information).  

Sincerely,

John Robb

How do you control Autonomous Weapons?

How do you control an autonomous system?

This isn't an academic question.  Some organizations already employ truly autonomous systems and these systems are getting very good very quickly.

Since these systems are already in use, I think this question is about as important as it gets.

Unfortunately, these systems are so new, very few people are working on the answer to this question.  Worse, this question is devilishly hard to answer, because a truly autonomous system...

• will solve problems that only human beings can currently solve.
• will write its own "code" and build its own models for solving problems and making decisions.
• will continuously learn/change/improve its code and its models as it gains experience. 

Here's my early thinking on this.

You can't control these systems using the methods we built for controlling the human built software and machines we already have.  If you attempt to control autonomous systems in the same way you control automation, you will fail (and fail badly).

A new method of command and control is needed.  Here are some ideas for how to pull this off:

• Human beings must be paired with these systems.  These people must act as coaches, trainers, teachers to these systems.  They must take responsibility for failures in their training. 
  
  
• These systems must be continually certified for use in way (largely qualitative vs. quantitative) that are similar to how we certify human beings.  Put them through a series of real world exercises.  If they can handle them and explain why they made the decisions they made (optimally, using natural language), they are certified for use.   
  
  
• We need to develop and deploy something I'm calling BIG SIM.  This is a compliment to the BIG DATA that's used to bootstrap these systems to minimum capability.  BIG SIM provides a massive real world sandbox that will allow autonomous systems to undergo extensive training and testing to suss out problems.  BIG SIM can be completely virtual.  It can also be accomplished through decentralized real-world testing as we are seeing with Tesla's crowdsourced "autopilot" or via a corporate solution like Amazon's Mechanical Turk.  

Sincerely,

John Robb