Monday, November 03, 2014
In 2011 an IBM computer, Watson, beat human competitors at Jeopardy!
This was a new landmark in artificial intelligence - a computer capable of correctly responding to plain English questions, in real time, by figuring out their intent.
At the time Watson was a computer as big as a room, and it was the only one of its kind in the world.
The original Watson still exists, as discussed in this Wired article, The Three Breakthroughs That Have Finally Unleashed AI on the World, however it is no longer alone.
Hundreds of Watsons are now in operation - not as room-sized computers, but operating 'in the cloud', as distributed software across thousands of open-source servers.
People can access the intellect and computing power of these Watsons through any computing device connected to the internet.
Even more significantly, like many artificial intelligences, Watson is a learning machine that gets more knowledgeable and able to find insights the more it learns. Whenever a Watson learns something, making a new connection, that knowledge is shared with every Watson - making it a distributed intelligence, able to learn at rates far faster than even a single supercomputer, or human, is able to learn.
The power of Watson isn't in the revolutionary algorithms that power its learning, it's in the network itself - how separate Watsons can share knowledge and learn from each other.
This is how humans evolved civilisation - by capturing, codifying, storing and sharing knowledge in sounds, images and words to pass it on from one individual to another.
However Watson hints at a more robust future for human intelligence, and for how we govern ourselves.
Humans have proven over the centuries that having more learners with better knowledge sharing means faster progress and better decision-making. Books, universal schooling and the internet have shown how dramatically a society can progress when appropriate knowledge sharing systems are in place.
The key is to focus on the size and complexity of the networks, not the expertise of individual 'nodes' (you might call them humans).
For computers this means that the more Watsons we create, and the more complex the knowledge sharing between them, the faster they will learn.
For governments this means the greater the transparency, and the more informed citizens are participating in knowledge sharing, the better the decisions and outcomes will be.
Now this isn't how government is currently constituted. The notion of representative democracy is that governance is handed to experts and specialists who live and breathe government so the rest of the population doesn't have to.
We elect politicians who are supposed to representative the interests of their electorates, and appoint bureaucrats whose role is to provide specialist knowledge and operate the machinery of government - develop policy, design and deliver programs, enforce laws and support citizens in emergencies.
By its nature this approach to government relies on experts who are placed separately to the population - often even physically removed and concentrated in a city like Canberra, Washington, Ottawa, Brazilia, Naypyidaw or Putrajaya.
This group (elected and appointed public servants alike) tend to become inwards focused - focused on how to make government keep working, not on whether it actually works and delivers for citizens.
Particularly inwardly focused governments tend to become so removed from their citizens that they are overthrown - though they've usually replaced with a not-dissimilar system.
Now we can do much better.
Rather than focusing on electing and appointing individual experts - the 'nodes' in our governance system, governments need to focus on the network that interconnects citizens, government, business, not-for-profits and other entities.
Rather than limiting decision making to a small core of elected officials (supported by appointed and self-nominated 'experts'), we need to design decision-making systems which empower broad groups of citizens to self-inform and involve themselves at appropriate steps of decision-making processes.
This isn't quite direct democracy - where the population weighs in on every issue, but it certainly is a few steps removed from the alienating 'representative democracy' that many countries use today.
What this model of governance allows for is far more agile and iterative policy debates, rapid testing and improvement of programs and managed distributed community support - where anyone in a community can offer to help others within a framework which values, supports and rewards their involvement, rather than looks at it with suspicion and places many barriers in the way.
Of course we need the mechanisms designed to support this model of government, and the notion that they will simply evolve out of our existing system is quite naive.
Our current governance structures are evolutionary - based on the principle that better approaches will beat out ineffective and inefficient ones. Both history and animal evolution have shown that inefficient organisms can survive for extremely long times, and can require radical environmental change (such as mass extinction events) for new forms to be successful.
On top of this the evolution of government is particularly slow as there's far fewer connections between the 200-odd national governments in the world than between the 200+ Watson artificial intelligences in the world.
While every Watson learns what other Watsons learn rapidly, governments have stilted and formal mechanisms for connection that mean that it can take decades - or even longer - for them to recognise successes and failures in others.
In other words, while we have a diverse group of governments all attempting to solve many of the same basic problems, the network effect isn't working as they are all too inward focused and have focused on developing expertise 'nodes' (individuals) rather than expert networks (connections).
This isn't something that can be fixed by one, or even a group of ten or more governments - thereby leaving humanity in the position of having to repeat the same errors time and time again, approving the same drugs, testing the same welfare systems, trialing the same legal regimes, even when we have examples of their failures and successes we could be learning from.
So therefore the best solution - perhaps the only workable solution for the likely duration of human civilisation on this planet - is to do what some of our forefather did and design new forms of government in a planned way.
Rather than letting governments slowly and haphazardly evolve through trial and error, we should take a leaf out of the book of engineers, and place a concerted effort into designing governance systems that meet human needs.
These systems should involve and nurture strong networks, focusing on the connections rather than the nodes - allowing us to both leverage the full capabilities of society in its own betterment and to rapidly adjust settings when environments and needs change.
We managed to design our way from the primitive and basic computers of the 1950s to distributed artificial intelligences in less than 70 years.
What could we do if we placed the same resources and attention on designing governance systems that suited modern society's needs?
And it all comes down to applying a distributed model to governance - both its design and its operation, rather than focusing on the elevation of individual experts and leaders to rule over us.
It's a big challenge, but for a species that went from horses to spaceships in two generations, it surely isn't an impossible one.
And given that societies thrive or die depending on how they are governed, are we willing to take the the risk and hope that our current governance and political systems simple evolve into more effective forms within a human lifespan?