Raison d'etre

Alternatives

Theory

Transformation

Technology

More...

 

Resourcism - a new economic system for the coming age of automation

A Paradigm Shift in Thinking

There is difficulty for most observers to understand why we must move to Resourcism.  The challenge is having a clear understanding how technically these new autonomous systems can and will work in future. Once this becomes clear, the implications of this (fairly limited) technical advancement challenges the thinking behind current economic models.

On further analysis it becomes obvious that no amount of tinkering with current economics can handle the move away from a jobs based society. We need a new model.

Challenging Your Viewpoint - John's got a new motor

In March 2023 the charity named the new BMW all-purpose vehicle after their founder Mr John.  And so, the metallic blue BMW was commissioned at great expense.  This was a big step for the charity’s goals.  The car was a new type of completely autonomous vehicle, no driver required, not even a steering wheel.  Perfect for the charity's aim of providing a 24/7 car service for local disabled people. The BMW named John had wide automatic doors and a fancy wheelchair that could take itself to the doorway of the callers house or business and back to the car, safely installing its passenger.

Soon, the first call came from a registered user’s smart phone.  The new electric powered car moved away silently, the charity’s people instinctively waved goodbye, The BMW named John did not notice.  The registered user’s account was debited with a small charge for the use of the vehicle.  This charge was designed to cover all running costs including the long term depreciation so the car could be renewed and the charity's good work continue. 

After the ninth drop, The BMW named John decided it required a charge before another pick up.  So it amended its app for a twenty minute delay and drove itself to the nearest charge point.  Paying for the charge from its bank account, the BWM named John went about its business again. 

The weeks went by without a hitch, having covered 10,000 miles a service and check-up was required.  Logging off for half a day the BWM named John arrived at the service garage and was expertly attended to.  Knowing everything was correct, the BWM named John paid for the work and started back on its set task delivering a reliable service at a very small cost for the local community.

The media wrote about the new autonomous car, as its work was exemplary and people liked it a lot, but the BWM named John did not notice.  It just kept going 24/7 without delays, or days off and with perfect attention to detail.  The BWM named John was a huge success and soon many more joined it on the busy inner-city roads.  Time passed, and the charity realised that the BWM named John was due for a replacement as it had reached the optimum period for its battery renewal.

The BWM named John arrived back at the charity for the first time since it had left.  The charities people came out on the street to look at it again, they considered it an old friend now.  But there was a difference, the BWM named John was in a darker blue livery.  And it seemed a little different, almost proud with a higher chassis and a wider wheelbase.

The BWM named John had purchased its own replacement from the original dealer.  Paying for its renewal directly from its own bank account, the BWM named John was now a second generation model.  This model had improved functions and its new app provided additional communications so that it could find out about its passengers and deliver a more personal service.

It drove up to the pavement adjacent to the director of operations of the charity and opened its doors, the director was urged by his colleagues to take a ride.  The BWM named John took the director on an inaugural lap around the city.  While the director was enjoying the smooth and quiet drive, the BWM named John texted the director with a simple message, 'John's got a new motor'.
 
(The charity’s lawyers also received a text – it just said; please change the registered owner to ‘the BWM named John’, thank you.’)

-------------------

In June 2017 Intel reported that they expect the Autonomous vehicle market to be worth 7 Trillion Dollars by 2050 (& save half a million road deaths)

Rolls–Royce says you can expect to see a fleet of 'drone ships' on the high seas by 2020 & Fully Autonomous Ships by 2035 (see pdf)

-------------------

Understanding the Technical Issues

Few people are have the technical oversight to understand the many technologies involved with creating autonomous smart systems. These will be able to do many different tasks and communicate well with humans.  For most of us this is a step too far in our normal appreciation and acceptance of what is possible.  We need to break the steps down into distinct parts to enable this understanding.

The run up to Smart Robotic Architecture (SRA) may be some years away.  However, the base technologies are already in place, although not fully implemented.

The core technologies revolve around the following processes;

So, we can see a path, not too distant, for these technologies to mature to create the first tentative steps to Smart Robot Architecture (SRA).



Development


The key areas for development towards smart autonomous systems are those of context and reliable action based on this context. Recognition for both visual and voice is already advanced. For image recognition the error rate has dropped down to 4.7%, this is almost the same as human capability.  Recognition for human gestures and facial expressions is also improving fast.

The Rise and Rise of Neural Networks

The path towards Smart Robot Architecture
(SRA)has been focused on systems that attempt a small range of actions.  Such systems are used in mainstream manufacturing, being trained on specific such as the assembly of parts.  These are not smart in the sense that they can apply wide actions in context.  There are already neural networks that provide good results for specific task based on massive data sets.  These include areas of medicine, where the neural system provides more accurate up to date information than the medical consultant can achieve.

We have recently seen neural network trained systems win games against humans.  IBM’s Watson convincingly won the American Jeopardy game show against several past winners.  This demonstrated that it is possible for a neural network to interpret speech and context to a high level.  Google’s Deepmind system beat the champion go player in Seoul last year, and has just won again, beating the current champion, in May 2017.  This game is not like chess, it has a creative aspect in which humans excel and the first win astounded many who thought it impossible.

In January 2017, an insurance firm in Japan replaced more than 30 employees a system based on IBM's Watson Explorer.  The system was introduced to increase productivity by 30 percent, and will give the firm a return on its investment in less than two years.

Emerging systems are now available for interacting with humans to provide a service. Enfield council worked with IPsoft, to build a new “cognitive agent”named Amelia. Her personality and social skills are based on natural language processing and the system learns how to interpret the emotion expressed in a human voice, so as to know how to respond appropriately. The hope is that callers won’t even notice that they’re not dealing with a human.

Over time we shall use more powerful neural networks and better training processes that continually improve the ability of such systems to interact with people and use particular actions depending on the environment they detect around them.  There will become a point in time when these systems seem truly smart and can apply a large range of skills, in context.

How Do Neural Networks Work ?

The underlying process for this recognition and context is not too difficult to understand.  They all use what is called machine learning.  This does not use logical programming that has to be written in computer code.

The machine learning approach is to collect a lot of examples, and specify the correct answer for a particular input.  A machine learning equation then takes these examples and produces, on its own, a best fit for the answer. This best fit system weighs up the examples learnt to produce the best guess.  If we do this correctly, such a learning system works for new cases just as well as the ones it's trained on.  And when the data changes, it can easily be retrained on the new data. 

The Next Step

All these current systems use a huge amount of computing power, mostly within special units containing racks of linked processors and memory.  These also use a massive amount of electrical power.  Currently most of the neural network systems work using GPU (Graphic Processing Unit) that were initially developed for image possessing. Due to their parallel nature they fit the processing tasks for neural networks that require massive amounts of data throughput.   

The next advance will see these systems become more efficient and consume much less power.  There are a number of specialist new types of electronic processes, (such as TrueNorth), that should help this transition.  These are built to create an electronic system that mimics the way neurons and synapses work.  Some of these neuromorphic systems (eg BrainScaleS) use analogue circuits and produce spikes emulating brain activity.  IBM’s TrueNorth uses only one thousandth of the power of a normal chip to do the same image processing task even though it has 4 times the number of transistors of a normal CPU.


email: cst@commonsensethinking.co.uk