1. This site uses cookies. By continuing to use this site, you are agreeing to our use of cookies. Learn More.

Chapter Six: What’s needed for full automation?

Discussion in 'Eventually Everything will become free.' started by amusicsite, 26 May 2018.

  1. amusicsite

    amusicsite dn ʎɐʍ sᴉɥ┴ Staff Member

    Location:
    UK
    It may seem like we are a long way off full automation but there are only a few short boosts in technology we need for it to work. The first one is the conversion of everything to run on electricity wherever possible. We have already done a good job of that with being able to cook, heat and light our buildings using electricity. In fact we consume more electricity every year since it’s invention and that’s even with us massively reducing the amount of power needed by things like computers, lights, TVs and other items. With the upcoming electric vehicle revolution the demand is going to grow even more. This is good though, because it’s a lot easier to get electricity from sources other that fossil fuels. It’s very portable and electricity is easy to generate anywhere on this planet, even in space or other planets and having most things run on it will reduce a lot of the supply chain for our energy.

    Solar generation will eventually generate most of the local electricity we need on Earth and even now there are people looking at producing solar roof tiles that could last hundreds of years. If every time you replaced or built all new roofs in the future was made to capture solar power, eventually you will have more than enough power to supply at least 25% of the worlds needs. This will happen when the new solar tiles become cheap enough that it barely makes sense not to use them. The rest will come from large solar instillation in deserts, on artificial islands or other waste land.

    There will probably be other things in the mix like wind, tidal, wave, geothermal and nuclear power generation but I think solar will be the most common. Especially if we move some of the heavy work into space which will reduce the power needed on planets. With the addition of storage systems and a global grid that can move around electricity, then you have a clean cheap form of fuel that can run most things.

    By the time we get to that point we will want to transform any non-electrical process to be electrified. Whether it be a furnace that melts metal, a tunnel boring machine, delivery truck, shipping container or eventually even aircraft. There will be a massive drive to run them on electricity because it will be cheaper than paying for any other fuel. A lot of this is already underway with people looking at electric boats, planes, production facilities and machinery. There are not many areas that look impossible to convert to electric and, as the cost of electricity drops, it’s only going to get more tempting to convert everything over to using electricity as soon as possible.

    The other main advancement needed is batteries, or at least portable electricity in some form. Mostly this will benefit the transportation industry and the power grids. The demand for longer range vehicles, higher power for large ships or planes and smaller, lighter batteries with more power and faster recharge times. There are some promising tech on the horizon that looks like it could solve the power / size / weight recharge problems with ease. When that is sorted you will be able to have a small box in your house that could run your whole house for weeks. This will also tackle the typical solar problem of “but sometimes here we get no sun for weeks”. Eventually we will may get to a point where electricity is so cheap and clean we barely think about how much we use.

    The next thing we need is robots, lots of robots. The advances in batteries will really help this area too, as one of the main drawbacks of robots at the moment is powering them when they are not near the mains. It’s fairly easy to have a robot that can stay in one place and make a car or move large chunks of metal. It’s quite another to have one that can go to the middle of a field and pick crops all day. You could probably make the robot that will pick crops work in the lab with ample supply of electricity but add on the weight of a current battery and it suddenly gets a lot harder to do that task the field.

    Once we have power packs that can keep a robot going for more than 12 hours then I think we will see a massive explosion in the use of robots. Electric vehicles may well have a roll to play in this, after all they are basically large batteries on wheels. Just having more electrical vehicles on the road is going to drastically increase the amount of electrical add-on devices you will get that can run off your car’s spare power. For robots they can be delivered by an electrical vehicle that then goes off, plugs itself into the mains and recharges it’s self, then come back when the robots are almost empty, charge them back up with it’s spare electricity and go off to recharge it’s self again. 24/7 remote robot work where they never need to plug directly into any electrical generation source.

    Obviously robots need more than just power to work well, they also need to become cheap enough and good enough to be more economic than humans. It does sometimes seem like this is happened all the time, recently I heard about a new machine that could lay bricks faster and better than humans could, a few months earlier it was a machine that could plaster a wall quicker and better. Though both of these examples are too expensive to be cost effective now but the price will drop over time. It starts with the simple repetitive tasks like this then as these get tackled the emphasis is put into solving the next task. After all you don’t need to continuously improve everything to get to full automation, you only have to make it slightly better than humans at all the tasks humans do. Constructing buildings would be one of the key tasks to sort out. We use a lot of power tools and automation already to construct building, sometimes whole cities, in an incredibly quick time. Often with 2-3 shifts of workers so production is going on most of the time. Though you have change over periods of no work, toilet breaks, 2-3 times the amount of staff to pay, 2-3 times the number of people who could be ill, hurt or just not turn up for work one day. If one robot could replace three shift workers and truly work 24/7 then those people are out of a job.

    We already have automated diggers and earth movers that work by GPS or lasers to work the land. It can’t be that hard to automate laying foundations, we have the brick laying robot, ones that plaster too and probably not too hard to get one to put on the roof with solar tiles. The tricky things at the moment are the fiddly jobs like wiring in the mains, doing the plumbing, screwing in the doors and things like that. For these tasks we have the tools but they just need applying to the task and a bit of improvment. The latest tool is machine learning A.I. which you set up to be able to watch and understand a few fundamentals. You show it hours of footage of humans doing the task where it learns the ins and outs of the job. Then you let it have a go and rate it each time it completes the task. Over a short space of time it learns the best way to do the job. This seems to be able to be applied to just about anything. You can even get them working in virtual environments to try things without wasting resources. So for example give the plumbing A.I. a virtual house to practice on or get them playing about with virtual machine building tools to experiment with ways to make the robots that will do the tasks better than a human.

    Just like with the physical automation, it’s the simple things that are being done now. Get the robot arm to watch someone pick up boxes and stack them, then learn how to do the same. Ultimately though we will move onto more complex tasks where these learning A.I. machines will work out how to improve everything, even themselves. In fact some of the new massive servers running the latest high level A.I. systems are built on hardware that A.I. machines have designed to make the task of running an A.I. system better, faster, more efficient and more powerful. The robots are starting to build the robots.

    Now this could escalate quite quickly. If you can make a robot that can make a copy of it’s self, then you have two robots than can make copies of it’s self. Then you have 4, 8, 16, 32… Within a few generations you have hundreds. Then you can turn 80% of them into robots that make other things and the number of robots increases as does production of other things. Within a few years you could have thousands of robots making millions of things. Within a few decades you could have millions of robots making billions of things. Building the A.I. systems to make them better and learning how to make new things. Building bespoke robots to perform specific tasks. As well as building the things we need or want.

    The last thing we need is interconnectivity, where each automated part of the process can easily access the next part of the system. This is most useful for the current electricity supply systems, where we have regional or national grids that evenly distribute power around the place. With the increase of solar these will be supplemented within local grids that share power around and intercontinental connections that link up the world into a super grid. There are plans for both these already.

    There have been experiments in some cities of solar + car storage and supply. The concept being that you have a connected grid of solar panels on a few roofs which are connected to car charging points and possibly home batteries. During the day the solar charges up all the batteries that are connected. At night the home and cars batteries power the homes, then as demand drops they move the power to all the car batteries or take power from a car not going to be used for a while and use it to top up cars that need more power the next day. As soon as the sun comes up in the morning everything not at 100% starts charging up till it’s full. Even on the small test cases it’s been proved that you don’t need many cars or charge points to make the whole thing work very well. Once scaled up then you will get even better performance as the cars will probably be able to top up whenever they stop anywhere for a while.

    On top of this you will have the regional grid that could share this power to other regions if there is a lack of sunshine or higher than usual demand in one area. With self driving cars, especially if they can also plug themselves in, you could even have cars with full batteries drive to areas that need a boost in power and they can offload some of their charge. Electricity is fairly easy to move around. Alongside this there is already talk about a super-grid or backbone that could carry huge amounts of electricity between countries and continents. So Africa could share power with the Americas while it’s sunny in one part but night-time in the other. Or the hemisphere in summer can send power to the other side of the planet in winter, where more power is probably needed.

    It’s not going to be cheap or easy to build these massive backbones but China seems to have already seen this as the future and is rolling out what has been called the new silk road. It’s plan is to have high speed rail road and grids connecting up every major country on the planet. A railway from China to France, from Spain to the far east of Russia, from Alaska to the bottom of Argentina. Coast to coast north-south and east-west connecting all the ports and countries together. Eventually with a tunnel under the sea between Russia and North America to join them together. One of the principle aims is to be able to catch a train in London, go through Europe and Russia, under the sea then via Canada and the USA to end up in New York. Along with routes across Africa from somewhere near the Middle East to the west coast and from east to west coasts of South America where the aim would be to reduce shipping times by loading the goods onto trains to cut across land rather than having to go around the coasts on long journeys.

    The aim of the new silk road is not to make money off it directly but instead to increase trade along it. History has shown that if you can make it easy to get somewhere and send goods there, then trade, ideas and technology flows to that place. It has often been noted that the main problem with 3rd world countries is you have a lack of good roads and transportation links. Improve this and you improve the country. Also if these trains are electrically powered then you can build the electrical backbone alongside it to move power around. There is even big plans to move fresh water around to where it’s needed but in short supply. This is not going to happen overnight, especially at the moment as geopolitics makes it tricky with some of the countries you will have to go through as well as it being very expensive. I’m sure if they continue with this ambitious project then the technology and automation needed to do it cheaper and quicker will come along.

    This new silk road will also be something that machines and robots can hop onto and use to get around the world to where the resources are or where the work is needed. So the right rail repair robot could jump on a train, travel to where there is a landslide or other maintenance job needed, get off the train and fix the lines then be dropped off at the next place. Self driving vehicles can move them around more locally to fix a broken pipe or repair a roof or move a number of them to a site to build a new factory or housing.

    So while things like putting solar on all the roofs, making enough batteries to power every vehicle and machine along with a road or rail network that covers the whole planet may seem like massive tasks that will take forever to do. As we progress towards full automation the price and feasibility of doing them becomes achievable. A lot of it is not much more impossible than what we put into place for cars. If you went back 100 years and said there would be 1 billion cars in 100 years time, which is about what we have now. People would have said you would need millions of miles of roads, thousands of petrol stations hundreds of oil refineries and the like. All things we take for granted these days.

    The more we automate things, the cheaper they become and the more we can achieve. Just like there is no point having a car without good roads and a local petrol station, you need the whole system to work for it to be useable. The same is true of full automation, you need to get the whole system working well. Recently Tesla tried to fully automate making cars, the same as GM tried in the 80s. Both were a failure as the main problem is getting the whole system working well together. If you try and add 30 new automation machines to a system with little track record, then you will end up with hundreds if not thousands of problems to solve. The roll out of automation works best when you add one more layer at a time, debug it, get it working well and then add more automation. Just like the road network we have not was not built in one go. You start of with the main routes, add in local roads, improve and expand the roads and the maintain them. The same will be true of automation. You could say at the moment for automation we are at the level of having a few motorways and trunk roads but not a full road network. Maybe like in the 60s, the UK’s first motorways were finished in the late 50s, by the end of the 60s we had a reasonable road system, like the reasonable level of automation we have now. Over the next few decades roads exploded and covered the world. I think automation could be entering the same sort of time.