Let Your Machines talk

Let Your Machines talk -

M2M Technology Partnering event

Marina Congress Center, Finland

3 March – 4 March 2005

Chairman, ladies and gentlemen,

I worked with Helsinki Telephone during late seventies. I connected my computer to the telephone network. It called random numbers asking: “Do you want to play a game, press one if you want to play.” Many people really wanted to play but I quit the experiment. Someone actually paid me to have fun. Well, not this much fun. I was afraid they sack me. I had built a robot for harassing people. Today it is illegal.

Technically I had few problems. Some telephone exchanges did not properly signal when the call was connected. Another problem was that touch tone phones were not yet widespread. Pulses from the dial did not propagate clearly through exchanges.

My next experiment allowed people to call my computer, which then called another computer with tens of thousands of pages of news and other information. My computer retrieved text from the other computer and read it aloud using a speech synthesizer. Users only needed a touch tone telephone. In 1981 this was turned into the world’s first widespread electronic home banking system.

Since then things have not developed as fast as I would have imagined. World Wide Web has naturally fulfilled Ted Nelsons prophesy of Xanadu. PC:s of the world have been united, but different devices are still too often strangers to each other. Few years ago my fridge and my doorbell started talking to my mobile phone. Many wolves and reindeer send their GPS-positions as SMS messages. Home automation is integrated and sensors and actuators talk with each other to provide optimal environment to each situation with minimum energy. Much else has also happened as we have heard also here today, but I would have expected more. I guess we will have more but it just takes longer.

Dear colleagues,

we have gathered here to talk about our machines. Would it not be funny to hear that our machines are having another conference just across the street? Life holds many surprises but our machines seldom surprise us. What if Joyce the Android would gossip like servants in the old books. Some viruses already publish random content in newsgroups and children molesters are searched with verbots in chat groups.

Smallest machines in research labs are built from few molecules. They could be built to gossip with our DNA. Newest studies show that corruption of mitochondrial DNA is perhaps the major reason why we grow old. Newer, more complex parts of our genetic structure can repair themselves much better, but mitochondrion cannot. Perhaps nanotechnology gives us machines to service our mitochondria.

We use machines to manipulate our physical world, but we also use machines to replace it. Physical traffic signs seem already obsolete. Directions could be viewed from virtual windshields, call phones or augmented reality glasses. Perhaps we also could soon smell, feel and taste virtual reality.

Negroponte saw atoms changing into bits. Electronic devices are now used to build muscles. We have a long way to go before bits sustain us fully, but some people already consider computer viruses to be a life form that should be preserved.

There is no end in sight what our machines can do. We do not even need much creativity to proceed. We could just study what the nature does; reverse engineer it and just imagine what wonders we could have.

Eagles hunt skilfully and carry heavy loads across the skies. Shopping would be easier if our grocery bags hunted food from shops and flew back home after a refill. Do not say it is impossible; engineers are just not as good as random chance in a few million years. Programmers have not even approached insect level. Bees and ants survey their environment, perform statistical calculus and communicate efficiently. As a group they both seem to be more intelligent than many human organizations. Even bacteria communicate more efficiently than most of our machines. They copy each others best features. Most computers only copy destructive traits automatically.

Artificial machines have their history. Early machines did not contain a power source. They were most probably used to catch fish or wild animals. Actually I remember my mother’s sewing machine, which required human power. Hot air balloons might have been first vehicles to contain their own power source. Chinese also invented gunpowder and rockets. Greek machines, including water clocks, seem to have surpassed all others in complexity. Few years ago an ancient navigational computer caused us to rethink what other wonders history has forgotten.

Early machines did their job standalone. During the so called dark Middle Ages we started using water and wind as power sources for grinding grain. New Ages saw us developing separate engines to power our machines. They were called automatons.

It is not a new idea that one machine can do something useful for another machine. Early industrial age saw hordes of machines aligned in a row, powered by one huge engine, creating the style of long narrow factory buildings we still see today. Trend changed after the unsurpassed creativity of Thomas Edison. Power networks joined machines that produced electricity to others that consumed it. Later also information started flowing. The concept of a broker was borne to receive information or power in various forms and distribute it further to many others in a standard form.

Automation used to mean machines that did not need to be used. Today we seem to spend more and more of our time using or servicing our automated machines. Luckily we have machines that build our machines, but our machines still depend on us.

Our younger daughter empties the dishwasher every morning, but when she leaves home, I would like to have a robot for that task. James or Joyce the robot could also have evening tasks. A glass of whiskey to me before it sat down to play the piano. Joyce could perhaps relay my daughter’s real time image on its surface, relay voice and movements. We got used to machine relayed telepathy, but perhaps machine relayed avatars are too much.

This is still only a dream. We serve and operate our machines now more than ever. Some computers do communicate efficiently with each other and of course we seldom any more provide muscle energy for our machines.

Banks have made almost as much progress as factories. We operate machines that communicate with other machines. There used to be a time when we used machines to print paper that were sent to banks where people read the papers, keyed the information to their own machines and sent us paper that we again might key into our machines. Some called those routines work and complained when it was automated.

My late uncle used to tell a story of a little boy and his father. Always when the father went to plough his fields he took the boy with him to sit on the plough. He told the boy how important it was. The boy was proud to be of help to his father. Once, the boy was sick with fever. He worried about his father and his work. Father soothed the boy saying: “Don’t worry; I put a stone in your place.” My uncle used this story to show how bad automation was. To me the story told how bad it is to live in a lie.

Machines have few needs. They may need suitable environment, monitoring and maintenance; they need energy, raw materials, components, directions and paths. Some architects have compared homes to machines, machines for living. Similarly, machines can well supply the suitable environment for machines. Monitoring is often also a suitable task for machines. Atomic power plant and a jet plane are examples of machines that are monitored by hundreds of other machines. My home also has some hundred sensors and dozens of computer modules monitoring the conditions.

Energy supply is easy. Some automated vacuum cleaners can plug themselves into electricity sockets when they are running on low battery. I wonder why we still need to operate the gasoline pumps by hand. Maintenance is more difficult. Fault detection is now often computerised but robots are not generally used to exchange broken parts. Robots are not yet fit for the job. There are too many different devices to be fixed, too many faults. Robots would need to be programmed for each specific task.

But if I remember correctly, remotely controlled robots perform surgery and it is much safer to perform brain surgery through waldoes than by hand. Specialists could remotely assist robots in demanding functions and perhaps one specialist could direct dozens of robots if the robots could perform some basic functions themselves. This is certainly a field that is underdeveloped. Today millions of people use largest part of their time travelling from one machine to another and using few minutes to fix each. And we may wish such auto diagnostic machines and intelligent robots that robot maintenance would become practical. But let us proceed to other needs.

I have a small bronze sculpture which is actually printed with a special 3D laser. The device can manufacture any kind of 3D objects from bronze or pulverized iron. When these and miniaturized robots become common, we might all have small factories, flexible production lines home and then we need hardware stored only to buy pulverised iron and other materials.

Automated logistics is progressing. RFID enables much more. But we still use human labour to feed our manufacturing machines with components and raw material. And even though music and video CD’s today are manufactured at home, most of our products are collected and brought home manually.

Jules Verne understood already in 1863 when he saw one of the first internal combustion engines that Paris would change. Roads would become wider, leaving less room for houses, gasoline stations would be needed and suburbs would arise.

Most people considered automobiles worthless still decades later and did not believe they would become common because roads were not good enough.

Today our roads are not good enough for machines that drive by themselves. Robots need a clear environment. Computer vision does not equal that of a honeybee. Neurologists and computer scientists are making progress however. How our own brain understands what it sees, needs to be reverse engineered. It is actually much more difficult to create an algorithm to recognise a chair than to win a chess master. In the mean time autonomous computers can’t be allowed to roam freely amongst us.

It is quite surprising to notice that flying is much easier to computers than driving. Autopilots have landed us through fog, space shuttle has never landed manually and very few people are able to learn how to fly a helicopter. For a computer it is not so difficult. My first computer in 1977 used RCA 1800, which was designed for the first cruise missiles. Today flying is much easier. GPS gives directions; GSM connects autopilots to automated flight controls. Sadly small aeroplanes still follow designs from the First World War.

Moller has designed both the engines and technology that could change this. Moller’s engines give two horsepower’s for each kilogram of mass. Smallest engines will initially be used for hand held machine tools, but they could equally well be used for flying grocery bags. Larger engines have been tested in the first flying car. I attended one of the tests and computers lifted the car and hovering. The device is theoretically capable of carrying four persons 1000 kilometers in two hours rising from ones back yard. We have routers that direct IP packages from million of destinations to millions of other destinations automatically. It is not a huge problem to automate flight control and piloting. Nasa is actually doing that already.

Then I could put my mother in law to Skycar, and say “Mother in law home”, to my Skycar. This shows the importance of physical and electronic networks, standards and brokering to any logistics problem.

It may be a bad design principle to adapt our environment to our machines. Our environment should be adapted to us and our machines should cope with what suits us best. I guess if we make machines to be more like us, we simultaneously make them better capable to recognise our artefacts and surroundings. Most of our surroundings are built for our senses and intellect.

We will naturally have better vision and other senses for our machines. I would not be surprised if someday we would see electronic pigs searching through the woods for truffles and electronic dogs guiding the blind, especially in developing countries where people might eat the real dogs. New senses enable our machines also to find other machines and take care of their needs. Robots that search for landmines have a novel interpretation for terminal communication.

Future visions are like caricatures; they show directions where we could search for more moderate innovations. But let us now take a closer look on how we humans use our machines. User interfaces are naturally different for a hammer and a computer. But when we think of exchanging information with our machines, each new interface actually requires us to learn a new language. If I could choose I would communicate with familiar machines only and let my James handle all the others, just like C-3PO in Star Wars. Besides, if we carry Swiss army electronics in our pocket for multipurpose communications, why build similar capabilities to other devices at all.

It may seem strange to use your mobile phone to check air pressure in your Nokian tyres, but it would be perfectly all right to use your mobile to remotely program your DVD recorder. But can we really expect each device to understand each other?

We should finally take seriously semantic web and ontology issues. Berners-Lee has a new mission and it is the essence of M2M communication, more important than internet itself. We need to understand that information infrastructure is not just protocols and cables. It is semantic and ontological content. Naturally all new devices need to be available through IP-networks and naturally we may end up having computer viruses in our animated tattoos or brains but I’m getting ahead of my story.

We get a different viewpoint when we consider why human beings communicate. Protocols, semantics and even ontology can often be cast aside. The most important and basic human communication, positive and negative feedback loops – display and understanding of emotions is genetically and globally standardised. Ontology is also defined. However lacking, Maslow’s theory of human motivation is a major foundation to explaining our behaviour.

Way too often technically oriented people wonder why new devices become popular even though they are not useful. Useful refers in this case to saving or earning money. If you reconsider what useful really means, you understand much better what motivates people. If our physiological needs are fulfilled, we are motivated by security needs, then closeness needs, respect needs, cognitive needs, aesthetic needs, self actualization needs and altruistic needs.

I wish I knew what needs were filled with tamagotchis. It is clear that machines feel like pets and daily soap operas like family to some people. I feel my house respects me when it reacts to my presence. Many people are proud of their shining cars or huge machines that they maintain, own or use.

What if machines operate each other, where is our place? There seem to be many routines that cannot be automated. Sensors can turn lights off if nobody moves, but if you were quietly reading a book, you get irritated. We need to give situation related manual instructions when machines are unable to understand our needs.

Neural interfaces will lead to machines that read our minds. There is no need to notice light switches any more. Warm or cold is even easier. And neural stimulus directly to pleasure centres may end sadness and bring eternal joy. Our machines just need to sustain our physical body, but again I reach too far.

Today machines are still too dumb for most practical purposes. And they are especially stupid when they would need to understand dynamically changing environment or communicate with other machines.

Think about an anthill. Ants have simple behavioural rules. They use elementary communication protocols with each other. But this simple behaviour emerges into a collective that is very intelligent. An anthill studies its environment, optimises resources to match opponents and best sources for food and construction materials. Dynamic events are matched with very appropriate and fast responses.

An anthill represents swarm intelligence. New studies show that our own intelligence is also distributed or modularized. Society of the mind is one clever term for our brain. It may be that humanlike machine intelligence requires semi-intelligent purpose specific modules that communicate with each other.

Intelligence does not arise without intentions. Thermostat is a simple example of intentional device. Intelligence requires capability to navigate through obstacles. Thermostats are not very intelligent. Consciousness is related to intelligence. It contains awareness of memories and events that are relevant to our intentions.

Imagine a computer that wishes to socialise with people in chat rooms. It might have some analytical capabilities but it also might read trough every other internet discussion and copy the structure of its reply from similar discussions elsewhere. It might use several search algorithms and learn to find what opening lines work. How similar this sounds to what young men do when starting to court girls.

Marvin Minsky has studied emotional machines. Human face has got a number of muscles. It is now clear that tensions in these muscles correlate directly with our basic emotions. Computers can read from our faces what emotions we have. This would actually be useful for us men also because we cannot read women’s emotions as well as they can read ours. But machines can also emulate our emotions.

Think about any face to face discussion. Much of it is just emotional expressions. Imagine a computer face that smiles and soothes you when you are afraid, responds immediately with shining eyes to your smile. If tamagotchi got millions and millions of kids addicted, how far will this kind of communication lead us?

But emotions are not just beautiful faces or illusions. Antonio Damasio, Marvin Minsky, Steven Pinker and many others have shown that emotions are very important in our decision making process. We have many goals. Emotions select what goals we should pursue at any given moment. This leads to fast decisions in fuzzy situations. Emotions connect our rational mind to our intentions, needs or motivations.

Marvin Minsky has claimed that machines need to be provided with emotions to enable intentional intelligence. This should not require too much. Very low level animals have some emotions like pain or aggression to direct them.

Could you now stretch your imagination a little? You can see a sad little mail carrying machine sitting in your back yard. It has clearly run out of energy. Your household robot notices a new machine and reacts immediately to its sadness. It goes out, asks for explanation and soon provides the guest with new energy. Your robot has an intention to keep everybody happy in its realm. Guest smiles and flies away.

We are not far from a future where our environment is aware of our needs. Our clothes and other materials will be intelligent, our skin may have animated tattoos, and our mobile phones may arrange dates for us if interesting people pass close by. Our blood may be replaced with nano scale machines to carry oxygen. We will become cyborgs. This is Stephen Hawkins desire. Otherwise he sees that we will become useless as machines will do everything better.

This is a common horror scene in science fiction books. Chief scientist for Sun, Bill Joy popularised this viewpoint few years ago. He wrote about nano level machines that duplicate themselves and cover the whole world. He also wrote about machines that surpass our intelligence. He called for stricter control of science.

It is clearly a possibility that we create machines that create machines. We can also provide the machines with intentions that in the end are harmful for us. We have already created a world where people are unhappy and mentally ill. We have created machines that autonomously kill people and destroy other machines. We also have created machines that humour us better than other people. We have TV.

But should we stop or create a brave new world where research is banned. Or should we open up, build glass walls to see what our neighbours do. And could we just trust our genes. People who prefer machines over other people will not reproduce. The other kind will become extinct.

Humans will in the long run be necessary to other humans. It is clear that machines will increase stratification if current economic rules are not changed. Machines will however also become cheaper than ever. It is like Milton Friedman said: “Kings did not need running water, they had running slaves.” We need our machines, but we need to better understand the psychological, societal and economic effects. They will increase when machines learn to communicate and create their own societies where we do not need to act as their brokers.

Have a nice future, thank you for your interest.