Internet of Things (Stuff)
We can divide computers into the 3 fundamental orthogonal aspects of a computer: Compute, storage and communication.
We are building Quantum computers today.
I expect that over the next 10-20 years we will see a revolution in computing related to quantum computers. The reason quantum computers represent such a “jump” in technology is that as quantum computers increase the number of qubits they can operate on simultaneously the power of their computation ability climbs exponentially not linearly.
It is not clear why such exponential growth cannot continue for a long time so that within 10 years we may have a single quantum computer that is capable of more than the conventional computing power of the world combined today. The power of billions of computers in one machine. This might lead to a recentralization of computation.
Today’s Quantum computers from D-Wave are $10 million a pop and have to be supercooled to -270C. Companies such as Google are testing the technology. At 1152 qubits the current quantum computers are comparable to conventional computers. Considering their cost this is not practical for most uses today. D-Wave has doubled the number of qubits every year for 3 years. If they double the number of qubits next year the computer will be faster than the fastest supercomputers at doing some tasks.
Quantum computers are not good at doing everything but for the things they do they are potentially revolutionary. Many of the things they are good at relate to recognition tasks or optimization tasks. These tasks have been particularly hard to crack with conventional computers.
It is also quite possible that in the future higher temperature quantum computers will possible lead to quantum computers in robots, IOT devices and handheld devices. Some believe quantum computers are necessary for simulating brain function.
Without quantum computers conventional computing is reaching some breaking points. Moore’s law which predicts that we will double computing power every 2 years is starting to falter. We are dividing computation into multiple processors but this scaling of computation power depends on being able to divide problems into pieces that can be independently computed. Some problems are amenable to this and some are not. Without quantum computers we will see fewer significant advancements in speed of computation than we have in the past.
The most important thing happening now is cost reduction, size reduction and power reduction of computation. This trend may continue for a long time.
Computers are vastly less efficient than biologic computers in terms of energy consumed. A human brain more powerful than thousand of computers runs on 15 cents of materials and few watts of energy.
We are seeing dramatic improvements in size and energy consumption allowing us to embed computation into smaller and smaller devices. Ultimately we can imagine everything having “computation” ability. I don’t have a good sense of what that could be used for longer term but I’m sure that as you look at all the problems in the future they will find use for ever smaller and cheaper less resource consuming computation ability. Combined with nano-technology it allows us to animate nano-technology with complex functions.
I admit readily that some of the consequences of our inevitable path to advancement are obviously scary. It is up to us to use technology wisely.
10 years ago people at home frequently had thousands of bits/second to their home and their phones. Data communications over wireless was practically nonexistent or very slow at hundreds of bits/second for the few people who had it.
10 years later cell phone 4rth generation LTE is common which allows communication at 10s of millions of bits/second over wireless and many homes have 100s of millions of bits/second. We are talking about a 10,000 increase in throughput in 10 years. I am frankly shocked this was possible. Nyquist-Shannon showed in 1959 that there were theoretical limits to the amount of data one could transmit over a certain bandwidth. Todays cell phones seem to break these laws. They achieve these amazing feats by employing a tremendous amount of sophistication combining data from multiple antenna with mathematically complex calculations. Cell phones are able to do what should be impossible, transmit and receive 10s of millions of bits of data to each individual portable device over the open air with thousands of other devices in the same vicinity doing the same thing.
What if we could do this again and get another 10,000 increase in bandwidth? One question is what would be the use of 10,000 times the performance we have today? Such a level of performance would be mind-boggling and seemingly unnecessary. It may be impossible to achieve wirelessly but wired communications could certainly see such increases.
The purpose of such communication bandwidth for the average person could only be for virtual reality. If I could create a 3d impression of a distant place here to a realistic enough level I may not need to travel to X to basically experience X. This I believe is where we will be in 50 years. We will use this communication ability to transmit and receive virtual reality images and senses from alternate locations. There is no reason this has to be limited to conventional light and sound.
1) games with any possible worlds
2) physical touch – a suit we wear that allows the transmission of physical touch from an animated anthropomorphic robot on the other end.
3) smell, cold, warmth, radiance or any other external stimuli
4) new senses such as extended vision, extended hearing…
5) remote prosthetic control of a anthropomorphic robot or alternative physical device for surgery or anything you want to do to extend our physical reach
6) into the microscope or the telescope allowing us to have full visual field perception into the microscopic or the ultra-macroscopic realms or control over the level of magnification with possible ability to transmit through gloves to devices that can operate in that realm
I will categorize these things into the following types of virtual reality:
1) VVR – Virtual virtual reality – like the matrix completely made up worlds for play or learning
2) VAR – Virtual anthropomorphic reality – extended our presence through the network to physical devices whose purpose is to give us and others a human like experience
3) VFR – Virtual functional reality – extending our ability to manipulate and see real world things at a level humans can’t do today either macroscopically (large devices) or microscopically or just to hear or see or smell etc better
4) VER – Virtual extended reality – extending our physical capabilities beyond their current abilities possibly needing brain implants or other more direct stimulation to the brain to translate the new sense to the human brain
The continued acceleration and improvement of bandwidth makes it possible to do more and more over virtual connections than physical connections. Today you can buy a device for a few thousands dollars that rolls around with your face on a screen. The device is cute and allows you to be virtually in someplace. You can control the remote robot and run into people in the hall, come up to them at their desk and talk to them.
It’s not hard to imagine that these devices become more and more anthropomorphic. If the remote “me” was connected to my brain in such a way that I could control it simply by thinking then I could believe I was in this remote place. I might need sensors that could feel the other location, smell it, taste it remotely. In this version of virtual reality I have a “alternate physical self” that I can be connected to and experience life from where I am.
Another version of virtual reality would not have a physical presence of yourself but the other place would be presented to you in full 3d. This is a simplified virtual reality but still compelling and useful.
It is clear bandwidth requirements of the future are huge if we want to go towards this virtual reality.
I did not anticipate the increases in bandwidth possible and similarly I am shocked by the gains in storage density. We have chips that fit on the surface of the end of your baby finger that have 256 billion bytes of information. In order to do this we have to write 1.5 trillion transistors onto a surface 0.02 square inches in size. I didn’t think it was remotely possible or if it was possible that it would cost a fortune and be available in 100 years. It is consumer cost and available last year. These things go beyond our ability to comprehend. It is virtually like magic.
It is clear that we need huge advancements in storage to handle the future. Right now people are fretting over the amount of data a few billion stupid IOT devices will generate. I think we have not thought carefully enough about the issues of bigdata becoming really big, i.e. a million or billion times more than we have today. So, as incredible as the storage increases we’ve made are we need lots more.
Fortunately storage is one of the ones that one can imagine easily could be made almost infinite. A simple expansion of our current storage technology is possible simply by making 3d versions of the storage devices we have today. I have no doubt we will find incredible increases in storage are possible. I would be surprised if the future did not see storage devices millions of times more dense than we have today not by going farther in 2 dimensions but by utilizing 3 dimensions.
Technology available very soon
Will allow terabytes of storage at main memory nand speeds and cheap with much longer life utilizing 3d technology as well as 3d conventional nand and ssd are in development. We can easily expect storage densities to explode over the next few decades.
The future of AI is complicated. When I first started studying computers everyone thought the human brain must be fairly simple combination of nerves and that we could figure out how to make smart computers pretty quickly. That was really wrong. 40+ years later we are no closer than we were then making me wonder if this is actually impossible. It is certainly a way harder problem than it first appeared.
In another blog I have written that one possibility for explaining this difficulty is that the brain is actually a number of quantum computers. Quantum computers are potentially incredibly powerful. Such computer power if necessary for intelligence on the order of human intelligence would mean that we are limited by the hardware and are easily decades and decades away from building machines capable of human level intelligence.
We have built a couple quantum computers and they are pathetically small compared to even the smallest possible brain quantum computer if that is the way the brain works. The quantum computers we have built besides being capable of only a thousand qubits compared to the brains potentially trillion qubits cost 10million dollars and need to run at supercooled temperatures close to -270 degrees C. This makes them impractical for all but the most esoteric applications. We need to make several really big jumps:
1) High temperature quantum computers
2) Millions or billions of qubits.
3) Cost reduced by a factor of a 100 billion or so per qubit
This is not going to happen overnight. It may be impossible.
It is also possible that human level intelligence can be achieved without this kind of advancement. Some are worried in fact that the current neural technology will achieve human intelligence eventually using conventional computer methodologies.
The European Union has approved a research effort similar in size and scope to the human genome project launched by the US in the 70s. This project is called the Brain project and the goal is to understand the neuronal structure of the brain down to the ion channels in individual neurons. It is funded for more than a billion Euros. This is a fantastic commitment to try to understand the brain. I am not sure it will achieve what its promoters want but it will undoubtedly make some jumps in our understanding.
There is a project funded by several very wealthy people to create computer programs that mimic brain structure to a high fidelity. Clearly these efforts are being funded because computer scientists have failed utterly to create very smart programs. It’s also true that the research we’ve done on the brain so far has failed massively in terms of understanding the basic functions of the brain.
Some people like Sir Roger Penrose, Prof of Mathematics and Physics at Oxford and probably the smartest man to ever have lived believe that not only do we need quantum computers to build a brain but that the brain reveals capabilities that actually require NEW physics that nobody has contemplated yet.
Computer brains may never show the ability to comprehend reality (consciousness) like humans do and to solve problems like humans do and potentially even some animals do. If so, then it will be necessary to figure out how to recreate a human biologic brain to get to “AI.” This may all go along the path described in my health section around transplanting brains. It may mean the goal of transplanting human consciousness to a computer or to a static storage device is impossible. It may mean we never have to worry about competing intelligence from robots or computers.
However, as an engineer I believe eventually we will figure out how to do it even if it means we have to do a lot pretty much the same way nature did it with a biologic brain of neurons.
Computer scientists have categorized the levels of intelligence into 3 categories and I agree with this categorization. Category 1 is what we have today which is intelligence in a single discipline. This is called ANI or narrow intelligence. These are computer programs that are good at chess or good at recognizing images in a photograph. Category 2 intelligence is what we have and is called AGI (General) intelligence. Penrose believes that this level of intelligence may be unachievable with conventional technology or even conventional quantum technology. Category 3 intelligence may not exist but is called ASI and it is the feared intelligence that we can create something much more intelligent than us that makes us feel like a house pet.
The ability of our science to create intelligence will have far reaching implications on advancements and society over the next 50-100 years because a lot of improvements in our capabilities and leisure are connected with making robots that are smarter (category AGI). If this turns out to be really hard problem that eludes us for decades then we may be stuck dramatically different environment than if we create Robots with brains that can think at a rudimentary level (ANI) and a vastly different environment that if we create AI that is superior to human brains (ASI)
The answer of whether we will be able to do each of these other possibilities (AGI or ASI) is unknowable. It depends on more than our engineering acumen. It is more like long distance space travel in that it may depend on what the nature of reality turns out to be and the limitations of physics.
I don’t believe there is any one thing that suddenly lurches us into AGI or ASI. We are NOT on the cusp of these events.
If AGI or ASI turn out to depend on using quantum capabilities this may mean that things like the singularity event for “recording a human” or transplanting a human are impossible. It is not known how to copy the quantum state of our brains if they are quantum computers and in fact it would violate quantum mechanics to readout the quantum state of our brains. We could readout a particular state of our brains but without knowing a lot more it is probably not possible to have that information be useful for creating a clone of that brain or running that information in a general purpose computer without having the exact brain at hand that held that quantum state.
Robots / IOT
I do think that at a minimum we will have smarter ANI type intelligence in computers and that will mean a lot of utility to humans. Such capability may enable us to build useful robots but it is unlikely that we will ever think of such robots as like humans or a threat to humans.
Whether robots become sentient or not we will have robots capable of doing many things people do today soon. This may become an ethical societal problem to figure out what to do with most people if they don’t have a job?
How does our society restructure around the idea we don’t need conventional labor or workers? Will the society devolve? How do we compensate everyone in such a society? Will “white collar” jobs proliferate fast enough to take up the slack?
I have little doubt that technology will allow the creation of more and more functional robots in every area we operate so that our need for working people diminishes. I think this could be 20 years from becoming really impactful but it will.