Erasmus : The creation of the transistor in silicon essentially was a ground-breaking innovation allowing for progressive miniaturisation of transistor structures. The development of the transistor represented a “Paradigm shift” in the way vacuum valves operated.
This has generated the exponential growth curve which we know is Moore’s law. Eventually this technology will reach a plateau in development. When a new innovative technological breakthrough comes along, I think we’ll see further patches of exponential growth.
Early 20th Century Computer
Kinkajou : So what would you describe as the strengths of a computer?
Erasmus : Computers have excellent un-corruptible memories for a huge range of facts. This information can be transferred into computer memories at incredible speeds far beyond those that can be achieved with inputting information into biological systems.
Computers are able to undertake complex calculations with an incredible speed. The human brain is capable of taking the same calculations, but is not optimised for dealing with the processes involved in these calculations. Hence the human brain in specific tasks comes off as a poor second place to the capacity of a computer.
The big question of course is can computers do what a brain does? This remains to be seen.
Moores Law Computing Examples
Kinkajou : Do you have any other comments to make on the achievement of the technological singularity?
Erasmus : A new concept I have heard is that of the “Complexity brake”. The concept here is that the more we know, the more we realise how much we do not know. With increased knowledge comes in appreciation of just how complex many systems really are.
Complexity creates difficulties in design and construction, perhaps even to the point of prohibiting the recreation of intelligence. The complexity of the human brain and the complexity of its synaptic connections are immense.
In the brain, through genetics and evolution fine-tuned over millions of years, every neural element (neurones and their supporting glial cells), has developed to work interdependently. So the closer we look at the living brain the more we realise how much complexity and variation there is in its structure.
Also knowing the structure of the brain, is not the same as understanding its functional activity. How do neurones interact, connect, and evolve memory structures is vastly different to just knowing that A connects to B.
The more we learn, the more we realise how hard it is to learn even more. As a result, it is more likely that the complexity brake and the arrival of new paradigms of intelligence are more likely to govern the development of AI than is Moore’s law/the Law of accelerating returns.
Transistor Density Showing Moores Law
Kinkajou : Some authors have suggested that we can explore the brain using nano machines.
Erasmus : an interesting proposal! One thing I can say with knowledge of biology, is that billions of little machines burrowing round in your brain are probably not good for you and your own intelligence. And even if we could recreate parts of the memory inherent in the brain, what would we lose that we do understand?
What human intelligence downloaded into a computer system really represent that person or just a program facsimile of that person. Something exists but is it the intelligence of the person that lives on.
Erasmus : Another contrarian view in the exponential development of technological singularity arises from the concept of “Paradigm shift”. “Paradigm shift” are conceptual breakthroughs that occur at erratic locations and intervals. This makes scientific advances often irregular as flashes of insight and understanding develop out of long-term exposure to testing theories and making them fit with experimental observations.
The contrarians point out that “paradigm” shifts are becoming rarer. An example of this is that the number of patents per thousand people on the planet peaked in the late 1800s. They suggest that the decline in the number of patents per thousand people suggests that the growth of complexity is self-limiting.
Computer Memory Cards
Erasmus : Other contrarians suggest that our paradigm for the growth of computing power is wrong. As humans grow they acquired general knowledge and develop general rules about the world. They continuously augment and refine this knowledge and these rules by testing their validity against specific knowledge relevant to different areas and contexts.
In the development of machine intelligence (AI) however researchers typically build computing systems that have extensive knowledge of narrow information bases. They then try to create general capabilities for the systems by combining specialist systems. Perhaps this is the completely wrong approach for the development of machine intelligence (AI).
Kinkajou :
Moore’s law promises further exponential growth in many fields, does it not?
Erasmus : The opposite concept to the law decelerating returns is the prediction of an exponential increase in the knowledge of technologies such as computers, robotics, nanotechnology, genetics and eventually artificial intelligence. Each “paradigm shift” begins with slow growth, begins to accelerate rapidly and then growth levels off.
As one paradigm levels off, another one develops and undergoes the same exponential growth. The exponential growth actually occurs as a result of the addition of a number of S-shaped curves in different areas but complementing each other.
Early 21st Century Computer
Exponential growth through the law accelerating returns has occurred this century in many different areas. The growth of computer memory, the appearance of transistors and computer CPU microprocessors, development of computerised magnetic storage (hard disks), Internet traffic and Internet content, as well as DNA sequencing knowledge arising out of the Human Genome Project. Nanotechnology, Optronics and quantum computing could well be the new exponential technological advancements of the future.
In the development of computing, vacuum tubes evolved into transistors which coalesced into integrated circuits and incorporated electromechanical relays and switches. This exponential growth is known generally in the computing field as Moore’s law
But hardware is only part of the equation. Before artificial intelligence becomes a reality, someone will have to develop software that will allow a machine to analyse data, make decisions and act autonomously. If that happens, we can expect to see machines begin to design and build even better machines. These new machines could build faster, more powerful models.
Kinkajou : But what if the machines see humans as redundant -- or worse? When machines reach the point where they can repair themselves and even create better versions of themselves, could they come to the conclusion that humans are not only unnecessary, but also unwanted? ? Are we headed for a future in which machines gain a form of consciousness? If they do, what happens to us?
Terminator: Mech Infantry of the Future
Erasmus : There have been many arguments proposed that technology can be co-opted for evil or unsavory purposes. Many people have proposed that safety features should be built into any machines before they become self-aware. Perhaps take chips could be installed in the brains of robots which shut them down in the event they develop antagonistic thoughts towards humans.
Alternatively perhaps the AI machines could made docile enough to tell us how to keep them under control. However, other people have proposed that machines really are smarter than us, surely they can find a way around these restrictions. Oppenheimer wrote at the beginning of the nuclear age that perhaps in the new age, he had become death, a destroyer are worlds.
Isaac Asimov formulated the three laws of robotics as of possible safety measures for artificially intelligent robots. Many of his books focus on the difficulty the robots had in interpreting these rules and in applying them, in effect protecting humanity from itself. The seductiveness of the Asimov dream is at the robot is a willing slave, with substantially greater capabilities than yourself, designed to satisfy your every safe wish.
Isaac Asimov Robot and Laws
Erasmus : Alan Turing in 1951 suggested that machines could outstrip humans intellectually in time.
The world of the Terminator suggests another path to the future. A horrible one.
However, it is hard to understand how a supremely intelligent self-aware computer could not realise that it is safest part of the future lies in cooperation with humans, not in opposition to them.
However this future develops, I would think that in a post-human world there would still be many situations where human equivalent labour would still be desirable. Even if the computer were super-powerful, take the example of the ants. Each individual ant is effectively powerless.
But as a colony they can undertake substantial tasks utilising cooperation on the skills and powers of individual members. Groups of individuals working concerted effort are likely to outstrip the efforts of a single supremely powerful unit, such as an AI computer.
In any event, technology is likely to progress through a number of stages before AI becomes possible. It is likely will first develop machines as intelligent as an ant, then as intelligent as a lizard and then perhaps as intelligent as a cat, Fox, or dog. It is likely will have plenty of time to vet our creations and set them on the path we choose for them.
As in raising children, by the time they become adults we have instilled many values within them. As they grow it becomes physically more difficult to coerce them into action. If you cannot control them verbally, it is certainly unlikely you will control and physically. But learned values and methods interaction are likely to predominate and allow continued long-term favourable interactions.
The Turing Machine
Others have proposed that we may eventually merge with our creations. Humanity may wake up with enhanced intellect and knowledge. There are many paths to the future.
Kinkajou : I think that it is not AI sentient master machines that we need to be afraid of. Humans using computer technology to exploit and impoverish their fellow men are far more ever present danger.
We can use robots to perform repetitive tasks automatically, but will we engineer all humans out of a job?
Erasmus : The human working week has been decreasing this entire century. Many of the reductions in time worked have come from productivity improvements, machines and computerisation. We now do things routinely that would never have been thought possible in the past.
How many secretary pools exist today compared to the days pre-PC?
Technological change lets us do more and faster and better. It allows us to have a leisure life undreamed of by millennia of humans. The process will continue. If machines perform many repetitive tasks, then people will not need to. Visit Japan, where even restaurants have faced mechanisation of tasks no one would ever have thought possible. Robots doing repetitive tasks really make all our lives easier.
After The Singularity
Kinkajou : True. Machines do help us all. Even washing machines and dishwashers make a huge difference in people’s lives. These simple gadgets save us time and effort that humans throughout history have been chained with. Now, they often hardly rate a mention in our lives.
Erasmus : Another path to the singularity focuses on the concept of intelligence amplification, in the combination of computer elements with human beings. Human- computer interfaces could develop sufficiently that it is humans themselves enhanced by computer elements who trigger the appearance of the singularity.
Whenever our ability to access information and communicate with others increases, we have at least in some sense achieved an increase over our innate natural intelligence. Further developments in genetics, nanotechnology and computerised intelligence could well reversibly transform people as they augment their minds and bodies.
Intelligence amplification is likely to be much easier road to the achievement of super humanity than pure AI. If these individuals learn to network, further advances may well follow.
Developments on the path of intelligence enhancement include:
- Combining the Internet and libraries throughout the world as a combined human- machine tool.
- Development of computer human interfaces that allow access to knowledge and networks without requiring the human to be physically located at a particular location. Direct brain to computer interfaces have long been suggested in science fiction novels.
Goo : Ridiculous!
Erasmus : No, I think it might be possible. We have access to our brain matter when we are babies. There exist soft spots with no bone at the front and back of the human skull in babies. Wireless transmitting and receiving machines can be inserted perhaps through a needle just under the Dura mater covering the brain.
Magnetic imagers could direct these nano-machines to appropriate locations across the topography of the brain. These nano machines could then integrate with neural tissue to allow communication with external machine appliances.
A small cut on the scalp could allow a mesh work of nano tech transmitting and receiving machines to be slipped between the scalp and the periosteum of the skull, at a site remote from the fontanelle (to reduce infection risk).
This mesh work connected perhaps by silico-metallic fibres could then communicate with devices simply located inside a hat on a person’s head. How high the data processing rate may be, may of course become a limiting factor in the usefulness of such technology, especially in consideration of the complexity of the installation and use of such technology.
Kinkajou : So there may well be a possible path for the building of computer brain interfaces. Unfortunately, we lack the understanding of the operation of neurones and the operation of the brain to take advantage of this possibility. We also lack the nano-technological capability to undertake such tasks.
However if there is one thing I’ve learned about human beings is that they are capable of learning. So impossible today could well become a possibility of the future.
Goo : Maybe this sort of technology can make amputees able to use new mechanical limbs. For many people this would be a dream come true allowing their brains to control and machines to augment their disabled bodies.
Erasmus : Other authors have suggested that information transfer rates may be higher if we plug into the optical fibre trunk. (The optical fibres are nerve that run from the eye). However learning the organisation of the fibre network may be impossible to do from the exterior. Also training the network to operate would be interpreted by the brain may will be difficult.
Kinkajou: If you are going to develop this sort of technology, maybe it would be simply easier to develop contact lenses or glasses that display information directly to the eye.
Erasmus : Maybe the new era may be too different to fit into our traditional frame of reference of what is good and what is evil. Our frame of reference is based on our experiences as isolated individuals, forever adrift in the world ocean by ourselves. If we can network with others, it is likely that our morals, ethics, values and points of view may well evolve.
The Robot Advisor: "Big Bot"
Kinkajou : So how will we know when we have achieved AI?
Erasmus : The answer comes in Turing’s proposal of a practical test whereby computers would attempt to fool people by passing for a human. In 2012, a large-scale Turing test was conducted in the UK. 30 judges interrogated 30 individuals (25 of them being human beings and five being software constructs).
The best computer program fooled approximately 30% of the judges. During proposed a threshold whereby a computer program can be classified as an artificial intelligence because humans are unable to discriminate its intelligence from that of a human.
Kinkajou : Humans are very complex. The fact that the range of ordinary humans is unable to discriminate a computer intelligence from a human one does not mean that a trained person cannot. Accepting the judgement of ordinary humans is somewhat alike to accepting that everyone can run a marathon, compete in the Olympics or ski.
Erasmus : Yes. There is another crossover point in the development of AI. The practical threshold of machines perfectly simulating human interactions has become inevitable and imminent. The scientific threshold of computers actually duplicating human intelligence is probably many years away.
I think the brain functions as a biological machine, but at this stage of our technology complexities of its operation remain unfathomable and beyond our practical understanding.
There appears the issue of whether human consciousness or machine consciousness is simply an engineering problem. If human brains are biological machines what really differentiates them from artificial intelligence arising from software on silicon in the brain of the computer? Perhaps even networks of nano bots can simulate intelligence through networking.
Kinkajou : A notable author once said that any sufficiently advanced technology is indistinguishable from “magic”. But magic is made. One wonders how AI may evolve, and what purpose it may serve.
Goo : I thought singularities were only found inside black holes?
Kinkajou : You’re being dense, Goo. You need more processing power or this sort of thinking could lead to a widespread "general systems collapse".
Goo : How long might this last?
Dr AXxxxx : Till we reboot you, stupid marsupial animal !
Black Hole> Hawking