Grand Challenges for Engineering - May 23, 2013
Combining artificial intelligence with robotics could revolutionize medical treatments and factories. How closely will these combinations be able to approximate humans? What other areas could benefit by such advances?
I think it would be wise to investigate the possibilities of creating a booster for the part of the brain that controls the heart, as opposed to installing an artificial heart beating device.
Quantum mechanics does indeed play a role in biological processes. Virtually all of today's computers avoid that randomness however, using high voltage levels to ensure that all detected electron pulses are intentionally sent. Low voltage/energy, high efficiency, slightly innaccurate (on the order of 2% max error, according to the article I heard of them in) computer chips exist, but these are relatively new and not widespread.
Also, as others have mentioned, most computer processors can only do one calculation at a time, though at an incredible speed. The brain does many thousands, even millions, of 'calculations', or neural impulses, simultaneously. True, we are limited to single digits for conscious thoughts/actions, but there are a great many unconscious, automatic processes necessary to keep you alive, functioning, sensing, percieving and processing the flood of sensory input, etc. Each requires the cooperation of many neurons. Very probably conciousness/awareness, and moreover selfawareness/sentience requires constant communication between large numbers of neurons.
As to the question "How close can machines come to human intelligence?", the predicted date for comuputers capable of uploading and/or functioning as the human brain is around 2050. Whether they will be sentient remains to be seen--the physical origins of the "mind" are even less understood than the functions of the "brain."
One more thing: the identifiable storage capacity, or available memory space, of the human brain is about 2 Petabytes, or 2000 TB.
I run www.techdimwit.com and work with BCI systems in terms of typing and to operate a robotic arm using the P300 based EEG systems available today. This is based on exponential growth and the law of accelerating returns, we will reverse engineer the brain and recreate neocortical columns to work as portions of the brain and then run them in real time and upload skills and build human intelligence based upon our structure and it will grow and surpass human intelligence then continue to do so more efficiently in accordance with the laws of physics. this is the most important project in humanity and in the 21st century we will soon then upload human thoughts and ideas into supercomputers and build upon them and merge with the machines and augment our intelligence and parts of our brains with this technology.
Even the term "systems" fails to take into account that biological communication and specifically neurological effects are many process's and events that happen in an apparent lack of, synchronization of all factors and indeed require different aspects of even methods of communication at each "phase". To simply build a neuron cluster even many of them and connect them as if they are just neuro-chemical's that produce EMF echo's as just artifacts, if we simply approach as neuro-synthetic "circuits" I don't think it is likely to work. System's endure, Organisms adapt, and seem to "love it". You do have at least a 3.5 billion year head start on any of our designed systems...
There is no distinction between HW&SW in the brain. Both the geo-topology of neurons physically change indeed "morph" and thats just for starters. What happens in neuron cluster's may resemble and does look very linear, but evokes change in higher dimension non-linear "areas", to regions to global-brain, by encoded (fractal) ? But with stochastic forcing, this scalable-information may transfer to a very low energy-information "pulse", that may be coded as EMF and "other' carriers of change. This may be a soliton wave that requires almost no energy to maintain all need information charge the "state of the destination" areas. And we haven't gotten into what may be most important about the brains low energy, ultra-redundent nature: quantum information transfer.
Add biological systems self assemble and correct. We have detected quantum events in photosynthesis: tunneling, bi-locality at the minimum. Certainly any real hope for "AI" has got get over the "circuit like view". Take advantage with what biological and abstract thought are based on. It works.
An example of little energy that transfers information is "the wave" we all know at sports stadiums. A trigger of, not energy exchange except as information. And in its own way "the wave" is pleasantly infectious.
As long as technology essentially mimics the square-based spreadsheet (because the base level of computer switching is on/off), we will not be able to understand the full methodology of human brain function. The next logical step to structural simulation is necessarily the hexagon, yet no one bothers to look toward the benefits of a hexagon structure in neurology before deciding the concept isn't sufficiently technological for groundbreaking ideation. This mental state of pre-closure would have prevented the discovery of penecillin, yet technology adheres to it like superglue.
Reverse-engineer the brain
The intersection of engineering and neuroscience promises great advances in health care, manufacturing, and communication.
For your information; A brain is part of a body and it did not have any direct access to an environment. By that reason it cannot process any information about causes in the World.
There are no facts for support commonly spread opinion about existence of mental functions.
Modeling of a brain will not provide a clue on how it works because it works in conjunction with a body.
I knew how to develop an artificial subjective system capable to demonstrate an intelligent behavior without wasting funds on such fruitless projects, or spending a billions on the research within the AI paradigm. Semiotics is another example of current research in style, which is based on the misleading assumptions about an functional and meaningful links among the words in a message.
The truce is in the simple conclusions: Meaning is in the behavior of an recipient, not in a message of any kind.
Best regards, Michael
Is reverse engineering the brain done on human beings without their consent? Are human beings tortured and violated to develop Artificial Intelligence? Are people chipped against their will so that you people can test your dreams?
How many people have you saved? You seem to be good at murdering people, and not the other way around.
One of the behaviors of human thinking is its nature of growth and improvement ,in a greatest amount and speed, to fit its environment. Machines are static.They can no longer run than what has been dictated inside them. 10Q.
R&D in AI mostly have illusions as a basis. For the beginning, why one have to reverse engineering of a brain to have a system capable to demonstrate a reasonable behavior? Biological brains did not have access to the World. Because of that a brain can not process the information about the events in the World. On another hand, a brain is dynamic system, with a structure reflecting the history of the body, not a history of the World. The tempo of developing of the new synapses in a live brain is pretty high. As it was discovered recently, from 5000 to 8000 synapses are created every second. It is no reason in further discussing. Reverse engineering of a brain is practically impossible. It is possible to have an artificial creative system? There is no law of the Nature that could prohibited such development, but its should be provided on the factual basis. What we could see today? Unsupported by facts statements are used as a theoretical basis for many projects. Who, for example, could provide a factual support for believes in existent of any so called mental functions? I could lead the team of professionals in development of an artificial subjective system capable to demonstrate a reasonable behavior. Such systems could be used for various purposes. In health care they could be used as a Robonurses, for example. Best regards, Michael
Thank you for the great information on scieintific breakthroughts. The idea of reverse- engineering the brain is due to be a wonderful achievement, but I believe that scientists may not be looking in the right direction, not learning from the existent, as said in the article. Why don't scientists find a way to preserve the brain, transplant it into a robot, and allow it to continue life from one who is deceased? Yes, this may sound extremely science fictional, but bear with me. Then, introduce the robot with a human-brain to a robot with a minor artificial intelligence wiring system. Allow each to speak to each and let the AI robot learn from the human-brained robot. Yes, the task of supplying blood, oxygen, fluids, and an appropiate skull is tough, but we also thought that uncovering the human genome was impossible. Also, there are rumors floating around that the reverse- engineering of brains has been accomplished in a privately-owned lab here in the US; I cannot wait to hear their findings. To add to my idea, how can you conceptualize the idea of thinking to a machine which must first conceptualize thinking. I am not skeptic of the machine becoming smarter than a creator (human), but smart on its own. I mean as in thinking or acting upon its on thought train. Who knows if a sci-fi re-enactment of I-Robot will occur? But, we must first study the brain much, much more to merely adhere the thought of reverse- engineering the brain to our agenda.
Re comment 1, no "philosopher" ever reached a definite conclusion (or even one that is likely to be roughly correct). Read our very short paper on neurally-based intelligence at http://ine-web.org/filead min/templates/_docs/NME3- 1Final.pdf We argue that the key to intelligence is simply accurate adjustment of individual synapses. Since we have a quadrillion of these inside our skull, this is not trivial. We think the circuitry of the cerebral cortex can pull off this amazing trick. I am skeptical any currently proposed machine could achieve this, but perhaps I am wrong.
For philosophical reasons, i dont think its possible for a creation to be smarter than its creator.... (don't confuse faster with smarter)
One Question to the headline (How close can machines come to human intelligence?) As far as I am informed, no one claims to know how the brain works (parts of the brain, Neurons, transmitter and what have you - sure, but....not much more than that, right?). So if we don't know how the brain works how can we say that something which we don't understand is equal to a machine et vice versae? Next question: How closely will these combinations be able to approximate humans? I guess, AI can support a doctor - sure why not? My car supports me when I want to get from point A to point B. Therefore, thank you so much for your interesting questions.
Perhaps it is possible that AI developement can help to improve a numerous variety of fields. Thinking only about the medical field is far too small. The creation of AI robots could lead to a global revolution. It may, one day, be the downfall of our necessary evil, money. Robots can be created to work for our desires, and the science fiction thought that they could take over is absurd. If you view this situation maturely there is no way for that to happen if we were to create them to "our" desire
I think that its very much possible to create robots looking and working just like humans.What's the dark side to this invention is should also be taken into consideration.The right questions should be asked, like what if goes against us?What if this tries to outpace the humans itself.Some things look very interesting and intriguing when we try to build them , but sometimes we might have to regret for it also.So nothing of this sort should be done that is counterproductive.A solution should be looked from both the angles We shouldn't take a myopic view while carrying on such kind of huge experiments.While this was the technological impact of the inventions but looking an from "PEST" viewpoint is very important.This was a critical analysis of the whole idea.If we think that yes its verified from every corner and each aspect of it has been looked into then we should go for such big inventions. The impact on all these four environments could mean a lot to humanity.
Technology has already been developed that not only matches, but also exceeds human decision-making ability when the design is constrained within certain predefined conditions. However, it is difficult to predict whether or not other aspects of human "thought" can be captured in a non-biological machine. Many processes, including emotions, memory, and learning are not simple to quantify or physiologically understand. One possible restraint to the development of an adequate biomimetic system may include our desire to have a single logical answer for any given situation. Considering current practical uses of technology in manufacturing and computing, it makes sense that man-made devices are precise and consistent. Relaxing this restriction and incorporating mores stochastic elements into design will inevitably create circumstances that could potentially cause delay or failure because a device simply "decided" to do something different. After all, failure is an extremely important part of the learning process. If a machine were developed that perfectly mimics these human characteristics, the usefulness of such a "creative" device must be assessed. Moreover, devices like these may cause society to reconsider what is meant when it speaks of "consciousness".
"Human intelligence is superb, and until we are able to fully understand how are brain works we won't be able to give machine anything close to human brain. "
Human intelligence is superb, and until we are able to fully understand how are brain works we won't be able to give machine anything close to human brain. Human brain is more adaptive, robust, quick decision making than any machine. To make a machine as intelligent as Human brain it need to have a learning skill process, take a kid for example, they learn new things so quickly because they can have high learning cycle.
Research in this field is without doubt very intriguing and interesting. To think a machine doing everything that a human can, even emulating human emotions, is truely fascinating. Despite all the movies that we see nowadays about robots going bad and destroying the earth, if the brain can really be understood and a really human like robot is made, it will definately bring enormous benefits to mankind, some which has already been mentioned in this sight. The primary focus of the research should be such that the reverse engineering enables the enhancement of the medical benefits to human beings. For instance if a blind man can be given his eye sight back or if a deaf and dumb person can communicate with others. I believe reverse engineering will be enormously fruitful, albeit the time that will still be required for it to be fully implemented
"To truly reverse engineer the brain is to find that it, like all phenomena, cannot be found to have an independent nature."
To truly reverse engineer the brain is to find that it, like all phenomena, cannot be found to have an independent nature. A fundamental shift must occur in the way we think about our world for there to be a complete reverse engineering of the brain.
I can't say for sure that robots will ever be able to do everything a human can do but it is possible. People say that robots can not preform emotion. That is false. Robots can easily be happy sad angry or annoyed that is if we program it to do so. For instance, if you program a robot to not be good at somthing like opening a bottle and you program him to do the body language of a physical human and use the speech of the physical human when he does not do what you tell him to do right it is personifying anger because when humans are angry they tend to use there hands and mouth more than anything else. I think that there will always however be somthing a human can do that a robot can't...
Current research should be focused upon emulating a certain function performed by us humans rather than trying to create an entire human robot. This strategy should be promoted as the matter at hand, the engineering of the brain, is a complex one. Copying the brain entirely the way it is might not be impossible, but it is time consuming for sure. For quicker results the folowing startegy should be used: 1. Select the problem to be tackled. 2. Evaluate the feasibilty of using AI in place of NI for the particular problem. 3. If AI selected, narrow down on the specific functions that need to be emulated. 4. Reverse engineer accordingly. 5. Implement results through a pilot project. 6. Explore expansion opportunities for the researched techniques. The strategy can be used to create a library of AI tools that can be integrated through time to handle more intricate tasks.
There is precious little available to practice neurosurgical techniques outside of a real patient in a no-risk setting. Simulation-based education for healthcare could greatly benefit from advances in artificial intelligence in addition to virtual reality.
Intelligence per se can be better understood by attempting to re-create intelligent beings. At an extreme I envision that intelligence could supercede current cognitive activities to embrace mystical qualities. I "define" what we call intelligence today as - The ability to work efficiently in areas that have complex, incomplete and ambiguous patterns. To enable intelligence we need a system that is both sub-symbolic (neurons based) and symbolic (higher level of processing, more analytical). We need a system that will easily integrate processing sensory signals with the processing of experiences and creative thinking. To do this I believe that we have to first identify what I call the Most Primitive Conceptual Entities of the world. I believe that somehow these are implanted in us human beings at birth. We are then capable of developing these Primitives into the conceptual primitives for a particular domain (like mechanical engineering or medicine) and subsequently learn the more advanced topics in those areas. In a similar manner we have to first identify these Most Primitive Conceptual Entities and then represent them 'adequately' in the system and then provide the right learning mechanisms to build cognitive structures for a domain, from them. This work would help in identifying the reasons for cognitive impairment in children and provide the option of implanting, nurturing the growth of the required primitives that will eventually remove the impairments.
This challenge is very closely related to one of the UKCRC's Computing Research grand challenges summarised here: http://www.ukcrc.org.uk/g rand_challenges One of those grand challenges is particularly close, namely GC5: "Architecture of Brain and Mind: Integrating high level cognitive processes with brain mechanisms and functions in a working robot". It is described here: http://www.cs.bham.ac.uk/ research/projects/cogaff/ gc/ I have elaborated on the importance of the more general study of biological information-processing (not only in brains, e.g. development of an embryo uses chemical/molecuar information processing) as a major engineering challenge here: http://www.cs.bham.ac.uk/ research/projects/cogaff/ misc/synthetic-biology.html Aaron Sloman http://www.cs.bham.ac.uk/ ~axs/
first of all, I would like to thank Yu Chou for his wonderful remark. and to my opinion, Human brain is the most complex architecture in existence. But understanding it and implement the understanding to design newer intelligent machine can really contribute. Like in the field of AI and machine learning. we cannot totally imitate it, but we must try as close as possible to get better approximation and data interpretation. And I personally feel very interested in this subject .
By far the most important goal in your list in my opinion. But the question one is really trying to answer is what is knowledge and how it is organized--not how to reverse engineer the brain. Certainly the brain is one way in which this is done. However, arguably the brain does this poorly--unable to handle more than 7-8 variables at a time, inaccurate memory recall, yet it is still much more usable than anything we can program up in silicon. I find the naming of the computer science field--artificial intelligence--amusing since any "bird-brain" (e.g. eagles) biological system can visualize and utilize sensory input much better than anything we can program up. Call that intelligence seems rather a hyperbole. Let's just say that there appears to be some very efficient algorithm done with very slow biological hardware with just a few tens of thousands of simple switches (the visual systems of birds) that we have no clues to. The visual task is low level and arguably devoid of any real "intelligence". Once the data is "recognized" by the visual system, then the real "intelligence" part of organizing the data into knowledge begins. Our problems here are several. The first of which is a lack of algorithms for simple tasks--going to more involved depth of simulating brains is not going to solve it. At bottom, no one will suggest that you can understand how computers calculate the weather by simulating the transistors of a CPU. Saying that the answer is to just simulate the hardware is an admission that we are clueless. For that matter, the airplane analogy is correct. It is only because we don't have the first clue about how to make an air foil that there is suggestions that making a bird would be easier and more direct. We need better ideas here--not necessarily more money to make bigger computers. But coming up with good algorithms is hard, asking NSF for more money to make a bigger computer is easier. Finally, perhaps before one goes on with this project, somebody actually defines knowledge and intelligence first. Most rely on circular logic or equating humans with intelligence (Turing Test for instance).
I am not a programmer or engineer, but I have an intense interest in this field and i don't really know why. But here is my 2 cents on this matter: Asking questions seem to be a simple and fundamental part of being human, yet i have not read any articles saying that any computer has yet asked questions. So, can we program this ability to ask questions into computers?
I do research in computer vision computer programming, which is one area of AI, and which utilizes findings from brain research. I currently have the opinion that part of consciousness is a fundamental characteristic of living matter that is not conducive to mathematics and computer implementation, and other aspects of the brain are more mechanical and conducive to mathematics, so that they are implementable in a computer. For example, the perception of a color like green seems to be a fundamental aspect of living matter. We all know what the color green looks like, but physics and mathematics can only describes the color green as a wave of certain nano meters. Does a photo diode "see" green as we see green instead of just sending out a voltage? If not, there does not appear to be a way to program a computer to see the color green as we see it. Image recognition may have aspects that can and cannot be programed or reduced to mathematics, although a greater part can probably be treated by mathematics, and thus programmable in a computer. Multi-celled organisms evolved relatively recently in Earth's history as oxygen was built up by photo-bacteria. The human brain while complex in terms of the number of nerve connections is probably to be relatively less complicated in terms of the number of brain structures, as it evolved in the recent history of the Earth. Many of the capabilities exhibited by the brain have been simulated by the computer like printed character recognition, speech recognition, simple decision making as well as less perfect simulations of face recognition, free writing recognition, music composition, and general image analysis.
One great contribution to human civilization of deciphering the workings of the human brain is to figure out how to detect when the brain is lying and not telling the truth. A fool proof lie detection method will greatly promote world peace in the modern world in my opinion. There are great distrusts amongst people of the world, and they are building great arsenals including nuclear weapons in a great part due to this distrust. Almost every country say that they want peace, but few fully believe these sayings, as there are no ways to determine whether these saying are lies. While world peace is much more than just about controlling lying by countries, such understanding of the brain can potentially improve world peace.
Greetings, I applaud your goal of "reverse engineering" the human brain. However, I doubt a complete working (hardware) prototype would yield much benefit. I'm a computer programmer (award winning) and have "programmed" many models and varieties of computing machines. It's (sometimes) painfully obvious that the "hardware" of the machine bears little upon the outward characteristics of it; quite the contrary in fact. It's the software, the programming that defines what it does, the "firmware" that defines the interface to the hardwares "how". The actual processor that does the work? It's essentially a "media". Virtually any algorithm can be executed on virtually any processor; again, processors are just media for algorithms. It's the algorithms that are the key to intelligence, algorithms that likely can be run on a variety of media, not just the human brain. And the algorithms running now, enabling you to read this, have been developed over hundreds of millions of years, steadily evolving, improving themselves generation after generation. However, all of our attempts at simulating intelligence have fallen woefully short. We don't even have a decent model of the human intellect. Machine Intelligence has proven as difficult to grasp as sunlight itself. How does one go about grabbing sunlight? Perhaps there are clues, clues in our own folklore and history. For instance, there's an old psychology saying which (I believe) provides some salient insight into "intelligence": It is: "All of our decisions are simply designed to please ourselves". Cute psychological device, or the First Rule of AI? I would suppose the latter. Think about it.
The major issue here is the ultra-parallel nature of the cortex and the major submodules like the cerebellum, thalamus and spinal cord. Over millenia there has been biological engineering of subprocesses through evolution that we may not have the insight to reverse engineer.
Fusion nuclear for all People
I think there is an issue missing in the "What is needed to reverse engineer the brain?" section: intracellular processes. Learning and memory are hinted at in the above sections, but not discussed in this section. The key feature of neurons that makes them different than transistors is that they are plastic: they change their behavior over time. Although this may involve some rewiring, a large portion of this plasticity is due to changes in synaptic weight caused by intracellular processes. Some known work involves long term potentiation and depression via glutamatergic channels. Other work has shown that receptor activation can set off intracellular signalling cascades that may switch between bi- or tri-stable states. Finally, further work shows that receptor activation by neurotransmitters can initiate gene expression changes which result it differing amounts of intracellular and membrane-bound proteins. These three mechanisms may be the heart of "learning" and are not discussed at all. As for "How close can machines come to humans?", I believe it is more a question of "when" rather than "how close". Ray Kurzweil' book "The age of spiritual machines" predicts that computer power, if it continues to grow at the rate it is growing, will have the capacity to do as many calculations as the human brain by the year 2020 in a machine that costs $1000. The problem will then become, "how do we make this machine that is as computationally powerful as the brain act like a brain". I think that may take another century to figure it out completely, but in the meantime computers will become smarter and smarter. We will probably see individual systems being replicated before that time, and we have some of those in crude form today: speech recognition, optical character recognition (recognizing handwriting), feature discriminators (used by military to find camouflaged enemy artillery), etc.
AI will be key in helping us design our next generations of toolsets as humanity is approaching its limits as to what it can design due to the scale, speeds and complexities involved at the frontier of nearly every science. Evolutionary algorithms, machine learning and computational theory will increasingly play key roles in nanotechnology, biological complexity, quantum physics and the many simulations and virtual worlds we will design in the coming centuries.
After the machines have replaced all those labor-intensive works, they will come into the field of design and analysis aspect of works. But the fact that machines combined with higher and higher level of artificial intelligence should always be treated carefully.
Defining "thinking" is the most important. Whether thinking is a deterministic or random process is a difficult question.Or it is indeed a single process with million or billion Monte-Carlo step? Moreover, it is not a 0 and 1 process. The ouput(s) may depend on the energy level or chemical composition of the input(s). I always suspect that even one can mimic a single brain cell, there shoud be a critical number of cells or connections before it works as a "thinking" brain.
Why would you condemn a species to slavery before it even exists?