Wednesday, 13 November 2013

End of TWC module

In a flash, the last TWC lecture is over, and so does the high workload from this module. While it seemed to be a burden at first, as the weeks passed, I noticed reviewing my previous postings on the past lectures did play a key role in reviewing what I had learnt. Writing these thoughts down has placed endless questions in my mind about technology, sparking my further interest in this limitless subject.

While the module seems to be on the side of science, it indeed does have a deeper message within as the same key points about taking the initiative rather than waiting around were being reiterated in every lecture. The lessons I learnt from this module have shaped a more curious and open mindset for me, and it will indeed prove to be useful in my future modules at SMU.

TWC Week 13 (Session 12)


This week's lecture was on presentations again and was the last in this TWC module.

One of the key ideas and concept that captured my attention was during the presentation on the answer to world hunger. While there are many future solutions to solving the issue of food supply, it seems that many of these solutions have opened new doors for large corporations to step and take advantage of the market situation. While GM food has brought about numerous benefits, the bullying behaviour of Monsanto and the the unclear health risks of such modified food cannot be ignored. It seems that for food, there is indeed enough food to go around, and the problem lies in its distribution. Hence progress in technology in this field may not be the key to solving this problem.

A key take away point from the presentations on trans-humanism and from my group's research on mind-controlled prosthetics would be that technology is indeed progressing at a very fast pace, but at the price of opening many doors to the unknown. With creations like mind-controlled exoskeletons and prosthetic limbs that are more powerful than the original human body itself, it seems likely that the risk of misuse in the future is high. As such it would be wise to start developing an ethical and legal framework on their use before such equipment become widespread around the world.

An issue for further discussion would probably be on the rich-poor divide. How much further can this social divide be pushed with increasingly powerful and efficient inventions from the 'richer' countries while the poor lag behind even further in the technology race?

8/10

Sunday, 10 November 2013

TWC Week 12 (Session 11)

This week's lecture was on web presentations by 3 groups.

One of the interesting observations I made was during the presentation about futuristic buildings, where ideas were drawn from the past, with one of the key drivers being adapting the house to natural disaster zones like tornadoes and earthquakes. Another interesting concept was from the presentation on nanotechnology where contrary to the popular mind set that progress means bigger and faster, technology could also progress in the form of becoming smaller.

One of the key takeaways from the lecture were the importance of analyzing current trends as shown in last week's lecture, as it has an enormous influence on how we shape our future. Another would be to always keep sustainability in mind as seen in the presentation on renewable energy.


7/10

Wednesday, 6 November 2013

TWC Individual Paper (Final)



Artificial Intelligence[1]

Executive Summary

With the recent rapid development in artificial intelligence, future implications and ethical concerns are being considered and definite guidelines have to be drawn to prevent such technology from becoming developing into something beyond our control.
Firstly, this paper documents the dawn of artificial intelligence and shows how far it has developed till today. It next examines the current situation and level of artificial intelligence as well as their impact on the world now. Finally, it focuses on the important future considerations and its enormous influence on the future especially if these serious implications and dangers are not addressed early.

Introduction

Artificial intelligence (A.I.) is a wide field with different definitions but all of which involve the science of creating machines with intelligent behaviour (AISB, n.d.). The general basis of A.I. is the assumption that the human brain’s thoughts and activity can be explained and replicated in mechanical terms and reasoning. A normal human brain is estimated to have around 100 billion nerve cells. However it is the neural connections between these cells known as synapses, which enable our brain to think and function, and they number over several hundreds of trillions (Fry, 2011). As such, this places the power of the human brain far beyond the reach of any supercomputer in existence today, with the potential capabilities of the human memory still unchartered (Reber, 2010). Mankind has crossed numerous new frontiers in endless fields, creating wonderful feats of technology and engineering, as well as exploring the boundaries of outer space. However, we still have not managed to fully understand something much closer to us, the human brain. (Walsh, 2013) As we will see, mankind has been attempting to explain and replicate the human brain function mechanically throughout the ages. While it can be said that while we are still ultimately unsuccessful, the development of A.I. has brought about positive changes to our lives in terms of convenience, efficiency, minimising safety risks and errors as well as endless other benefits ().
Firstly this paper seeks to document the rise of A.I. before moving on to the current situation and future considerations. Through detailed examining of the current state of A.I., clues and signs can be gathered about the direction of where A.I. would be heading towards. It is evident from the current situations, some of the potential hazards and implications we could be facing in the future. The numerous concerns about where we could be heading to in the future with regards to A.I. are clearly not unfounded and will be analysed to seek out potential solutions to solve them.
With regards to limitations of this paper, the key factor would be the lack of any real world examples or case studies to substantiate the future complications as the possibility of such scenarios happening would be decades away in the future at least. Another issue would be the need for endless assumptions and specialized expertise in A.I. to draw up precise and exact solutions to answer the problems faced in future considerations. As such, my thoughts and analysis will be made from studying of the current situation as well as from the theories of other researchers in this field.

Historical Perspective

We will never know for sure where the idea of A.I. originated from. Stories of intelligent robots abound in historical myths of different cultures, ranging from the golden robots of Hephaestus in Ancient Greece, to Yan Zi’s automaton in Ancient China (Grachev, 2006). But what we do know for sure is that in the 1950s, several milestone events which led to the birth of A.I. as we know it today occurred. Alan Turing’s published paper on the possibilities of machines as intelligent as us, together with A.I. research being accepted as an academic discipline led to the Dartmouth Conference in 1956 (Turing, 1950). Many leading researchers in various fields were gathered and set the guidelines and foundation for the development of A.I. for many decades to come (Stewart, n.d.). This also sparked government interest over the next two decades which led to extensive funding in A.I. research, especially on machine translation (Slocum, 1985). As such, A.I. developed at a rapid pace with many programs and algorithms developed to solve problems as well as replicate isolated areas of human thinking (“Stottler Henke”, n.d.).
However in the 1970s, government reports were showing that the results of artificial research were not living up to the great expectations promised and as such, led to a heavy cut in government funding. This led to a period of time known as the “A.I. winter”. It also exposed the lack of genuine understanding and underestimation in how A.I. works, as programmers could only create machines with highly limited functions rather than the intelligent thinking machines that were expected (Smith, McGuire, Huang, Yang, 2006). Overall, progress in artificial intelligence was heavily affected and did not recover until the 1990s.
Fast forward to today, it is impossible for one to ignore the huge leaps in technological advancement with regards to A.I. While we are still nowhere near creating anything that can fully replicate a human brain, robots have been created that have some form of autonomous function. (Veloso, 2013) We are also rapidly entering a new frontier littered with many potential problems and concerns which will be examined in this paper.

Current Situation

The level of A.I. today is a far cry from that of in the 1960s-70s. No longer are machines limited to mathematical functions and the like. A.I. has now become more widely common in our everyday lives and have become increasingly autonomous functions. Various machines and robots have also been created to exceed the level of the human brain in limited areas (“Artificial Intelligence”, n.d.).
An example of such machines would be in the game of chess. Back in the 1960s and 1970s, the idea of a computer beating one of the world’s best chess players was seemingly unthinkable. That all changed when IBM destroyed that notion in May 1997, with an improved version of its original Deep Blue, beating arguably the greatest player in the history of chess, Gary Kasparov in a 6 game match (“Deep Blue”, n.d.). Today’s chess programs have evolved from the complex hardware Deep Blue had, which required it to evaluate every single move possible at each point in time, into chess software, resulting in much more efficient processing and thinking. The strongest chess engine in existence today, Houdini has an inbuilt ‘instinct’ which allows it to select only potentially good moves for further examining, while merely glancing over weaker moves, making it much more efficient and faster than previous programs (Chessbase, 2012). Such an ability has been only previously been capable of by human players, and for a computer program to possess such a skill represents a huge step up for the level of A.I. This also opens more possibilities as to how A.I. could grow from here with evidence that instinct can be programmed in as well. It may soon become a question not of how, but when would robots become human-like in thought.
Today’s increasingly autonomous A.I. can be seen in the example of Google’s driverless cars. Extensive testing has shown the driverless car to be accident free at over 500 000km (Guizzo, 2011). One may argue that such technology is not new as aeroplanes currently have an autopilot feature where the plane can be flown automatically with minimal interference from the pilot. However the autopilot in aeroplanes today is merely to help the pilot perform his tasks, rather than take over his responsibility. (Kramer, 2013) It is also incapable of more complex tasks like taking off and landing. What makes the driverless car amazing is the environment of which the autonomous machine is operating in. The different sensors of the car working seamlessly together without any human interference to drive the car along crowded streets and highways, successfully avoiding walking pedestrians and moving cars is definitely a major engineering feat. Such a car is still far from being commercially available, with many liability and legal issues still being unanswered (Hars, 2013). Such a level of autonomy in A.I. today demonstrates that complex machines and mechanisms can be programmed to be autonomous despite being in a challenging and ever changing environment. Once again, such an event further shows the magnitude of possibilities for A.I. in the future.
A.I. also plays an increasingly important role today as seen in the military. Armies around the world are constantly seeking new ways to reduce the risk of soldiers being killed in battle and have turned to A.I. for alternative ways of warfare (Gaudin, 2013).  An example would be the Samsung machine gun robot developed in South Korea. Developed for use along the Demilitarized Zone between North and South Korea, the robot is capable of performing sentry duties, with its sensors being able to detect human movements and challenge intruders through audio or video communications. Though the robot would still be under the control of a human, it does have an automatic mode where it can make the decision to fire at an intruder when he is unable to provide an access code (Global Security.org, 2011). Today, the possibilities of Lethal Autonomous Robots are being extensively researched into, with ideas of a robot having full decision of who to kill without any human interference being developed. The technology for such a machine does exist currently, as seen in the increasingly autonomous drones being manufactured. Although there is an element of human control for all autonomous robots currently, the possibility of a fully autonomous killing robot being developed in the near future is highly likely (Pilkington, 2013). Such technological advancements not only have the potential to change the mode and tactics of warfare forever, but it also has the ability to limit the loss of human lives by reducing the need for actual human soldiers. While the potential benefits of such robots are enormous, the potential risks of these autonomous robots operating out of our control cannot be ignored, and will be discussed later under future considerations.
A.I. has also been developed to a point where it can replicate human emotions and feelings to a limited degree. In the past, films like Blade Runner could only portray our speculations about the possibilities of robots with human-like emotions, through its story about a group of robots who have escaped to Earth from off-world colonies in an attempt to prolong their lifespans. However, in 2010, the first robot that could develop emotions and form bonds with humans was unveiled, being modelled on the attachment process that chimpanzee infants and human babies have (BBC, 2010). Another important milestone would be the development of robots that could evolve their programming code and learn the ability to lie in controlled experiments (Christensen, 2009). With an increased understanding of A.I., humans have begun to make progress in replicating basic human emotions and feelings as well as moving on to more complex ones. Though the range of emotions and feelings portrayed are still very limited as compared to those of a human, such progress is definitely fascinating as it leaves us to wonder if robots in the future can one day be ‘human-like’. The possibilities are endless but so are the future implications and dangers, which will be discussed next. 

Future Considerations

With A.I. developing at a rapid pace, we are moving into a new age faced with many ethical concerns and dangers. Disastrous scenarios have been portrayed in many movies like The Terminator series as well as I, Robot. Many scientists and researchers have also warned about the potential implications and dangers of a developed A.I.
The first issue would be the danger of a technological singularity, a scenario when A.I. would have evolved past the point of human intelligence to dominate human civilization and human nature itself. Numerous A.I. researchers have argued that there is no motivation as well as being mathematically impossible for a highly developed A.I. to maintain any ‘friendliness’ with the human race. As such, this places them as a potential threat to mankind (Keiper, A. & A.N. Schulman, 2011). While the possibility of such advanced robots dominating the human race seems ages away, the seeds of such technology have already been planted as shown in the earlier examples of increasingly autonomous robots in the military. Different proposals to prevent such a disaster from occurring exist. One side of the argument emphasises the need to base all future A.I. developmental goals on a friendly relationship with humans, so as to ensure that the machines will never turn on us. However the level of engineering and coding to install such a behavioural pattern is beyond the limits of technology today and even anywhere in the near future, which makes it an unviable option at the moment. Less pessimistic proposals include the extreme view which concludes that the level of A.I. today is sufficient, and trying to improve it will lead us on an inevitable road to a technological singularity (Goertzel, 2011). With such proposals looking seemingly unfeasible at the moment, it would be best to take a more neutral approach, with development to continue as it is now, and having us humans keeping a close watch on progress to reduce the chances of such a singularity occurring. A good example would be the United Nations calling for a halt on further research and development of lethal autonomous robots until international guidelines on their usage can be decided upon (Bloomfield, 2013).
Another key issue would be with regards to the liability of the robot. A.I. has been developed to such a stage today where robots are increasingly responsible for doing highly manual and routinized jobs of humans as seen in the commercial sectors, leading the question of who is to be responsible for any breaches of the law committed by them. With such increasing responsibility, it is highly likely that a robot would be placed in charge of a life and death situation in the future. For such difficult questions, once again, a wide variety of opinions do exist. While there are calls for detailed logs for autonomous system to explain the robot’s reasoning, such features will no doubt place a limit on the level of the decision making systems (The Economist, 2012). Applying ethical systems to such A.I. seems to be the best way forward, but with the sheer difficulty of transferring such guidelines into programming systems, it stands to say that such programming would not exist anytime soon. Overall, such technological developments are definitely without question, for the better of mankind, but it would be necessary to answer such critical questions and set strict rules and guidelines in place before embarking on more ambitious projects, to spare us the mind boggling problems in the future.
One path of A.I. progress that is being currently explored is the goal of not only creating a highly intelligent machine through replicating the human thought process but also, to develop a robot with a human-like consciousness that would be virtually indistinguishable from that of a human. With such a robot in existence, we would be inclined to treat it as a fellow human being, bringing in questions of the rights they should be accorded, that of a human or of a machine that we can shut down at any time (Roth, 2009). Currently when individuals are being arrested for damages to machines or robots, it’s the rights of people like the owners of these machines or robots that are being protected, not the machines. With the existence of such human-like robots in the future, the lines between man and machine will be blurred. With the case of civil rights, if they are accorded to such robots in the future, it stands to mean that they would be held responsible for any errors or laws broken committed by them, opening another huge can of worms with endless issues like those of the differences between software and hardware (Freitas, R.A., Jr., 1985).
These major issues have to be answered before technology spirals out of our control. The issue of A.I. is rather similar to that of all technological advancements. Like a fire, it can help us keep warm while on the other hand, consume our houses in an instant if we are not careful. As such, the author ultimately feels that the element of human control should always be present in A.I. as a safety feature, hence keeping machines and robots always under our command. This not only enforces the point of which A.I. was created in the first place, which was to help and serve our needs, but shuts down the possibilities of a technological singularity where machines ultimately dominate us. The element of human control as a safety switch also closes the door on questions of whether robots should deserve equal civil rights. With the case of autonomous robots, it is the author’s view that they should be kept to the commercial sectors involving simple repetitive tasks that do not require a level of intelligence or reasoning, rather than in high risk situations involving life and death, given the extreme difficulty of installing ethical systems in robots at the moment. 

Conclusion

We have indeed come a long way from ancient history till today. What used to be myths and stories that people dreamed about have begun to set foot in reality. The dawn of a new age started in the 1950s with extensive collaboration between many researches of various fields on the theories and guidelines of A.I., and was propelled by extensive government funding. Although the development of A.I. hit a major tumbling block in the 1970s, it managed to pick up momentum once again in the 1990s at a rapid pace. Machines with capabilities once thought of as impossible were being unveiled and mankind’s ambition sought to ensure even greater feats of engineering would occur. Such events would definitely announce the start of a revolutionary new age and greatly shape our ways of life.
However, the current situation poses enough clues about the potential implications of what could happen if we do not have a tight rein over our technological progress. While the future considerations raised earlier may not happen in the near future, it is indeed a possibility in the decades to come. The main concerns stem from the idea of robots one day operating beyond our control as well as having an equal or even high intelligence than the creators, humans ourselves. While proposals for ethical systems and guidelines to be built into artificial programming codes seem like the best solution, such engineering and technology simply does not exist currently and do not seem likely to be developed in the near future. As such, the idea of always having a human in overall control as a failsafe seems the most practical to the author. Such a system does have its disadvantages as it limits the setting up of an artificial neural network as well as the prevention of example-based learning systems which allow a robot to make improved decisions based on its experiences which are far more advanced than current programming systems based on mere programming code for the robots. However the elephant in the room cannot be ignored, and hence it is the author’s view that such measures have to be implemented until technological advancements have found a way to program such ethical guidelines into programming code.
Overall, A.I. in the future does have the endless potential to change our lives dramatically for the better. While there may be many concerns and potential problems to face, it is the author’s view that as long as we address them early and adequately, we would be able to harness the full benefits without creating unnecessary problems for ourselves.

References

AISB. (n.d.). What is Artificial Intelligence? Retrieved from http://www.aisb.org.uk/public-engagement/what-is-ai

Artificial Intelligence. (n.d.). Retrieved from http://curiosity.discovery.com/question/smarter-human-or-computer

BBC (2010 August 10). Nao the first robot with 'emotions' unveiled. Retrieved from http://news.bbc.co.uk/local/threecounties/hi/people_and_places/newsid_8900000/8900417.stm

Bloomfield, A. (2013) Lethal Autonomous Robotics: UN Calls For Global Moratorium On Killer Robots. Retrieved from http://www.policymic.com/articles/39017/lethal-autonomous-robotics-un-calls-for-global-moratorium-on-killer-robots

Chessbase. (2012 October 29). Houdini 3 – the world's strongest chess engine in the Fritz interface. Retrieved from http://en.chessbase.com/Home/TabId/211/PostId/4008591

Christensen, B. (2009 August 24). Robots Learn to Lie. Retrieved from http://www.livescience.com/10574-robots-learn-lie.html


Fry, A. (2011 July 27). A Cubic Millimeter of Your Brain. Retrieved from

Freitas, R.A.Jr. (1985 January 13). The Legal Rights of Robots. Retrieved from http://www.rfreitas.com/Astro/LegalRightsOfRobots.htm

Gaudin, S. (2013, October 10). U.S. Army evaluates self-driving, machine gun-toting robots. Retrieved from http://www.computerworld.com/s/article/9243120/U.S._Army_evaluates_self_driving_machine_gun_toting_robots

Global Security.org (2011 July 11). Samsung Techwin SGR-A1 Sentry Guard Robot. Retrieved from http://www.globalsecurity.org/military/world/rok/sgr-a1.htm

Goertzel, B. (2011 August 17) Does Humanity Need an AI Nanny? Retrieved from http://hplusmagazine.com/2011/08/17/does-humanity-need-an-ai-nanny/
Grachev, G. (2006 September 11) Humanoid robots existed in ancient civilizations. Retrieved from http://english.pravda.ru/science/mysteries/11-09-2006/84374-robots-0/#

Guizzo, E. (2011 October 18). How Google's Self-Driving Car Works. Retrieved from http://spectrum.ieee.org/automaton/robotics/artificial-intelligence/how-google-self-driving-car-works

Hars, A. (2013 September). Supervising autonomous cars on autopilot: A hazardous idea. Retrieved from http://www.inventivio.com/innovationbriefs/2013-09/Supervised-Autonomous-Driving-Harmful.2013-09.pdf

Keiper, A. & A.N. Schulman. (2011).The Problem with 'Friendly' Artificial Intelligence, The New Atlantis, Number 32, p. 80-89.

http://cacm.acm.org/magazines/2013/11/169024-legal-issues-with-robots/fulltext#

Kramer, M. (2013 July 9). Q&A With a Pilot: Just How Does Autopilot Work? Retrieved from http://news.nationalgeographic.com/news/2013/07/130709-planes-autopilot-ask-a-pilot-patrick-smith-flying-asiana/

Pilkington, E. (2013 May 29). 'Killer robots' pose threat to peace and should be banned, UN warned. Retrieved from http://www.theguardian.com/science/2013/may/29/killer-robots-ban-un-warning\

Reber, P. (2010 April 19). What is the memory capacity of the human brain? Retrieved from http://www.scientificamerican.com/article.cfm?id=what-is-the-memory-capacity

Roth, D. (2009 January 19). Do Humanlike Machines Deserve Human Rights? Retrieved from http://www.wired.com/culture/culturereviews/magazine/17-02/st_essay

Slocum, J. (1985). A Survey of Machine Translation: Its history, current status, and future prospects. Computational Linguistics, 11(1). Retrieved from http://acl.ldc.upenn.edu/J/J85/J85-1001.pdf

Smith, C., McGuire, B., Huang, T., Yang, G. (2006 December). The History of Artificial Intelligence. Retrieved from http://courses.cs.washington.edu/courses/csep590/06au/projects/history-ai.pdf

Stewart, B. (n.d.). Dartmouth Artificial Intelligence (AI) Conference. Retrieved from http://www.livinginternet.com/i/ii_ai.htm

Stottler Henke. (n.d.). Retrieved from http://www.stottlerhenke.com/ai_general/history.htm

The Economist. (2012 June 2). Morals and the machine. Retrieved from http://www.economist.com/node/21556234

                                                                                                                                 Turing, A. (1950) Computing Machinery and Intelligence. Retrieved from http://www.csee.umbc.edu/courses/471/papers/turing.pdf
                                                                                                                                          Walsh, F. (2013 October 7). Billion pound brain project under way. Retrieved from http://www.bbc.co.uk/news/health-24428162
Veloso, M. (2013). Autonomous Robot Soccer Teams. Retrieved from http://www.nae.edu/File.aspx?id=7300


[1] Peer reviewed by Sam Lim Kok Wei and Prashant Premchand Dadlani