Robotics

The Basics:

Since Leonardo da Vinci’s first sketches of a mechanical man, robots have captivated humankind. Today, in many domains, robots are mere tools. In others, researchers strive to build robots that emulate and perhaps surpass the physical and mental deftness of their creators.

In the simplest sense, a robot must engage with information from its surroundings and do something physical with this information. The precision and thoroughness with which robots engage with their environment and the level of creativity and functionality in their analysis of this environment are some of the parameters weighed by pioneering robotics researchers.

In 1959 robots entered the field of manufacturing and illustrated the vast potential for specialized, consistent, and accurate machines that could do work for humans in hazardous or inhumane work environments. As computer processors get cheaper and faster, top researchers are renewing their interest in robotics and developing ambitious goals for robotics, moving the field into the next generation—where robots are beginning to possess levels of sociability and versatility. These robots, like Domo¹ and Kismet² from MIT, incorporate theories from social development psychology, ethology, and evolution.

The field on the whole is looking ahead to an environment where robots may become capable of assisting in the more intimate, fundamental aspects of life. This research, spearheaded at MIT, sees robotics together with AI, as developing versatile capabilities and ultimately the ability to learn and improve.

The Procedures:

While traditional autonomous robots, such as specialized robotic arms working on factory lines, are designed to operate independent from humans, the next generation of sociable humanoid robots is being designed to interact, cooperate and learn from people.

This trend coincides with current trends in AI research, and in fact it is difficult to fully tease the two domains apart. At IBM for example, the follow-up project to Deep Blue, Joshua Blue,³ pursues advanced AI robotics by studying how children develop dexterity and intelligence from their physical and social environment.

This ‘learner’ paradigm plays an important role at MIT as well. Their fist major success with a sociable robot, Kismet,⁴ was designed with an altricial system, similar to a young child. Kismet included visual and auditory sensors, a processing system that modeled attention, behavior, perception, motivation, and emotion systems with evolution and developmental psychology-inspired algorithms and heuristics, and an output system that produced appropriate vocalizations, head and eye orientations, and facial expressions. “Domo,” ⁵ MIT’s follow-up to Kismet, applies these same principles to a robot that can locate a human by sound and sight, grasp an offered item, and place it on a shelf requested by a human http://www.youtube.com/watch?v=Ke8VrmUbHY8&NR=1.

Even though researchers all over the world have made enormous strides modeling human tasks, the simplest actions, like walking, still prove difficult for robotics researchers. At the Honda labs researchers have seen success with their robot ASIMO. But even while ASIMO ascends stairs using an onboard computer that works to neutralize stability-jeopardizing forces, it can’t mimic the fluidity with which a child moves http://www.youtube.com/watch?v=Q3C5sc8b3xM. Probably the leader of the pack in terms of agility is the four-legged Big Dog by Boston Dynamics http://gizmodo.com/368651/new-video-of-bigdog-quadruped-robot-is-so-stunning-its-spooky.

Relationship to Terasem:

As long as robotics research employs diverse scientific disciplines, fruitful theories and implications for sociable robots will have applications for Cyberconsciousness. Computer systems that emerge to enable a robot to sense stimuli from its environment, interpret the data, and operate in various novel ways, can be almost directly applied to Cyberconsciousness. Clearly these systems will rely heavily on AI and neuroscience research, and the three disciplines will likely advance and cross-pollinate in concert.

 

Furthermore, Nanobots, the smallest of all robots, pose great possibilities for the exploration of the human body and mind⁶. Nanobots will operate with the same central tenets of all robots (to sense, interpret, and act) but will aim to model biological machinery (such as cilia and flagella) to do their work on the molecular scale http://bionano.rutgers.edu/or.html.

These nanorobots, because of their size, will likely have the ability to enter human arteries to repair damage and interact with nerve cells to gather information about the still mysterious nature of consciousness. In their ability to gather previously inaccessible information, nanorobotics may provide breakthroughs for sociable robotics, Cyberconsciousness, and nearly every other domain of advanced science.

Sources and further reading:

http://www.shadowrobot.com/

http://www.ai.mit.edu/projects/humanoid-robotics-group/kismet/kismet.html

http://web.mit.edu/newsoffice/2007/domo.html

http://asimo.honda.com/default.aspx

http://www.thetech.org/exhibits_events/online/robotics/universal/

http://www.ri.cmu.edu/

http://www.nanobot.info/

http://bionano.rutgers.edu/or.html

http://www.youtube.com/watch?v=Ke8VrmUbHY8&NR=1

http://www.youtube.com/watch?v=Q3C5sc8b3xM

Artificial Intelligence

The Basics:

Alan Turing, the brilliant mathematician who is famous for helping to crack the cipher machine used by the Germans in WWII, put forth the possibility of artificial intelligence in 1950, in his seminal paper “Computing Machinery and Intelligence.”¹ Six years later, at a Dartmouth College conference spearheaded by the computer scientist John McCarthy, expert mathematicians came together to officially launch the discipline of Artificial Intelligence.

Artificial Intelligence (AI) is the engineering of machines and computer programs that possess intelligence. However, a definition of intelligence proves elusive. John McCarthy calls it the “computational art of achieving your goals.”² The “art” portion of this statement, more than 50 years after the AI Dartmouth conference, has the AI community still challenged.

Artificial Intelligence today should be broken into two realms: Narrow AI and Artificial General Intelligence (AGI). Narrow AI systems are confined to specific domains of intelligence. IBM’s Deep Blue computer for example, can process 600 million chess moves per second. By using computer algorithms that enable intelligence development in regard to positions and situations, Deep Blue was able, in 1997, to defeat the top-ranked chess player in the world, Gary Karsparov. ³

Deep Blue’s chess intelligence is astounding, but it is unable to apply this intelligence in another domain, such as global climate change prediction. Kasparov’s famous quip after his loss, “At least it couldn’t enjoy its victory,” expressed further limitations of narrow AI.

General AI research is largely pursued by universities and not-for-profit institutions because the task seems currently much more ambitious and much less marketable than narrow AI systems. The field seeks to develop software and robotic systems with intelligence that can be applied to a variety of environments to solve a wide array of complex problems. Like the exploration of space, the challenging task of general AI research draws on a variety of subjects, attracts some of the most brilliant scientific minds, and produces technologies and theories that benefit many other scientific pursuits.

The Procedures

The quest to achieve AGI, a young, immensely complex undertaking, is fractured into subgroups, each using different paradigms in their research. The main philosophical divide exists among the “neats” and the “scruffies.”⁴ In the world of AI the neats are composed of researchers that favor a clear and organized approach. They favor models of design that can be neatly proven correct. On the other hand, scruffies contend that intelligence is intractably connected with the irrational mystery of consciousness, and is too fluid and complicated to approach with the clear and rational tools of logic and applied statistics. The divide represents different perspectives on the nature of the human brain. Are we fundamentally rational or irrational? The disagreement also highlights the question of whether or not researchers should pursue AI with the powerful, but extremely messy human brain as a model. Presumably, other models of intelligence, different from the human model could arise and prove immensely effective.

The prominent, neat John McCarthy has famously said that he cares about general artificial intelligence, but not necessary via the imitation of humanoid intelligence. On the other hand, the prominent AI researcher and inventor of the palm pilot, Jeff Hawkins, believes that the key to developing truly intelligence machines lies in the complex neo-cortex of the human brain. His company, Numenta, develops computer memory systems modeled on the human neo-cortex.

Both schools of AI research find it necessary to draw on many resources to tackle their immense challenge. Search algorithms, logic systems, probability theory, economic theories, evolutionary computation, and neural network theories all come into play. Furthermore, new ideas, even radically different paradigms must be considered to continue the general AI pursuit. At the 50^th year anniversary meeting of 1956 Dartmouth conference the organizers affirmed that the metaphor of the brain as a computer must be discarded. The panel acknowledged considerable advances in narrow AI in the last 50 years but stated that, “what has been missed is – we believe – how important embodiment and the interaction with the world are as the basis for thinking. Quite recently it has become evident that many fields (linguistics, cognitive sciences, neuroscience, morphogenesis, artificial intelligence, robotics, and material sciences) are highly relevant in order to advance the state of the art. It is our conviction that breakthroughs can only be achieved by a strong cross-fertilization between these fields.”

IBM seems to agree with this advice. While working on their latest AI endeavor, Joshua Blue, researchers at IBM have made a point of seeking out experts on the neat and scruffy side in order to launch a project that seeks to model intelligence after a the brain of a young child.⁵ Joshua Blue incorporates natural language understanding, common sense reasoning, and to some extent, emotional intelligence capabilities. The software design company Novamente, applies a similar perspective as they develop computer programs that can reflect on their past actions and learn from those experiences. The hope, as the Dartmouth panel hints, is that the key to unlocking AGI may be in studying and modeling how intelligence develops.

Relationship to Terasem:

An AI system’s recognition of its own intelligence and an ability to improve itself - a sort of consciousness, will undoubtedly prove a necessary development as researchers pursue the most advanced AGI systems. Terasem’s mission of cyberconsciousness ⁶ plays an important role in the possibility of an AGI system because cyberconsciouness, although currently still theoretical, combines the massive parallelism, evolutionary history, and versatility of the human brain with the increasing speed, breadth, and longevity of the Internet. This powerful union could prove the ultimate form of AI, capable of solving extremely complex problems with diverse subject matter.

Undoubtedly, as the Internet becomes increasingly part of our lives, it will become increasingly incorporated in the AGI paradigms. Already we are seeing search engines and Internet advertisements develop a sort of narrow intelligence. With the development of web 3.0 (Semantic Web) the enormous content of the Internet may become readable and understood by computer programs, expanding the breadth of their intelligence. This development will allow websites such as Terasem’s lifenaut.com access to the vast stores of information on the web, exponentially increasing the data available to the site’s “Mindware” and increasing the chances for researchers to uncover the parameters and qualities of emergent artificial intelligence.

"AI's in the future will be able to recreate people from the information left behind about them if suitable backups of their brain were not made (in which case it would be straightforward).  Neural nanobots would obtain all the available information about them from other people's brains.  The AI would also consider all of the person's writings, pictures, movies, etc.  Also their genetic code.  And it could then create a person who would pass a Turing test for that person with their best friends as the judges.  For that reason it is worthwhile keeping your own files -- letters, emails, photos, writings, etc. Is this recreated person the same person?  It is an interesting question, but we could also ask today are we the same person as we were, say, a year ago.  The recreated person by the AI is probably at least as close as we are to ourselves after some time passage."
-- Ray Kurzweil

Inventor of the All-Font Scanner, Talking Book for the Blind & Kurzweil Piano

Creator of AI Music Composers, AI Poets and AI Artists

Recipient of National Medal of Technology

References and Further Reading:

http://www.aboutai.net/DesktopDefault.aspx

http://www.compapp.dcu.ie/~humphrys/ai.links.html

http://www-formal.stanford.edu/jmc/whatisai/

http://www.a-i.com/

http://www.aaai.org/home.html

http://www.csail.mit.edu/index.php

http://www.nytimes.com/2007/12/02/books/review/Henig-t.html?_r=1&bl&ex=1197003600&en=1c347f0cdd5b6e57&ei=5087%0A&oref=slogin

http://www.npr.org/templates/story/story.php?storyId=16816185&ft=1&f=1007

http://www.isi.imi.i.u-tokyo.ac.jp/~maxl/Events/ASAI50MV/

 

Pygmalion and Galatea Deep Blue ?

Multimedia Spider Deciphering Research

The Basics:

Web crawlers are the diligent workers of the World Wide Web. These methodical crawler programs comb the web, utilizing its extensive linkages to provide useful up-to-date lists of relevant web sites in response to human queries.

As the web evolves so do its web crawlers. The new web, (called Web 3.0 or the Semantic Web ¹) will require web crawlers to be capable of interpreting content embedded in diverse sources of text, photos, audio, and video. These ‘multimedia deciphering spiders’ are the next generation of web crawlers.

Traditionally, when web users search for information, web crawlers follow specific algorithms to seek out useful sites based on key words and link popularity. The web users then must search through the results to find exactly the piece of information they seek. Multimedia spiders could take over this task for humans. With advancements in AI research and standardizations of the web landscape, these spiders will drastically cut down information acquisition time and set the stage for a true harnessing of the power of the Internet. This isn’t a new idea. Tim Berners-Lee proposed it as early as 1994 at the first World Wide Web conference². Today, despite 50 years of AI research, the semantic web and its spiders remain largely unrealized.

The Procedures:

Multimedia spider deciphering research necessarily focuses on both the functional spiders and the semantic web that serves as their terrain. Tech start-ups (Radar Networks³), established titans (Google⁴ and Yahoo⁵), and public sector organizations (World Wide Web Consortium⁶) have all taken up this challenge⁷,⁸. Some organizations put emphasis on advancing the capabilities of multimedia spiders, developing their ability to interpret diverse forms of data with concepts borrowed from AI. Others take up the task of standardizing the language of cyberspace, making the spider’s job of extracting meaning easier.

The quest for this standardization proves enthusiastic in the life sciences where research demands the integration of heterogeneous data sets originating form separate subfields. These scientific communities, through the adoption of ontologies, are developing language standards for scientific web databases. Ultimately a researcher will be able to ask a specific scientific question, and multimedia spiders, through interpretation of the standardized language across all the relevant scientific web pages, will yield, not a website, but a specific answer from data culled from many websites and many forms of media.

The web agents of the web necessarily possess a strategy and architecture. They start with lists of primary URL’s to visit according to a certain topic and as they “crawl” through these sites they identify links to related sites, recursively adding them to their list of sites to be visited until they finally emerge with a list of the most promising sites (in the case of web crawlers) or a specific answer (in the case of multimedia deciphering spiders). The programmed strategy and architecture of web crawlers dictates how deeply they pursue peripheral sites and how quickly they pursue updated sites. The finest programmed crawlers emerge with a balance of volume, quality and freshness⁹.

Relationship to Terasem:

The pursuit of cyberconsciousness through AI advancement, advanced personality capture, or nanotechnology, will develop speed and scope as it incorporates multimedia spiders navigating the semantic web. Rather than humans searching through websites to extract relevant data, spiders could perform this task at a tremendous pace. For cyberconsciousness websites like lifenaut.com, spiders could comb the vast media offerings of the web to accumulate and connect sources relevant to a certain personality. The multimedia spidering software could also be used within mindfiles to search video itself for emotions, events, memories and link them with related text documents and chatbot conversations. The possibilities are infinite.

Like many of the science topics behind Terasem, multimedia spider deciphering research should not be viewed in isolation. The true potential of these spiders will undoubtedly be realized only in concert with AI advancements. Spiders combing the semantic web will provide seeming intelligent answers to questions, but the integration of these answers, to solve the most complex problems, will only be realized as the spiders achieve emergent AI properties. The potent combination of the semantic web and web spiders with some form of AI will enable exceptional fidelity of the personality capture available at lifenaut.com, and with time, advanced cyberconsciousness.

Sources and further reading:

http://money.cnn.com/magazines/business2/business2_archive/2007/07/01/100117068/index.htm?postversion=2007070305

http://dollar.biz.uiowa.edu/~pant/Papers/crawling.pdf

http://www.nytimes.com/2006/11/12/business/12web.html?_r=1&oref=slogin

http://eprints.ecs.soton.ac.uk/12614/1/Semantic_Web_Revisted.pdf

http://www.news.com/8301-10784_3-9824586-7.html?tag=nefd.top

www.obofoundry.org