Approaches to designing intelligent robots
In order to construct Turing Androids who can walk and talk like humans, what would be the best
approach of the following options?
001. Work on modifying human genes so as to create more intelligent humans whose memory and
intelligence would be superior. Use these superior humans to design Turing robots.
Negative aspect is the time it takes to raise human children
002. Work on adding electronics in order to improve the memory and processing speed of the human
003. Download Open cyc and get it to work. www.opencyc.org/
004. Just sit back and watch the web evolve into an intelligent being. Words and their definitions are at
wiki, cyc is working on symantics and parsing of sentences.
005. Very soon Microsoft, Yahoo and Google will be in a parsing race to see who will dominate the next
phase of search engines. The search for facts and knowledge rather than web site content.
006. Improvements in chip design set us an example of how to design robots. We need to develop the
tools needed to develop robots rather than attacking the problems directly.
007. Video games are evolving in the direction of virtual reality. If we watch the evolution of Wii and Play
Stations, we will see them evolve into intelligence. Play Station 3 has 8 processors.
008. Neural networks (Artificial Neural Networks). Continue to develop and evolve these networks until
we finally reach the complexity of the human brain.
009. Write computer programs to simulate ants in the format of artificial life. Evolve these ants into more
complex animals. Guide the evolution so that the resultant artificial animals evolve in the direction of
humans. End up with a simulation of humans.
010. Duplicate the functions of each part of the human brain. Kurzweil states that 20 modules have
already been simulated (In Depth interview on C-SPAN Books www.booktv.org/feature/index.asp?
segid=7515&schedID=457) He also stated that the total number of modules is "several hundred". (TV
show = www.booktv.org/ram/feature/1106/arc_btv110506_4.ram)
011. Create a system that learns, and then it will learn to make models of itself, and discover ways to
improve themselves. So far, this method has failed, but this is not proof that it will never succeed.
012. Marvin Minsky will put together a team of three system programmers and deliver a prototype in 3 to
013. Ben Goertzel will finally succeed in his attempts to create artificial general intelligence.
014. Ray Kurzweil's Ramona will evolve into an intelligent robot. He says that he will make Romona
more independent and more intelligent. He expects that Romona will pass the Turing test in 2029 so
that he can win his bet with Mitch Kapor. This is bet number 1 on longbets. www.longbets.org/1
015. Rodney Brooks will succeed with his Living Machines.
016. Develop and Intentional Programming System like Charles Simonyi's. (See Technology Review,
017. Jeff Hawkins is making advances in understanding the human brain. He will eventually deliver a
product, just like he delivered the Palm Pilot.
018. Toys like Robosapien will evolve each year or two and will display human intelligence in a decade
or two. http://www.rsmediaonline.com/
019. Robots such as Roomba now have connectors attached to WiFi modules which allows the average
12 year-old kid to play with the software. Eventually some kid will win his science fair by making a
roomba as intelligent as he is.
020. We first need more computational power. To obtain intelligent systems, we need to start out with
making faster and larger computer chips. Three-dimensional chips with optical communications to lower
capacitative crosstalk problems.
021. We need to move to cubic carbon technology. Diamond conducts heat ten to thirteen times faster
and can make faster transistors.
022. Develop reverse computing which lowers the power required to compute.
023. Use reverse computation to speed things up. sim.sagepub.com/cgi/content/abstract/82/1/61
Parallel simulation • discrete event simulation • optimistic simulation • reverse computation • plasma
024. Use parallel computers to simulate each nerve cell. Blue Gene /L is attempting to simulate neurons
and then simulate the human brain. Will supercomputers be the first to show intelligence?
025. Google used a large Rosetta stone style of texts to teach its software to translate Arabic to English
and English to Arabic. Nobody on the Google team spoke any Arabic. This indicates progress toward
Artificial Neural Network (ANN) learning how to think. blog.outer-court.com/archive/2005-05-22-n83.html
026. Roger Penrose will show how the human brain uses quantum computing to create its ability to think
intelligently. Engineers will follow his lead and develop quantum computing devices to simulate the
functioning of the human mind.
027. The heart of human intelligence is pattern recognition. R2 technologies developed a neural
network which recognizes breast cancer on X-rays as well as human radiologists who specialize in breast
cancer do. It does better than non-specialist human radiologists. Is one of the things we must have for
artificial intelligence a group of such pattern recognizers? Does the human mind also have faster
recognition systems that can be used to perform practical day-to-day functioning?
028. Savants such as Kim Peek have much larger memories than average people but cannot button
their shirts. Are his neurons redirected to different tasks? Does this indicate the complexity of functions
which humans take for granted as easy such as buttoning one's shirt?
029. Study the brain and psychology, especially evolutionary psychology in order to understand more
fully the workings of the human brain. Article The Moral Maze in New Scientist November 26, 2005.
030. Take a look at Expert Systems again. Intel started to tell audiences in 2003 that the expert
systems which failed so badly in the last century were finally coming up to speed. Check this out, is it
true? Cyc talks about using opencyc for developing expert systems. www.opencyc.org
Other expert systems: www.aaai.org/AITopics/html/expert.html or www.xpertrule.com/index.htm
031. Write papers for and attend the PBA 2007.
2nd Workshop on Parallel Bioinspired Algorithms (PBA 2007)
boinc.unex.es/pba, www.sigevo.org/gecco-2007/workshops.html to be held at 2007 Genetic and
Evolutionary Computation Conference July 7-11, 2007, London, England
032. Many more papers and conferences.
033. Try genetic algorithms, genetic programming.
034. Try evolution strategy, evolutionary programming.
035. Try probabilistic models, estimation of distribution algorithms, or probabilistic search methods
036. Try swarm intelligence, ant colony optimisation. Try to apply it to simulate intelligence.
037. Play with artificial immune systems.
038. Study the human brain and then try multi-agent systems.
039. Start a Open Source Code project (Like SourceForge.net, etc.) and everyone will volunteer.
040. Bioinspired algorithms.
041. Arthur T. Murray does it with what he calls Mentifex. According to his schedule, he already
achieved "True AI last June 7th and we are now in the "Artificial Intelligence Landrush" phase of the
schedule during 2007. His book is called AI4U.
See: Mentifex www.scn.org/~mentifex/ he lays out a timetable as follows:
III. Singularity Timetable
2006 June 7 -- True AI (achieved on 7 June 2006)
2007 -- AI Landrush
2009 -- Human-Level AI
2011 -- Cybernetic Economy
2012 -- Superintelligent AI
2012 -- Joint Stewardship of Earth
2012 -- Technological Singularity
- - - -
sounds like the singularity is really really near.
042. Check out A General Methodology for Designing Self-Organizing Systems by Carlos