Microsoft Word Kurzweil, Ray The Singularity Is Near doc

Space Exploration.
NASA is building self-understanding into the software controlling its unmanned spacecraft. 
Because Mars is about three light-minutes from Earth, and Jupiter around forty light-minutes (depending on the exact 
position of the planets), communication between spacecraft headed there and earthbound controllers is significantly 
delayed. For this reason it's important that the software controlling these missions have the capability of performing its 
own tactical decision making. To accomplish this NASA software is being designed to include a model of the 
software's own capabilities and those of the spacecraft, as well as the challenges each mission is likely to encounter. 
Such AI-based systems are capable of reasoning through new situations rather than just following preprogrammed 
rules. This approach enabled the craft Deep Space •• One in 1999 to use its own technical knowledge to devise a series 
of original plans to overcome a stuck switch that threatened to destroy its mission of exploring an asteroid.
185
The AI 
system's first plan failed to work, but its second plan saved the mission. "These systems have a commonsense model of 
the physics of their internal components," explains Brian Williams, coinventor of Deep Space One's autonomous 
software and now a scientist at MIT's Space Systems and AI laboratories. "[The spacecraft] can reason from that 
model to determine what is wrong and to know how to act." 
Using a network of computers NASA used GAs to evolve an antenna design for three Space Technology 5 
satellites that will study the Earth's magnetic field. Millions of possible designs competed in the simulated evolution. 
According to NASA scientist and project leader Jason Lohn, "We are now using the [GA] software to design tiny 
microscopic machines, including gyroscopes, for spaceflight navigation. The software also may invent designs that no 
human designer would ever think of."
186
Another NASA AI system learned on its own to distinguish stars from galaxies in very faint images with an 
accuracy surpassing that of human astronomers. 
New land-based robotic telescopes are able to make their own decisions on where to look and how to optimize the 
likelihood of finding desired phenomena. Called "autonomous, semi-intelligent observatories," the systems can adjust 
to the weather, notice items of interest, and decide on their own to track them. They are able to detect very subtle 
phenomena, such as a star blinking for a nanosecond, which may indicate a small asteroid in the outer regions of our 
solar system passing in front of the light from that star.
187
One such system, called Moving Object and Transient Event 
Search System (MOTESS), has identified on its own 180 new asteroids and several comets during its first two years of 
operation. "We have an intelligent observing system," explained University of Exeter astronomer Alasdair Allan. "It 
thinks and reacts for itself, deciding whether something it has discovered is interesting enough to need more 
observations. If more observations are needed, it just goes ahead and gets them." 
Similar systems are used by the military to automatically analyze data from spy satellites. Current satellite 
technology is capable of observing ground-level features about an inch in size and is not affected by bad weather, 

---

clouds, or darkness.
188
The massive amount of data continually generated would not be manageable without automated 
image recognition programmed to look for relevant developments.