	Generational Dynamics
	Forecasting America's Destiny ... and the World's

\| HOME \| WEB LOG \| COUNTRY WIKI \| COMMENT \| FORUM \| DOWNLOADS \| ABOUT \|

Generational Dynamics Web Log for 27-Dec-2010
27-Dec-10 News -- IBM vs Jeopardy! brings robotic warfare and the Singularity closer

Web Log - December, 2010

27-Dec-10 News -- IBM vs Jeopardy! brings robotic warfare and the Singularity closer

Jeopardy! broadcast is scheduled for February 14-16

IBM vs Jeopardy! brings robotic warfare and the Singularity closer

In 2005, the Department of Defense announced the Future Combat System program. (See "Army's 'Future Combat System' calls for autonomous robot soldiers by 2014.")

The plan was that a network of intelligent battlefield robots, including aircraft, ground vehicles, and units that look like toy trucks, would replace the duties of ordinary soldiers, and would have the ability to kill. However, at least initially, the robots would not be making autonomous decisions to kill, and humans would have to approve any killing.

Other countries took up the concept. The above is the scariest video you've ever seen. It's a 2007 promotional video for heavily armed robot sentries being considered by the South Koreans to guard the North Korean border. The robot would have the ability to detect intruders, give verbal warnings, interpret voiced responses, and fire if necessary.

The project was canceled in 2008, after a a year-long pilot showed that the robots were "unfit for combat," according to Global Security. However, Stars and Stripes reports that a new generation of the robots are now being tested along the North Korean border.

In 2009, development of DoD's Future Combat Systems was terminated because it was substantially over budget and wasn't showing results. Some of the individual components, including unattended ground sensors, were spun off and are still under development.

This is the way artificial intelligence has gone since its inception. In the 1950s, it was promised that computer algorithms would be developed that would make a computer the world chess champion within ten years. By 1970, computer chess could only play as well as a mediocre human player.

It's only in the last few years that IBM's Deep Blue computer could credibly claim to be world champion. Is that because IBM's scientists finally developed those clever algorithms that were promised in the 1950s? Hardly. In fact, IBM's winning computers used pretty much exactly the same algorithms (the so-called Minimax Algorithm) that was used in the 1960s.

The only major change from the 1960s is that today's computers are much faster and more powerful. A 1960s computer could look ahead only three chess moves; today's computers can look ahead almost ten moves in the same amount of time.

Artificial intelligence experts used to refer to solutions that depended on the power of computers by the pejorative term "brute force solutions." But as AI has failed in one objective after another, it turns out that brute force is the only game in town.

And now there's a new game in town that IBM's computers are going to challenge.

IBM's Watson computer will play the game of Jeopardy! against former game champions in matches to be aired on February 14-16, according to IBM.

In this game, the host gives a clue in the form of an answer to a question, and three contestants compete to give the quickest response in the form of a question. The game is filled with language tricks and riddles and double meanings, and so this match will be quite significant: It will be a major test of AI problem known as "natural language processing." For the first time, a computer will be required to quickly recognize and analyze arbitrary English language sentences, and respond to them appropriately, with no restrictions.

A NY Times article in June explained how Watson had been trained.

During the three years of Watson's development, the computer was fed tens of millions of documents, including "books, reference material, any sort of dictionary, thesauri, folksonomies, taxonomies, encyclopedias, any kind of reference material you can imagine getting your hands on or licensing. Novels, bibles, plays."

The computer didn't just store this material. It first did "natural language processing" on the material. This is an artificial intelligence technique that interprets an ordinary English language sentence correctly. Then, it did statistical analyses on the material and created a database of concepts and connections. For example, it would determine that the name "Sherlock Holmes" appeared frequently in the same sentence as phrases like "deerstalker hat" and "Professor Moriarty" and "opium," but never appeared in the same sentence with, say, "Super Bowl." Using this kind of database, Watson can examine the words in a Jeopardy clue, look up connections in its database, and hopefully come up with the correct response in the few seconds alloted.

This kind of computer solution is possible today only because computers have continued to become more and more powerful. Although you can call Watson an "idiot savant" solution to a problem, it's really a quite significant step in the development of computers with human and superhuman intelligence. And if all goes as hoped on February 14-16, then it will have been accomplished several years earlier than I had expected.

And that means that it's a very significant step on the road to the Singularity, the point in time when computers will be more intelligent and more creative than humans. At that point there will be a sharp bend in the technology curve, since super-intelligent computers will be able to develop new technologies exponentially faster than humans, including technologies to make themselves faster.

My personal epiphany on this subject came in 2004 when, as the result of an online discussion, I informally developed the software architecture that would be used in the first generation of super-intelligent computers. (See Generational Dynamics: The Singularity for a summary of the architecture.) Going through that process made the Singularity real to me, and I could see how it was likely to unfold.

The algorithm that I developed cannot be implemented on today's computers, because today's computers aren't fast and powerful enough. But computer power has been growing exponentially for many decades, doubling every 18 months, with no end in sight.

When will the Singularity occur? Various experts have suggested dates anywhere from 2020 to 2060. My own estimate is that computers will be powerful enough to fully implement the architecture that I developed by the early 2020s, and that the Singularity will occur around 2030. However, if the Jeopardy! challenge proceeds as planned, then it may turn out the 2030 date will have to be moved closer by a few years.

Even if you still believe, as many do, that the Singularity will never occur because the human brain has some magical powers that no computer could ever have, or whatever, it's still undeniable that computer power is doubling every 18 months, and that this will affect all our lives, as individuals, as a nation, and as a planet. Every domestic and foreign policy is affected.

For example, by the 2020s, we'll have everything from intelligent computer plumbers to intelligent computer nursemaids. The problem of skyrocketing health care costs will be resolved not by a government program but by offloading the vast majority of every day medical functions to intelligent robots. Robots will also be used regularly for warfare in that time frame.

As an example of international policy, consider "climate change." Climate experts are predicting that the seas will rise by a few inches by 2100, but that prediction is as ridiculous as a 1910 prediction that the world would be covered by horse droppings by 2000. Technology is moving too quickly for any such prediction to be relevant.

New computer technologies will solve resolve any climate change issues in the 2020s in ways that today cannot be predicted. And if the Singularity occurs by 2030, as I expect, then the entire climate change debate is total nonsense.

When the "climategate" scandal broke last year, I did a search of the hacked East Anglian e-mail messages for the word "Singularity," and it doesn't appear once. So you have all these brilliant scientists working to predict climate change, and you would think that at least one of them would have asked the question of one another: "Hey, what about the Singularity?" But apparently it was never asked once. It must have been forbidden subject, since it conflicts with their own claims.

The Jeopardy! challenge, if it's successful, will be a major step forward for computer intelligence and the Singularity. Policymakers around the world should be factoring it into their plans. Military planners around the world are certainly doing so, and we should see the results within a few years.

(Comments: For reader comments, questions and discussion, see the 27-Dec-10 News -- IBM vs Jeopardy! brings robotic warfare and the Singularity closer thread of the Generational Dynamics forum. Comments may be posted anonymously.) (27-Dec-2010) Permanent Link
Receive daily World View columns by e-mail
Donate to Generational Dynamics via PayPal

Web Log Pages

Current Web Log

Generational Dynamics Web Log for 27-Dec-2010 27-Dec-10 News -- IBM vs Jeopardy! brings robotic warfare and the Singularity closer

IBM vs Jeopardy! brings robotic warfare and the Singularity closer

Web Log Pages

Generational Dynamics Web Log for 27-Dec-2010
27-Dec-10 News -- IBM vs Jeopardy! brings robotic warfare and the Singularity closer