What will happen when machines outthink us?

"At the center of a black hole there lies a point called a singularity where the laws of physics no longer make sense.

In a similar way, according to futurists gathered Saturday for a weekend conference, information technology is hurtling toward a point where machines will become smarter than their makers. If that happens, it will alter what it means to be human in ways almost impossible to conceive, they say.

"The Singularity Summit: AI and the Future of Humanity" brought together hundreds of Silicon Valley techies and scientists to imagine a future of self-programming computers and brain implants that would allow humans to think at speeds nearing today's microprocessors.

Artificial intelligence researchers at the summit warned that now is the time to develop ethical guidelines for ensuring these advances help rather than harm.

"We and our world won't be us anymore," Rodney Brooks, a robotics professor at the Massachusetts Institute of Technology, told the audience. When it comes to computers, he said, "who is us and who is them is going to become a different sort of question."

Eliezer Yudkowsky, co-founder of the Palo Alto-based Singularity Institute, which organized the summit, focuses his research on the development of so-called "friendly artificial intelligence." His greatest fear, he said, is that a brilliant inventor creates a self-improving but amoral artificial intelligence that turns hostile.

T-minus 22 years?
The first use of the term "singularity" to describe this kind of fundamental technological transformation is credited to Vernor Vinge, a California mathematician and science-fiction author.

High-tech entrepreneur Ray Kurzweil raised the profile of the singularity concept in his 2005 book "The Singularity is Near," in which he argues that the exponential pace of technological progress makes the emergence of smarter-than-human intelligence the future's only logical outcome.

Kurzweil, director of the Singularity Institute, is so confident in his predictions of the singularity that he has even set a date: 2029.

Most "singularists" feel they have strong evidence to support their claims, citing the dramatic advances in computing technology that have already occurred over the last 50 years.

In 1965, Intel co-founder Gordon Moore accurately predicted that the number of transistors on a chip should double about every two years. By comparison, singularists point out, the entire evolution of modern humans from primates has resulted in only a threefold increase in brain capacity.

With advances in biotechnology and information technology, they say, there's no scientific reason that human thinking couldn't be pushed to speeds up to a million times faster.

Is the ‘nerdocalypse’ near?
Some critics have mocked singularists for their obsession with "techno-salvation" and "techno-holocaust" _ or what some wags have called the coming "nerdocalypse." Their predictions are grounded as much in science fiction as science, the detractors claim, and may never come to pass.

But advocates argue it would be irresponsible to ignore the possibility of dire outcomes.

"Technology is heading here. It will predictably get to the point of making artificial intelligence," Yudkowsky said. "The mere fact that you cannot predict exactly when it will happen down to the day is no excuse for closing your eyes and refusing to think about it.""

Researchers Turn Web Blather to Books

"A few simple keystrokes may soon turn blather into books. Researchers at Carnegie Mellon University have discovered a way to enlist people across the globe to help digitize books every time they solve the simple distorted word puzzles commonly used to register at Web sites or buy things online.

The word puzzles are known as CAPTCHAs, short for "completely automated public Turing tests to tell computers and humans apart." Computers can't decipher the twisted letters and numbers, ensuring that real people and not automated programs are using the Web sites.

Researchers estimate that about 60 million of those nonsensical jumbles are solved everyday around the world, taking an average of about 10 seconds each to decipher and type in.

Instead of wasting time typing in random letters and numbers, Carnegie Mellon researchers have come up with a way for people to type in snippets of books to put their time to good use, confirm they're not machines and help speed up the process of getting searchable texts online.

Humanity is wasting 150,000 hours every day on these," said Luis von Ahn, an assistant professor of computer science at Carnegie Mellon. He helped develop the CAPTCHAs about seven years ago. "Is there any way in which we can use this human time for something good for humanity, do 10 seconds of useful work for humanity?"

Many large projects are under way now to digitize books and put them online, and that's mostly being done by scanning pages of books so that people can "page through" the books online. In some cases, optical character recognition, or OCR, is being used to digitize books to make the texts searchable.

But von Ahn said OCR doesn't always work on text that is older, faded or distorted. In those cases, often the only way to digitize the works is to manually type them into a computer.

Von Ahn is working with the Internet Archive, which runs several book-scanning projects, to use CAPTCHAs for this instead. Internet Archive scans 12,000 books a month and sends von Ahn hundreds of thousands of files that are images that the computer doesn't recognize. Those files are downloaded onto von Ahn's server and split up into single words that can be used as CAPTCHAs at sites all over the Internet.

If enough users decipher the CAPTCHAs in the same way, the computer will recognize that as the correct answer.

"If we can correct these books so that they are really in good shape, then you can go and use these books in other type devices more easily" such as handheld computers or in programs for reading to the blind, said Brewster Kahle, co-founder of the Internet Archive.

Von Ahn approached the Internet Archive to get help in developing the new system, but it has not been put into use yet. Theoretically, von Ahn said the new book-based CAPTCHAs could be used in place of any CAPTCHA currently on the Web.

The project, named reCAPTCHA, is one of many projects that enlist computer users from the community to help out. For example, Cloudmark Inc. uses its base of users to judge what is spam and what isn't. News aggregation sites like Digg Inc.'s digg.com and Time Warner Inc. (TWX)'s Netscape.com ask visitors to recommend and vote on items to go on top.

For von Ahn's project, Intel Corp. (INTC) donated equipment and the work was sponsored by the MacArthur Foundation, which awarded von Ahn a "genius grant" last year.

Kahle, whose Internet Archive has about 200,000 books currently online, is working with libraries in three countries to digitize their books. Kahle said von Ahn's project is "harnessing human power in exactly the right way."

"It's definitely a barn-raising to try to build the great library," Kahle said."

On the net:

http://www.recaptcha.net
http://www.gutenberg.org