Singularity Summit Roundup

Ashley and I attended the Singularity Summit at Stanford yesterday – an event focused on examining the future when artificial intelligence and other advanced technologies reach an “event horizon” that dramatically exceed or augment humanity’s abilities. Four of the presenters in particular stood out: Ray Kurzweil, Erik Drexler, Bill McKibben, and Cory Doctorow.

Ray Kurzweil started the conference off by reading a page of text from his new book, The Singularity Is Near, using a compact device designed to read books for the blind. Using the device, Ray took a picture of the page with the device and had it read the page aloud on his behalf. This is apparently a project that Ray started working on with the National Federation of the Blind about five years ago, but at the time the technology was not sufficiently advanced to enable the application. At the time of original investigation, digital cameras didn’t have enough resolution to enable good pattern recognition, and pattern recognition algorithms had not yet been designed to handle the difficult environment in which such a device would need to operate. All this changed in the intervening five years. This ever increasing rate of change, which Ray terms the Law of Accelerating Returns, was the central focus of Ray’s keynote.

Ray blazed through a presentation that exposed a lot of the core data that he uses in his book to support his thesis of the accelerating rate of change. Particular items which I found to be interesting:

• The expectation is that traditional photolithography will allow 4nm features by 2018, providing sufficient computing power to emulate the human brain for under $1000.
• The majority of Ray’s slides consisted of graphs showing either capabilities trending linearly up and to the right (on a log-scale diagram), or price/performance ratios trending linearly down and to the right (on a log-scale diagram). What I found amazing was just how steady, how predictable these trends appeared to be, despite the fact that the success of any given individual project or technology’s development was essentially unpredictable. Not to sound overly optimistic, but I did find there to be something comforting in the apparent inevitability of progress these graphics presented when contrasted with the uncertainty we see in everyday life (for example: war, oil prices, global warming, American Idol results).
• Ray highlighted the possibility (or possibly the current reality?) of a respirocyte – an artificial red blood cell. Using such a device to replace red blood cells would allow you to do an Olympic sprint without taking a breath, or sit on the bottom of your pool for four hours!
• Advancement in state of the art brain scanning technology is starting to bear fruit, providing researchers with insights into how the brain processing information. One example of success is the work Ray and his team did in reverse engineering some of the transforms performed by the brain’s audio cortex as part of researching improvements to their speech recognition technology.
• The growth of the Internet is providing a vast data set to use in training pattern recognition systems. For example, Google has developed an English-to-Arabic (and vice versa) translation tool using the large set of texts available online in both languages to train the system. No one on the team developing the program speaks Arabic.
• In a rather impressive demonstration, Ray showed a video of himself demonstrating a speech-to-text-to-speech system that accepted voice input in one language, and output speech in a different language.
• An observation on biological and other systems that seems to apply increasingly in the world of online social networking tools, and other collaborative software: “decentralized, self-organizing systems are inherently stable”.

I found Erik Drexler‘s presentation to be somewhat confusing – he provides abundant examples of recent success in manipulation of DNA and proteins into arbitrary structures, and yet seems dedicated to a vision of a world where we manipulate individual atoms like so many Lego blocks. It would seem to me that the biological route is more likely, and in the nearer term will achieve the end goal of humanity being able to build whatever they want with arbitrary precision. I just can’t understand why Drexler is so dedicated to the idea of applying existing mechanical principles on the molecular scale, when manipulation of biological systems appears to be a more ready-made solution.

Cory Doctorow delivered a tailored version of his standard speech on the evils of Digital Rights Management, spinning his argument to focus on the need for users to remain in control of their technology. I thought his best point was that doing so provides the raw material for future innovation. For example: he proposes that consumer’s investment in unprotected CDs has paid an enormous ‘dividend’ in the form of the mass of unrestricted innovation resulting from the development of portable MP3 players (driven primarily by user’s ability to fill said devices with the contents of ripped CDs); the success of these devices drove the miniaturization of storage technology, large scale collaborative filtering web sites, online music stores, etc. This innovation is not available to the users of DVDs because of the provisions of the DMCA that prevent them from ripping their movies to use with other devices. Food for thought for legislators that appear more interested in protecting the current media regime, than trading up for a much greater public good.

Bill McKibben brought a welcome element of calculated sobriety to the discussion, questioning the drive towards additional technological development. In particular, Bill questioned whether or not our advanced technology have made us any happier. By analogy, he questioned the idea that more is better – if more is better, faster is better, why then would it be worthwhile to run the Boston marathon? We could just as easily drive the distance with our technology – but getting there was not the point! The arguments for pursuing the Singularity, in his opinion, sometimes sound like observations shared by concert-goers on the way out of a Phish concert.

A few items of note on the other presenters:

• Douglas Hofstadter followed Ray Kurzweil and delivered what can only be described as a thinly veiled personal attack on Ray’s book. While I believe his central thesis, that scientists in general are not considering the ramifications of the Singularity and taking it into consideration as they advance research, was reasonable, his style of delivery marred what might have been an otherwise constructive presentation. Though he lampooned Ray for “hand-waving” too much, Hofstadter provided little substantial material to support his own claims. A disappointing and somewhat unproductive session.
• Nick Bostrom provided a moment of humorous respite in his presentation considering possible changes that result in a catastrophic end to the human race – one of his possibilities: “We are living in a computer simulation and it gets shut down”.
• Sebastian Thrun showed a number of highly entertaining videos from the DARPA Grand Challenge. He presented a compelling reason for the importance of this work that I hadn’t previously considered: the potential to save significant lives and improve quality of life for the elderly by automating driving. Of particular note: there are 42,000 deaths due to automobile accidents in the US each year – equivalent to the number of deaths in Vietnam, and 3X (though I question this number) the number of deaths resulting from September 11th, 2001 – imagine eliminating these deaths through automated driving.

All in all, an excellent event – not the standard fare for tech events of late in Silicon Valley. My thanks to the Singularity Institute for putting on the event. See Renee Blodgett‘s blog for more in-depth coverage of the summit.