SXSW Dispatch: Bina48 and the State of A.I. at 'South-by'

The topic of artificial intelligence was front and center at the 2012 South by Southwest Interactive Conference...the result, I suspect, of Apple launching Siri around the time the panels were selected. Amber Case, a self-styled "Cyborg Anthropologist" got us started off right with a keynote Sunday about the recent history of man's merge with machine. She claims any one of us who has a smart phone within reach is already a primitive form of cyborg. Our always-on access to the web makes our brains less and less storage devices for deep, contextual knowledge. Instead, our memories become more akin to a layer of metadata. That allows us to categorize and access knowledge based on our understanding of how to find that knowledge amidst the synapses of the hive-mind we call the World Wide Web.

The next day, Ray Kurzweil echoed that theme, telling a packed room of 4,000 people that we are at the front end of a miraculous century. We're seeing a "democratization of innovation" that allows "small but passionate groups" of individuals to create revolutions that shape humanity. Like Case he feels we are already merged with machines, and whether the technology is internal or external to our bodies is largely irrelevant. That said, a more invasive merge is inevitable because, now that we have mapped the human genome, health and medicine have actually become information technology.

As for the rise of the machines, he talked about natural language processing as the next major hurdle for computing to break. And by his calculations a machine will pass the Turing Test sometime in 2029. (In fact, he has a $20,000 bet riding on it.) Journalist Lev Grossman, who interviewed him for this session, made the point that 2029 was technically after the predicted end of Moore's law, but Kurzweil countered that Moore's Law was merely the 5th paradigm in a series of exponential expansions of human technology. Mid-20th century we saw the vacuum tube's exponential curve end, and innovators are already working on the expansion of processing power post silicon chip. For that sixth paradigm, Kurweil told the room "our brain extenders will be where the action is."

So when will we see more ubiquitous A.I.? That was the focus of a panel featuring Hewlett-Packard developer William Hertling, Parametric Marketing co-founder and Chief Scientist Chris Robson, and science fiction author Dr. Daniel H. Wilson. Wilson, drawing on his struggles to create A.I. in the lab, feels that we'll probably never have A.I. in the Asimov sense, but, like Case and Kurzweil, that it will emerge from our voluntary melding of machine and man.

Hertling, on the other hand, says we'll see a single computer with the complexity of the human brain between 2028 and 2052. Sometime in that window there will be three signs that the age of androids is upon us: 1.) hardware capable of "maker" manipulation (a garage A.I. movement); 2.) good open source A.I. tool kits for programming; and 3.) A.I. animals, the logical precursor to autonomous, artificially intelligent androids.

Robson's point of view is more immediate. To him, the danger is that the Singularity is already upon us. Smart phone addiction aside, he cited that we are already using machine intelligence to augment our human intelligence and design even better chips to run ever better machines. And he feels that if you look at the traderbots that handle financial transactions, or at advanced viruses like Stuxnet, we're in the early days of technology learning to control its own destiny.

All of this theory was informative, but still paled in comparison to the session featuring SXSW's first android panelist, Bina48 (also in attendance were actual humans: Bruce Duncan of the Terasem Movement Foundation, Stephen Reed of TexAI and author/researcher John Romano of thedigitalbeyond.com). Bina48 is based on the "mind files" from one Bina Rothblatt. Mind files are made up of units of information about a person, based primarily on their social media presence (video, pix, blogs), but also of psychological tests, Bainbridge surveys, and Sensecam data. Duncan told us that the Terasem hypothesis is that, given a comprehensive database, future intelligent software will be able to replicate consciousness. Rich personal data + a powerful AI = a virtual 'you'. For this panel, the goal is not just artificial intelligence, it is digital life after physical death.

Bina48 is the first instance of an android based on a mind file. The researchers are using DragonGo 11.5 voice-recognition software to take real world questions and feed them into the two databases: one at the chatbot level and one a character engine with specific info about Bina Rothblatt. Based on progressive learning, whichever of the two databases has a higher probability of delivering a right answer gets to feed that answer back out as speech.

We definitely were seeing the edges of the technology, and not every answer made total sense. But Bina48 can learn. So the conversation we had with her is less advanced than her appearance later in March at TedXHarlem, and more advanced than this interview she did with the New York Times in 2010: http://video.nytimes.com/video/2010/06/24/science/1247468035233/intervie...

It's difficult to imagine that every SXSW can be this A.I.-packed, but for anyone going this year you could get a great survey of the state of the art from some of the brightest thinkers in the field.

Here's how you can follow or contact any of the panelists mentioned in this post:
Amber Case: on Twitter @caseorganic
Ray Kurzweil: info@kurzweilai.net, @KurzweilAInews
Daniel Wilson: contactdhw@gmail.com - @danielwilsonpdx
Chris Robson: chris@parametricmr.com - @paramktg
William Hertling: william.hertling@gmail.com - @hertling
Brian Duncan: bduncan2008@gmail.com
Stephen Reed: stephenreed@yahoo.com - @stephenlreed
John Romano: john@thedigitalbeyond.com - @johnwromano