Failed Screenwriter

For immersive story to work we first need AI to work, but it's the opinion of many that AI development is lagging. A recent post on Network World, Whatever happened to artificial intelligence?, might suggest that it is an issue of public attention.

Yet tools and embedded intelligent systems don't answer the "grand challenges" of artificial intelligence, including robots and language processing. Very few projects have captured the public's imagination.

NASA got great public response with their Mars rovers, but little was made of the artificial intelligence components. Artificial intelligence techniques considered pure research 15 years earlier guided rovers Spirit and Opportunity around rocks a world away.

- Whatever happened to artificial intelligence?

Public attention is important for drawing innovative people to the field and cold hard cash. Look how well NASA has done as the public bored of the space shuttle. Attention is also important in the development of AI in the way human interaction can accelerate development. Ben Goertzel's popular post, On the Merits of Parrots, explains how virtual parrots in the metavers can assist development of natural language processing AI. Usually the more people involved the better - Innovative community or research papers. But how exactly are we to garner that attention?

What we need is something sexy that doesn't over sell itself - repeating the common mistake of early AI pioneers. Having it sooner rather than later would help greatly, hence it can't be too far fetched. So maybe what we really need is something cool - sexy might be too much of a stretch. But that is something cool; the wonderful AI that powers Google search might be great but it isn't exactly cool to the average person.

My proposal is to concentrate on natural language processing. I know, how novel of me. Ben Goertzel has already laid out his proposal for virtual parrots, but I think we need to come at it from a different angle - that angle being to involve the artist. What we need is to use AI and more specifically natural language processing to power intelligent virtual characters that can be inserted into virtual worlds, or more importantly virtual scenes or rooms.

Us (developers)

Objective:Provide the artist with easy to use tools to create intelligent virtual characters that metaverse inhabitants can interact with using natural language.

Target user: The artist.

User (artist)

Objective:Create intelligent virtual characters that entertain metaverse inhabitants.

Target user: Metaverse inhabitants.

Empower the artists and they'll make something cool. AI developers get input from both artists and users. From artists AI developers get a captive audience interested in AI development, just from a different perspective. And if artists are successful in drawing a crowd it will mean reams of data for AI developers to churn through.

What will artists create to generate a crowd - doppelgangers of themselves, personal assistants (even if limited in capabilities), or just interesting characters.

Of course requirements for such a project will be a bit different than what most AI developers are used to. Just off the top of my head:

• Need something artists can get there hands into - Artists aren't programmers so the tools need to be WYSIWYG in a way. May seem like a stretch to cram complicated AI logic into a WYSIWYG interface but I don't see how it can be escaped. It's also possible to hand the more coarse controls over to the artist and keep the finer points beyond their reach.
• The resulting natural language processing doesn't have to be perfect - Short comings will be the responsibility of the artist to cover up. End users will also most likely cut the natural language processing some slack. It isn't speach to text where the margin of error is extremely low.
• Distribution through the metaverse, preferably virtual scenes or rooms - In theory these intelligent virtual characters could be distributed across the metaverse in virtual worlds, however I think they are more suited for the new scene or room craze sweeping the metaverse. By focusing on the natural language aspect, these characters would probably be limited in other aspects and a room nicely confines them. Rooms are also lighter and can be pushed to where the users are, as Vivaty does with their Facebook widget. Cobalt is also working on a browser plug-in.

Of course this isn't far from a chatterbot and those have been around for years, not generating much excitement. But I think the metaverse presents a unique opportunity to dust the concept off and reinvigorate it with an infusion of modern AI. The simple addition of a 3D body and context might do wonders.

I was happy to read about OpenCog on Ben Goertzel's blog. If you're not familiar with Ben Goertzel, he's the CEO of Novamente and the biggest proponent of artificial general intelligence, AGI. I first met Ben when he and The Electronic Sheep Company made an announcement at Virtual Worlds Conference October 2007 to collaborate on developing intelligent virtual agents. I found their collaboration of particular excitement for what it might mean for immersive story, however at the time everything Novamente did was proprietary. But like most companies these days, they have opened sourced much of that in OpenCog. One of the reasons for this is Ben's opinion that AI development has fallen woefully behind other fields.

However, we suggest that one viable path to remedying the situation involves providing the field of AI research with an open foundation comparable to the open foundations underlying popular internet applications.

- Overview of OpenCog

Making the move to open source goes along with his general theme to take advantage of the “wisdom of crowds”. But first you need a crowd, and that's where the metaverse and talking virtual parrots come in. The metaverse as a playground and distribution channel, the parrots as a means of direct interaction. He goes into more detail in his post, On the Merits of Parrots; advocating the placement virtual parrots around the metaverse to assist development of natural language processing AI.

For a virtual parrot, the test of whether it has used English correctly, in a given instance, will come down to whether its human friends have rewarded it, and whether it has gotten what it wanted. If a parrot asks for food incoherently, it’s less likely to get food — and since the virtual parrots will be programmed to want food, they will have motivation to learn to speak correctly.

And of course a bigger the crowd means more people interacting with the parrots, and more feedback.

What it comes down to it: A virtual parrot, learning language, will have lots of teachers, and that’s a good thing. The more customers we get for the parrot, the more teachers the AI underlying the parrot will have.

I think bumping into one of these parrots, or intelligent virtual agents, would be great but my problem is I'm not sure how long it would keep my attention.

A baby AI has a lot of disadvantages compared to a baby human being: it lacks the intricate set of inductive biases built into the human brain, and it also lacks a set of teachers with a similar form and psyche to it … and for that matter, it lacks a really rich body and world.

I know a lot of guys who would rather do chores than have to interact with kids, especially if they aren't their own. Once the novelty wears off, it's simply work rewarding what is essentially a virtual baby. If you can appreciate what's actually behind the virtual parrot you might continue, the rest of us will walk on looking for anything else that is entertaining. And that's where the early potential lies for these baby intelligent virtual agents - entertainment. If we need feedback and the more the better, we should try and wrap these babies in entertainment.

The question is how exactly do we go about trying to make these early intelligent virtual agents entertaining to the masses? The obvious solution is to make a game out of it, however I believe games will only invite cheating. A person playing a game wants a challenge, how challenging can a baby be? If we give prizes for winning the game I would think most players would see it more of a challenge cheating the system then to play accordingly.

The solution may lie in the same technique used by movies to mask poor special effects many decades ago - suspension of disbelief. Hook the user with the story and they'll stick with you through almost anything. Create an interesting character and they will cut them a little slack. The burden isn't so much on the technology but on the context. That's where the artist comes in with immersive story - dramatic narrative for the metaverse, where a player interacts in real time with computer controlled NPCs and virtual environments. I guess another way to describe it would be to create dramatic narrative through the manipulation of intelligent virtual agents in virtual worlds. Being a big supporter of immersive story, it's obvious why I'm excited to see projects like OpenCog.

Of course immersive story will have its own set of unique requirements when compared to OpenCog and intelligent virtual agents, but it's potential to draw crowds and fuel the feedback loop Ben Goertzel is hoping to stoke will make it an important piece of AI development in the near future.

Apparently a team at the Rensselaer Polytechnic Institute have created a Second Life avatar that is powered entirely by AI - so more specifically a bot. This isn't the first time I've read about such an occurrence, the metaverse seems to be the latest play ground for AI enthusiasts. I'm always intrigued because I want to use some of this technology in immersive stories, however that's when reality sets in.

At the moment, Edd goes where the team send him and he can answer questions like "Where are you from?" but understands only English that has previously been translated into mathematical logic.

Apparently the bot they created, Edd Hifeng, has the reasoning skills equal to that of a 4 year old. How much can I rely on a four year old to tell a compelling story?

But my disappointment probably has more to do with my lack of knowledge where AI is concerned. And no wonder, AI makes my head hurt - just try to wrap your head around the Chinese room argument. Apparently there are two types of AI, strong and weak. strong AI is where they try to match or exceed human intelligence, as they are trying with Edd Hifeng. Weak AI is much less pretentious, like the AI that powers some of Google search.

With immersive story we aren't expecting the AI to create the story but rather the artist, all we need the AI to do is assist in having it play out. For immersive story all we need is weak AI, we'll let the academics continue to strive for strong AI.

Nicholas Carr's latest book, The Big Switch: Rewiring the World, from Edison to Google is very insightful, however near the end he extends into fear mongering a bit. Some of that fear mongering springs from an interesting point, Larry Page and SergeyBrin's motivation for Google had more roots in AI (artificial intelligence) than being just a search engine.

They weren't just interested in perfecting their search engine, they said. What they really looked forward to was melding their technology with the human brain itself. "You want access to as much [information] as possible so you can discern what is most relevant and correct," explained Brin. "The solution isn't to limit the information you receive. Ultimately you want to have the entire world's knowledge connected directly to your mind."

Larry and Sergey are still pursuing their vision of AI.

"Every time I talk about Google's future with Larry Page," reports Steve Jurvetson, a prominent Silicon Valley venture capitalist, "he argues that it will become an artificial intelligence."

It is difficult to see exactly what scares Nicholas with this, other than general fear of omnipresent technology.

"Why not improve the brain?" Brin muses at one point. "Perhaps in the future, we can attach a little version of Google that you just plug into your brain."

Plugging AI resources into my brain doesn't scare me, if anything I'd be giddy about it as Larry and Sergey are. But what I'd like to see is the AI used not just to assist us, but to power our entertainment as well. And I don't mean in some pie in the sky, where AI constructs interesting stories for us on the fly. No I mean in a more practical way, say as to power language processing for virtual characters, enabling them to process slang and be topical with current events. That could be immersive drama's beach head as the search engine is for Larry and Sergey's AI ambitions, who knows where it could lead.

But alas I doubt Larry and Sergey's ambitions align much with entertainment. Entertainment is soft compared to their math and science backgrounds. Where as I see incredibly interesting data from immersive drama they may see too much ambiguity, but isn't Google's attempt to interpret queries fraught with as much ambiguity? Immersive drama represents a way to technically examine entertainment like never before. Just as we gather click trails for web sites, we'll be able to get the same for immersive dramas. We'll get metrics on engagement, time spent with a certain character, type of responses, etc. - practically everything needed to construct AI as Google has done by taping into people's mind through search. But if still Larry and Sergey can't wrap their heads around the entertainment aspect, maybe games will be the carrot they need. Games today has a little science built up around it, maybe that will capture their interest and eventually lead them to immersive drama.

One area I can see them moving quickly into is the metaverse, but not the one many of us think of when we hear the word. I see them expanding Google Earth, adding avatars, and integrating their AI. Think of this as an intermediate step to tapping directly into your brain with their AI - a digital interface to the AI before it goes direct.

The father of Eliza died March 12. With the creation of Eliza, Joseph Weizenbaum practically invented the field of chatterbots and popularized AI (artificial intelligence) in the 60's. But more than that he demonstrated the capabilities of computers to be more, especially where entertainment was concerned.

Well before the advent of multimedia, there was a moment in the history of the computer that demonstrated its representational and narrative power with the same startling immediacy as the Lumieres ' train did for the motion picture camera. This is the famous but often misunderstood moment in which the first completely computer-based character was created.

- Janet H. Murray from Hamlet on the Holodeck: The Future of Narrative in Cyberspace

While presently Weizenbaum's Eliza is more credited with sparking AI, over the next few years I believe it will be seen more as the central building block of immersive drama. Eliza's ability to create an illusion will prove more fundamental and long lasting to entertainment than AI.

In fact, the responsiveness of the conversation was an illusion, because Eliza was programmed simply to respond to certain key words and phrases.

In trying to develop immersive drama I'm naturally interested in artificial intelligence (AI) and how it might be used to create realistic characters. Given AI today, we can create the NPCs (Non Player Characters) we encounter in video games today. To create immersive drama we will most likely need more from NPCs and consequently AI. So you can understand my enthusiasm when I read posts like, Virtual character with child-like reasoning abilities enters Second Life.

Truly convincing autonomous synthetic characters must possess memories; believe things, want things, remember things.

A pretty tall order, but that's pretty much in line with the field of AI - grand predictions and a lot of failure. But maybe the researchers at Rensselaer’s Cognitive Science Department have figured out a way to make it work. Than again, some of the quotes from the post make me suspicious.

“Declarative definitions of all of the concepts central to a theory of the mind, including lying, betrayal, and even evil,” are covered as well, according to Bringsjord.

Evil - really. Perhaps that is stretching it a bit? The concept to try and recreate the theory of mind makes sense, however when do we cross over from trying to create the illusion of a character and creating a real entity? If Bringsjord's objective is entertainment he might be a little off the mark, especially when you consider a character in a movie is more a metaphor for human nature than true representation.

A character is a work of art, a metaphor for human nature. We relate to characters as if they were real, but they're superior to reality. Their aspects are designed to be clear and knowable; whereas our fellow humans are difficult to understand, if not enigmatic. We know characters better than we know our friends because a character is eternal and unchanging, while people shift - just when we thing we understand them, we don't. In fact, I know Rick Blaine in CASABLANCA better then I know myself. Rick is always Rick. I'm a bit iffy.

- Robert McKee

When I read about AI initiatives like this it reminds me of my issues with interactive story - where the usual approach is more that of a problem solver than artist. Sometimes I question if researchers are aiming more for simulated reality than entertainment and if they're aware of that?

I first met Udi Mander while we were both at Yahoo! Now he's at Google tackling the hard problems of search. How hard - well he recently outlined just what makes search such a hard problem. As interesting as that was, what I found most intriguing and use in the metaverse was what Google was doing to tackle the problems.

• Users and Queries - Ability to understand that 'GM' stands for 'General Motors
• Models - Ability to interpret questions, such as 'How much does it cost for an exhaust system' to ' cost of an exhaust system'
• Languages - Localize the above two techniques

Google is of course in an enviable position, collecting instant feedback from millions of a search every day. So it isn't too far fetched to believe that Udi and Google will overcome the hard problem of search. With that Google can look to further dominate the search market.

But what they're building can be applied to more then just search, and maybe spelling utilities. What if they turned that search engine power over to the Metaverse ? Not just to find stuff but to power natural language support for bots? Box all that in a black box and let bots ping it. Such a black box wouldn't supply responses but would greatly increase the interpretation of user inputs. Responses could be supplied by combining the black box with AIML chatterbot technology.

Community projects could sprout up a la Wikipedia style - letting the community supply appropriate AIML responses. Feedback to the black box could determine which response were found to be the most appropriate. This enhanced black box could be the starting point for bot creators, a default. Creators could create unique responses for their bot and rely on the enhanced black box for everything else, ensuring their bot would never be stumped.

This is just wishful thinking of course as Google has yet to announce their metaverse, despite the long running rumors.