CYC-ology - Using AI to Organize Knowledge

Doug Lenat, an artificial intelligence researcher and CEO of Cycorp, a company that aims to build general artificial intelligence, gave a talk at the Center For Inquiry Austin. He examined why AI is so difficult to create, and how CYC is approaching this task.

Why haven't we been able to create a program that would pass the Turing test, i.e. be able to converse in such a way as to be indistinguishable from a human? For a large part it's because human thinking is faulty in ways that are very hard to approximate in software. Doug Lenat calls these idiosyncracies of human thinking translogical behaviors. Those are illogical but predictable decisions that most people make; incorrect but predictable answers to queries. Lenat listed some of those behaviors in his talk. He also addressed them in his article, "The Voice of the Turtle: Whatever Happened to AI?" (PDF). Here are some examples, compiled from both the article and the talk.

Flawed memory and arithmetic ability: while a human will correctly tell you what day of the week was yesterday, he or she will most likely be wrong if asked what day of the week was April 7, 1996. For the same reason, humans are likely to give wrong answers to math problems, but certain answers are more "human" than others. 93 - 25 = 78 is more understandable than 0 or 9998.

Conjunction Fallacy: Most people will say "A and B" more likely than A. For example, asked to decide which is more likely, "Fred S. just got lung cancer" or "Fred S. smokes and just got lung cancer," most people say the latter.

Incorrectly estimating probabilities of various events. People worry more about dying in a hijacked flight than the drive to the airport.

Failure to discount sunk cost; also, skewed perception of risk and reward. People estimate risks and rewards very differently if it means losing something they already had, as opposed to investing into something they don't yet have.

Reflection framing effect. Let's say, before adopting a certain public health program (e.g. medical screening), 500 people a year used to die from whatever this program is supposed to prevent. If you market it to the public as saving 200 lives a year, many more people will vote for it than if you say "300 people per year will die".

Another example. In two neighboring countries organ donor rates are 85% and 15%. Lenat asked us to guess the cause of this drastic difference, considering that the two countries are very similar in their socio-political and economic situation. It turns out, the only difference is that when you get a driver's license in the country A, you have to opt-in to be an organ donor by checking a box on a form; in country B, you have to opt OUT of it, also by checking a box. When opt-in is the default, 85% people opt-in; the reverse is also true. 85% of people just don't bother to check the checkbox either way. Who would have thought?

Failure to understand regression to the mean. That's a kind of translogical thinking I found the most poignant. Many parents punish their child after he or she gets an abnormally bad grade, and reward them after getting a good grade. However, after an unusually bad grade, the next one is statistically likely to be better without any punishment or reward. Similarly, after a good grade, the next one is likely to be worse. So parents who react to grades with punishments or rewards, get an idea that punishment works, but rewards don't. Historically it explains a lot of cruelty among humans, says Lennat. In reality, he believes, nothing really has any effect on human beings.

Scott, Steve Bratteng (center) and Doug Lenat (right) chat after Doug Lenat's lecture "CYC-ology - Using AI to Organize Knowledge" at the CFI Austin.

Despite these uniquely human weaknesses, people can easily make inferences about the world that computers can't. Even the best search engines today are falling short of putting together simplest facts about the world, and drawing conclusions from them. If you ask Google "Is the Space Needle taller than the Eiffel tower?", you'll get tons of pages that give the heights of theose objects, but no page that tells you which one is taller. You also won't get an answer to "who was U.S. president when Barack Obama was born?", because search engines still can't string together two facts: Obama's year of birth, and the identity of the president that year (John Kennedy). Today's search engines only handle syntactic search, while these queries represent examples of semantic search.

As we saw, human reasoning strengths and weaknesses are not the same as AI reasoning strengths and weaknesses. There is an opportunity for synergy here, says Lenat. What is the missing piece to bridge that chasm? CYC is ready to bridge it with their (a) ontology of terms, which includes over 1 million of general terms and proper names, (b) knowledge base of general knowledge, which is 10x size of the ontology, and includes such facts as unsupported objects fall; once you're dead, you stay dead; people sleep at night; wheeled vehicles slow down in mud, and (3) a fast inference engine. So for example, if you query CYC system for an image of "someone smiling", it will retrieve a picture with a caption "A man helping his daughter take her first step". The system achieves this by putting together the fact that when you are happy, you smile, with the fact that you become happy when someone you love accomplishes a milestone.

(As an aside, I think this level of sophistication is easy to foil. Human emotions are more complicated than this example describes, and someone watching their child take first steps could easily have tears in their eyes. So an AI would have to know that there is such a thing as "tears of joy". But how would it tell between those, and tears of sadness? An AI would have a long, long way to go before it could recognize similar emotional nuances.)

So are we making any progress towards AI? Doug Lenat believes that the current semantic web "craze, fad, or trend" (his words) is moving us in the right direction. Instead of syntactic searching like Google is doing now, in a small number of years we might be able to see semantic searching. What would be the signs that our software is becoming more intelligent? Look for speech understanding systems, like Dragon Naturally Speaking, to stop making dumb mistakes, says Lenat. When they no longer garble your words in ways that a human would never misunderstand, that would be the sign that speech recognition programs have some semantic awareness.

Are we on the road to Singularity, then? Nobody in the audience asked Lenat this question outright, but he admitted he believes it's only a matter of time until artificial intelligence crosses human intelligence.

Ambivalent towards Google Wave

Apropos my previous post): I got a Google Wave invite, but after logging in I haven't done anything with it. I don't know anyone I could "wave" at, and am not very motivated to find such people -- or to wave at them. It's hard enough just to keep up with email, Facebook and Twitter. I communicate with most of my real-life friends over Facebook, so Wave would be duplication of that functionality. If, however, I started "waving" at a completely different set of people, that would double the time I spend on my virtual social life, and I can't afford that. So I haven't been doing anything with Google Wave.

But I saw this great article, What problems does Google Wave solve? by Daniel Tenner, where he argues Google Wave is not so much an enhancement of your social life, as a corporate collaboration tool. This ties back to what Tristan Slominski said in his Innotech presentation (see my previous post). I can see how a Wave-like IDE plugin might enable programmers to work on shared pieces of code. Then again, many if not most companies would not let their code past corporate walls and firewalls. To route confidential information, or even source code, through third-party servers is considered unthinkable in most companies.

But if Google Wave is so great for collaboration, I can think of some uses for it. Perhaps it could help me and my friends to work on our fanzine more efficiently. (I spoke about our fanzine in this post.) Preparing each issue of the fanzine involves bouncing documents back-and-forth multiple times between authors, translators, copyeditors, and illustrators. So if Google Wave really works as advertised in Daniel Tenner's article, it could be useful for that. The only problem is I would have to convince my friends to use it. For all I know, they might feel the same kind of resistance to it as I do.

Innotech 2009

Still cursing Austin's prehistoric self-pay parking lots, I walked into the Innotech Beta Summit panel, a showcase of select Austin startups. To my surprise, parking meters were mentioned there. A representative of Infochimps, a company that specializes in "making large data sets sexy", said a collection of locations of parking spots in downtown Austin is an example of those sexy data sets. (Another example is TAKS scores.) Anyone can put a data set on Infochimps web site, and if some organization or person is interested in it, they can buy it. Well, I'm sure glad some company is interested in locations of parking meters in Austin downtown. Whatever they are up to might make parking easier one day, who knows?

Wouldn't it be nice to have an application that showed all parking spots within a certain radius? It could direct me to a nearest free spot as I wind my way through downtown, looking for parking. But maybe such an app for iPhone already exists? I don't have an iPhone (and now that I've been laid off I don't anticipate buying one soon), so I don't know.

I expected Beta Summit to be the most interesting Innotech panel, and it was. It featured 6 or 7 Austin startups. The first one, BuzzStream, did not impress me all that much, perhaps because I came to the talk late. The only impression I got from it was that it did some kind of fancy contact management, integrating your contacts with social management sites. Or perhaps it was yet another social media aggregator, the kind that gathers all the content your friends have posted on various other sites, into one news feed. We all have seen social media applications that claimed to be the aggregator to end all aggregators, and then a few months later no one remembers them. Plurk comes to mind.

The next presenter was Gendai Games. Their product lets you create an iPhone game in minutes, even if you are not a programmer. As a demo, Nestor Hernandez recreated the game Labyrinth in front of the audience. He did that he dragged widgets, such as Accelerate and Collide, on the screen. Those widgets made a ball accelerate when iPhone is tilted, or bounce off the "walls", i.e. the sides of the screen. Pretty neat.

Eric Moujaes from Gendai Games	Tristan Slominski speaking on Operational Transformation: The Key to Understanding Google Wave

One web application everybody could relate to was Gelato, a dating site that works differently than most dating sites. On an ordinary dating site members have static profiles that are often misleading or uninformative. People often post younger and thinner pictures of themselves. Steve Odom, the founder of Gelato, believes that people's social media data streams reveals much more about them than their self-proclaimed love of "long walks on the beach and candlelight dinners". Their Tweets, Flickr photos, YouTube videos or Netflix queues, or soundtracks on internet radio stations they listen to, reflect a much fuller, dynamic picture of their tastes and their preferred ways to spend time. So Gelato aggregates all that into a user's profile. Is that a bit too stalkerish? Users can opt out of whichever feeds they prefer others not to see.

As a demo, Steve Odom pulled up a woman's profile on Gelato. "Would I want to date her?" he asked the audience. Her Twitter word cloud (another neat feature of Gelato) showed that her most commonly used word was LOL. "Hmm, maybe not," Odom concluded.

Another web application of broad appeal was PetMD. It's analogous to medical information sites, only about pet health. It also helps you find a veterinarian, even an emergency vet if need be.

There may have been other startups, but I left early.

This conference was a mix of technical and soft presentations. It's always hard to know which ones to choose; an appealing tile can be misleading. So there was some kind of "work/life balance for entrepreneurs" panel, where a self-proclaimed work/life balance coach did nothing but slung cliches about success like "you are your own worst enemy"; I spent half an hour before concluding it was BS. But by then I missed the first half of "Operational Transformation: The Key to Understanding Google Wave" presentation by Tristan Slominski. When I came in, he and the audience were up to their eyeballs in the APIs. Operational Transformation is kind of like a platform on which Google Wave is written. Knowing its API, you can write your own clients that will be able to communicate with other Operational Transform clients. For example, Slominski says, those could be plug-ins for IDEs (i.e. development environments -- tools in which programmers write, compile, build and test their code.

Innotech, or a digression about parking spots in downtown Austin

Innotech is an annual one-day technology conference in Austin, TX. This year it took place on October 29. Before I can speak about it, I should talk about getting there, and the most complicated part of getting there is finding a parking spot in downtown Austin. The process of searching for one also provides the greatest thrill you are likely to have at Innotech, if by thrill we mean a nervous rush. I used to think of it as my personal deficiency that finding parking in downtown Austin (or most other cities, for that matter) stresses me out so much. But here's an article "Why Speakers Earn $30,000 an Hour - Confessions of a Public Speaker"> by Scott Berkun, where he says finding the right address and parking in unfamiliar places is stressful! Ah, I feel so validated.

The parking garage at the Austin Convention Center was full at last year's Innotech; this year I didn't even bother to check it. And street parking often has 3 hour limit, and costs $1 an hour. A bit steep for the whole day. I thought I'd park in a public parking garage, but they seem to be on every corner when you're NOT looking for one, and damn hard to find when you are. Every garage appears to be reserved for employees of that particular office building, or if they are open to the public, they charge thereabouts of $20 a day. Steep.

Downtown Austin also has a scattering of paid parking lots; it's been a bit of a mystery to me why those parking lots aren't used by offices or restaurants attached to it. Maybe I haven't been paying attention, and there is nothing attached to them, or whatever it was has closed or was torn down. Often those parking lots don't have attendants. What they have instead are kind of vertical boxes with slots you should stuff money into. They take coins and bills. There is also a piece of metal hanging from a string that you should use to stuff coins and bills into the slot, should they get stuck just inside the slot.

Fortunately, parking at one of those lots costs only $5 a day, so it's a good deal. This year I stuffed my $5 bill into my slot without difficulty, but last year was a different matter. I didn't have enough bills, but I had a purse full of quarters back from the days when I used a paid laundromat at an apartment complex. I don't have much use for those quarters, so I stuffed 20 of them, one after another, into the slot, and I had to wiggle that metallic stuffer really hard. A guy who parked in the same lot at the same time, asked if I was going to Innotech. He was going there too. However, he didn't have enough $1-$5 bills or coins. So I fed quarters into his slot too (as I said, I don't have much use for them). He was thankful and said it must be true what they say about people being friendly in Austin. (He had moved here from California.)

Just as last year, I wondered how ironic it was that I had to use such an outmoded, awkward way to pay for parking before I could get to a conference on all things high-tech. Besides established tech companies, this conference also features selected Austin startups. I wished any of those startups that gave 8-minute talks on the Beta Summit panel had poured its energies into technologies that would let you pay for parking with your cell phone. That has to be possible, right? I've heard it's already possible in some parts of the world. And you don't necessarily have to install fancy, expensive parking meters that would interface with your phone directly. You could simply pay by a text-messaging the company that owns the parking lot.

But now that I thought about it, I'm not sure companies who own parking lots would want such a thing. They don't have to care about convenience for customers, because they don't have to compete for customers. Parking is very hard to find downtown. Space is limited and it won't grow magically. People will put up with inconvenience just to get a parking spot, because what choice do they have? So probably nothing will happen, unless it could be somehow demonstrated that companies would save expenses by adopting a more efficient way of collecting payments for parking.

From here I can seque to Innotech. That will be my next blog post.