Tuesday, March 14, 2006

Conceptual AI

Jantunen talks about artificial intelligence, which is capable on bringing so-called "technological singularity". This means that the AI is human-like and intelligent enough to perform various informational tasks, which require independent assimilation of facts. When the AI is sophisticated enough to design software and hardware, we can tell it to design a better AI, and so we have reached the technological singularity.

First, I characterize what I mean by conceptual AI, and then I describe how it may become reality.

This writing is about conceptual AI; an artificial intelligence, which can communicate with humans in natural language. It can also think with concepts, which are not hard-wired. Although current state-of-the-art game AIs can beat most humans in many areas (Deep Blue, Masters of Orion, ...), they wouldn't be conceptual AIs even if they had a natural-language interface, since they can only handle a finite set of concepts. The game AIs couldn't use new concepts to analyze the situation.

The conceptual AI can be divided roughly to two parts: the language unit and the domain model. The language unit could analyze textual descriptions of the domain model. It would try to map the concepts of the natural language to understand what domain model phenomena they denote. It would recongize and handle various linguistic problems: indetifying what pronouns, substantives and verbs denote, recognizing ambiguous exrpressions, guessing the standard of comparison when somtehing is described as "high", "bad" or "special", assimilating new concepts based on definitions, and recognizing when the other end is using a concept in a wrong way.

The domain model would contain the mechanics and dynamics of the subject matter. For example, if the topic of discussion is stock markets, the domain model would contain a database of financial information about the companies, about their production methods, dynamics of the industries, recent news events affecting the economy, etc. This domain information would be structured in a format, which is suitable for the AI. Some parts of format would depend on what concepts the AI considers relevant, and the format would be open-ended to enable extension when new kind of information is hypothesized to be relevant.

The conceptual AI wouldn't contain emotions, goals (except answering the user's questions and performing the tasks given to it), self-interests, visual memory (except if the domain model requires visual thinking), nor other human characteristics.

How Could Conceptual AI Be Developed?

I'm going to handle 3 scenarios; to reject 2 of them and give green light to one.

Rejected: Expert system is generalized into a Conceptual AI. The scenario is basically that initially there is a special-purpose AI, which has a complex and extensive domain model (expert system). The expert system is sharpened with a restricted natural-language interface to make it more usable and accessible for the people who need it. Gradually, the natural-language interface is extended and made more flexible, and metaconceptual elements are added. Why not credible:I haven't heard about expert systems, which could answer to a wide variety of different questions, nor questions which they are not designed to answer. Even if the domain model is complex, it still can probably handle only a finite types of questions. Therefore, there is little or no benefit from the metaconceptual lanugage module. The flexibility wouldn't be useful because of the rigid domain model.

Rejected: Data mining. The scenario is such that we have a huge database of numerical and textual data on some topic. We also have means to do some kind of preliminary analysis for the data: to put it to a relational database in nice tables, to parse the textual data, etc. First, a straightforward natural language interface is added to answer simple queries about the data. Secondly, the AI is sharpened with metaconceptual facilities. It is trained to recongize "normal" levels for various attributes based on distributions, so it can also recognize "higher than usual". It is made to understand definitions. It is made to point out trends and outlier cases. This time, there is both a advantage from the metaconceptual facilities. Why not credible: My impression is that the development is towards more flexible query languages. SQL (the current standard database query langauge) could be replaced by prolog-based query languages, providing a lot more flexibility and deduction. Therefore, there is little incentive to put a natural-language interface, when specialized query languages still have a long way to go.

Probable: Tech support. Suppose that we are talking about an ISP support phone. The AI would be developed gradually:

  • Phase 0 (where we are now): Asks the user to press 1 if the problem is in broadband connections, 2 for phones and 3 for other.

  • Phase 1: Asks the user to describe the problem. Uses the information to redirect the call to the person, who has handled this kinds of problems before.

  • Phase 2: Asks the user for some initial information to search the customer information. Uses the description to checks if the call reports a problem, which has already been reported by someone else.

  • Phase 3: The AI is extended to solve very simple and but common problems.

  • Phase 4: The metaconceptual side is extended so that it can better understand ambiguous descriptions and provide some initial information before the call is redirected to a human. This way, the queueing time is less wasted.

  • ...

  • Phase 20: The domain model is extended with information on how specific applications (Word, etc.) work.

  • ...

  • Phse 56: The domain model is extended with a computer. If the user describes a problem, the AI can try to reproduce the problem in the computer.

In this scenario, the incentive for a good metaconceptual unit is great right from the start, since user descriptions are seldom very clear nor logical. Secondly, each additional step towards more intelligent AI is economically justified, as it directly replaces human effort. The AI is useful right from the start, and there isn't any discontinuity point or big gap, where a big technical leap is needed to make the AI more useful. Therefore, there is constant incentive to take small steps towards a much better AI.
Why it hasn't started yet:This is possible only after a suitably robust speech recognition technology is available. The speech recognition technology must be able to recongize normal speech practically without errors, although it can ask the user to talk clearly. When the speech-to-text conversion is avalable, the work can start.

Sunday, March 12, 2006

Habbo Hotel Goes Mobile

Last Tuesday, Sampo Karjalainen, a co-founder of the Sulake Corporation that maintains and develops Habbo Hotel, kept a lecture in Tampere about the past, present and future of Habbo. Habbo is virtual hotel, where people come to chat a bit like in IRC. The participants are represented by avatars, which walk in the 3D hotel rooms. The users see the rooms in the same perspective as in Populous, newer Civilizations or Settlers, and not in the first-person-shooter perspective. From the communication point of view, the main restrictions compared to IRC are that (1) one person can only be in one room at a time, while IRC allows many channels, and (2) the messages are shown in the top of the avatars, and not logged.

Gradual Development

The current version of the Habbo software is running localized hotels in ~20 different languages. However, the software started local and was developed gradually with many intermediate steps.

Seven years ago, the first prototype was located at the website of a band called Mobiles. From the screenshots Sampo showed it seemed to have 2 rooms, where the avatars mainly walked around. Mobiles was a non-commercial project and was not advertised, but it established a user base and demonstrated that there is demand for this kind of service.

The second version, and the first commercial version, was a Radiolinja advertisement game named Lumisota (snow war). There, the avatars first teamed up inside a skiing hut, and then went outside to throw snowballs at each others. The third version was Hotelli Kultakala (Hotel Goldfish), a Finnish virtual hotel.

The presentation was non-technical, so it is not possible to track down exactly what features were added at each point, and how the code size, complexity and technologies developed. Probably most of those facts are trade secrets anyway. The prototyping clearly had a dual role both in making the technology mature and finding out the target group and business model. I got the impression that there weren't any big risks taken; each step forward was clearly justified by extrapolating experiences from the previous prototypes.

Small Torrents Create a River

In Habbo Hotel, the users can get their own room for free. Initially, the room is empty. The user can buy furniture to make the room less hollow. The main sources of are advertising and the sales of Habbo furniture. They also sell ringing tones and logos.

The small unit price of the furniture items requires a solid micropayment solution. The presentation said that Sulake has 160 local partners in arranging the micropayments. I didn't ask how automated the purchasing was: it would have been interesting to know how many systems were directly interfaced with the Habbo engine so that no human intervention was needed in purchases, and what kind of difficulties had to be overcome to make it possible.

The estimated income for year 2005 was 30m euros. There were 5 million different users each month visiting the hotel. This means that the average user produced 6 euros in purchases and advertising income. It's amazing that a whole company can run on personalization - on the willingness to make the environment reflect the identity - and more specifically on virtual personalization.

Smells Like Teen Spirit

The age group 12-18 contains 90% of the Habbo users. Sampo pointed out a central double standards with regards to privacy and censorship in the Internet: With adults, listening to other peoples' conversations is spying, a violation of privacy, a threat to freedom of expression and in general a bad thing. With children, not watching the conversations is a comparable sin, since it exposes the underaged to paedophiles, drug dealers, foul language and bullying.

Sulake was actually contacted by Kuluttajavirasto (Finnish Consumer Agency), and they had to explain what precautions they take to protect the users. They have a special moderation system, and English conversations are watched 24h/day, while local conversations are watched only during times when there are some people in the hotels.

Mobility And Its Discontents

The first link to mobile phones was the micropayment by SMS. There was also a failed attempt to make WAP client to Habbo.

With the arrival of Symbian Series 60 phones, with the performance of native C++ applications (as opposed to Java Micro Edition) and socket connections, porting the Habbo environment became a realistic opportunity. Writing the Symbian client started in 2004 and was gradual. The first phase ensured that the phones are powerful enough to run the visual interface, which they were. The second phase concentrated on solving network-related issues.

Sampo showed an impressive demo, where he run the Habbo environment on a Symbian phone. The screen was big enough to show several avatars. The scrolling was smooth, and there was no visible lag. When the user typed, a red ball appeared on the top of the avatar to indicate that he or she is about to say something. It seemed like most technical problems in the client side were solved.

There were also some Habbo single-user games. The games were warm-ups of simple arcade and puzzle games of the past. The link between the games and the Habbo hotel is the Habbo brand, and they have little to do with the hotel environment. Apparently the intention is to use the huge user base to efficiently advertise the games.

I asked how they tested their software on wide number of mobile devices and languages. They co-operate with Universum to test Java applications. The Symbian application hadn't reached the point were this kind of testing is needed, so didn't get any tips on that.

The Mobile Context Of Use

The difference between the computer and the mobile phone is not only in screen size, keyboard, computing power and network conection, but also in the way people use them. Mobiles are ideally suited to be used in various waiting situations: trains, classrooms, meetings, pauses between events. Another typical feature for mobiles is partial attention: If a person is using a computer, he or she probably sits in front of it and notices quite quickly what happens. On the other hand, on mobiles the person may be doing something else, and only sporadically watch the Habbo interface. This creates the problem that avatars guided from mobile phones seem to be passive lampposts most of the time.

This had at least two consequences. Firstly, they intend to make a separate hotel for mobile users, instead of putting them to PC hotel. Secondly, the shift of emphasis from the Habbo groupware to single-user games reflects the problems of the mobile context of use - they don't consider mere discussion to be fancy enough in itself in the mobile context.

Monday, March 06, 2006

Limperin Hilma

From the point of view of Panu-style ATM activism, Limperin Hilma is probably the most realistic song about the relationships between men and women. Unfortunately my poetic abilities are not good enough to translate this hilarious song to English.

Vieläkö muistatte Limperin Hilman
koulumme oppilaan niin herttaisen
mukanaan aina toi kauniin hän ilman
mietimme ken tekee valloituksen

Anttilan Augustin kaihoisat laulut
herkkähän mielehen tehosi kai
lipevä kieli ja korkea kaulus
parhaimmat pisteet ne kilvassa sai

Hilma se hääräili pellolla illat
päivisin tehtaalla työskenteli
********** kehräsi naapurin villat
Että hän mökkinsä lunastaa voisi
Augusti työstä ei piitannut lain
mielummin Augusti laulais ja joisi
eläisi helpossa maailmassa.

Vieläkö muistatte Limperin Hilman
vihdoinkin rauhassa levätä voi
ruohoinen hauta on ristiä ilman
Augusti mökkinsä viikossa joi.

A note about the translation: "Aukusti" (August) is a Finnish male name.

Do you still remember Hilma of Limberg
pupil in our school, most pretty and smart.
Her presence always brought weather a glimmer
we speculated who'd conquest her heart.

August's emotive songs and harp
made deep impression to poetic mind.
Collar so high and tongue so sharp
guaranteed best scroes for lifelong bind.

Daytime she was in factory employment,
fieldwork was how she evenings spent.
(don't remember original, can't translate)

She worked hard to pay the debt for the house
August turned out to be a full dork
He dedicated to music and booze
preferred to live in a world without work.

Do you still remember Hilma of Limberg
at last she's getting the rest that she needs.
August sold house and spent all on beer
grave has no cross and is growing weeds.

Sunday, March 05, 2006

The Morals Of Narnia

When I went to see Narnia, I expected a story with some kind of christian morals. What I found was a mere Disney spectacle.

The movie started very well. The opening scene is a prime example of how action scenes should be written. It introduced the characters, and the tension between the defiant little brother and the responsible or bossy big brother. It was an integral part of the plot - motivating the move to the countryside - and not just a fill-up whose signifigance boils down to who dies and who survives. And it was short.

The quality of the movie decreased linearly with the time. The final battle scene was a prime example of how not to write action scenes. It revealed no new information about the characters. The solution was known all the time. It was artificially lenghtened. I often close my eyes during such action scenes, and open them when the noise goes silent. Not once have missed important plot events this way in any movie.

Anyway, back to the moral side. The moral choices of the characters consisted mainly of yes/no choices: "yes" continue the adventure, and "no" to return from Narnia to the common world. There were also some multiple choices. The characters didn't plan, didn't try to predict consequences nor predict risks. They were spectators who watched the story unfold. Even christian moral is more fine-grained than the moral of Narnia.

Next, I'll define a moral imperative, which is consistent with transhumanist ideology. I call it "techno-optimist moral imperative".

A pareto improvement in efficiency is a development, which increases the productivity of an employee without making his or her job more straining or irritating. For example, digging a gutter with a digger machine instead of a shovel is such an improvement.

Two things are required to turn the misery of the past into our current comfy Western civilization: pareto improvements in efficiency and political will to spread the bigger cake to masses in the form of democracy, rule of law, education, healthcare, etc. (You could add that a third requirement is some kind of middle-class attitude to life. I'm not going to dwelve into that.)

Since the implementation of rights (I mean real rights, like right to vote, right to a phone call, various contractual rights, etc. and not the kinds of subjective rights that are often ridiculed in other Erektus blogs) requires pareto improvements in efficiency, and the adoption of pareto improvements in efficiency usually enhances the rights of the masses in the long run, it is reasonable to rise these improvements into a moral imperative:

Techno-optimist moral imperative:

It is immoral not to adopt a pareto improvement in efficiency.

Tommi said in one of his writings that he seldom encounters moral choices. Multiple-choice moral decisions have been removed from the life of ordinary citizens, and outsourced to judges, jurys, members of parliament, those who prioritize various life-critical services, etc. The techno-optimist moral imperative concerns a different set of people: those who can influence their own work.

This moral imperative establishes a tight link between knowledge and morality. In order to know if some modification in the way you work is an improvement, you must have good knowledge of your own field. Unlike in Narnia, where the magic bow aims to the target by itself, techno-optimist moral imperative requires training and experience. It also requires the persons involved to plan and speculate what the way of working would look like after a change, and how the various side effects of the practice should be dealt with.

In many areas of informational work, the productivity can not be measured accurately enough to detect small changes (like 10% or so). In these cases, it is necessary to resort to secondary signs of efficiency. For example, if some problem is a recurrent source of frustration, then solving it is likely to also improve efficiency. For example, continuous integration may be such a practise. Also, reducing the time required to do some subtask is a secondary sign of efficiency.

Also, the moral imperative doesn't say that you should believe just any person who says that something will be an efficiency improvement. It is quite common that marketers claim that something improves efficiency, while the real effect is questionable.

Lastly, the imperative doesn't talk about the political feasibility of adopting an improvement. Suppose that there are 6 workers in the team, and a practise can be adopted if 3 people strongly agree that it is good, and no more than one is strongly against. In this kind of situation, the percieved benefits of the new practise are of paramount importance. Here the christian moral tradition offers a good motto.

God give me serenity to accept the things I cannot change, courage to change the things I can, and wisdom to always tell the difference.