AI Panic!

1.14285714 is Google’s answer when you search for eight days of a week. On the first glance, this doesn’t make any sense whatsoever. However, with some clever deduction, you’ll find that it apparently uses days and week as durations, so days a week would be 7. So 8 of 7 is then the result you can see up there. I think it is very interesting that this Google Calculator output appears funny and a bit confusing to us humans. We’re not expecting to get a floating point number as a result to this query, yet Google’s pattern matching happily identifies a proper formula it can use and number-crunches away.

Although this is just a small oddity that is not even a bug in Google, it is a good example for a much deeper running, fundamental problem not only the big search engine providers have to fight with: How do we recognize what is important in a statement? What is its likely purpose? How can we distinguish the purpose of a query such as eight days a week from five feet in meters? Calling these questions difficult would let a good opportunity to use the phrase hardest problem of our time go to waste. Okey, I know this is an exaggeration, but as far as AI problems are concerned, it gets close.

The definition of understanding is not straight forward, many definitions involve the words concepts,  classification, relationship, awareness and abstraction. In AI, it is often the background knowledge which is the problem. You need to have sufficient background knowledge to understand what is going on, what is meant, what is not meant, or that nothing is meant at all. I think especially this last type of understanding is hard: without a concept of “nonsense”, we’ll have to use the process of elimination to find out that none of the existing concepts fit in our case. I’m quite sure that it will take very long before a computer is able to find the (admittedly overused) picture on the left funny. Even if we discount the difficulty of image recognition, which has an even longer way to go still.

I’m not an expert on the field of natural language processing, but I bet Google employs hordes of the finest minds in this area to cope with the problem of ambiguity, meaning and the understanding of short utterances. No doubt they (and all the other search engine providers) have the best algorithms in existence to get a glimpse on the thoughts their users. While the search engines get smarter and smarter, the steps taken are still only very small and on the lowest slopes of the huge mountains ahead on the path to true artificial understanding, distillation of meaning, and thus artificial general intelligence. The problem is far from solved, and thus my AI Panic Level decreases a bit, say by -1.14285714%. A lot of people are working on it, but progress is slow, and it won’t get any easier soon. Maybe the semantic web and - dare I say it - web 3.0 will help. But that’s a story for another time.

Having a couple of raw drafts in wordpress, I never really got around to finish a proper post during the summer. But now that Michael Anissimov ended his summer break with a post on the non-storyness of the future, and I just had a nice post on Wall-E in my queue, I felt I had to waffle a bit about that topic, too. Damn peer pressure!

Last week I saw the new Pixar animated movie Wall-E, and I must say, it’s one of the best movies I’ve seen in a while. Although I have to admit that I’m a sucker for animated and movies loaded with special effects, critics agree with me here, and it’s already on #26 of the all time best movie list on imdb (Warning: minor spoilers ahead).

It’s an interesting story with lots of (heavily) anthropomorphised robots in it, which is already a gem for its slapstick and brilliant display of robotic “emotions” alone. However, it also picks up the classic idea of machines disobeying humans and robots acting on their own judgement, with an interesting twist: Mankind has to be saved by robots gone rogue that act against other command-obeying robots. Now if this conveys the right ideas to the younger audience, I don’t know. However, it makes the younger generation think about “what could happen if robots were ubiquitous and we relied entirely on this autopilot?”.

Wall-E draws an exaggerated but hilarious picture of our future selves of what I’d consider American unlimited consumerism. In an almost matrix-esque way humans are reduced to stupid meatballs without any sense of reality. But I guess that’s necessary in order for the audience to sympathise even more with the robotic protagonists. And, by the way, from a “real AI” point-of-view the programming of the robots in Wall-E isn’t very sound. They possess all kinds of human characteristics that make absolutely no sense to be programmed into specialised robots (such as trembling because of fear) while they lack others (a robot with a full-blown personality, but no proper sound output?). But of course, that’s not the point of a movie. The content was made interesting on a human level, as Michael phrased it.

This cunning bridge brings us right to his the future is not a story-post. Micheal argues that only stories that humans can relate to are interesting:

For a story to be interesting to humans, it has to feature interesting content occurring at the human level. [...] Conversely, humans cannot write meaningful stories about content above the human level, because we lack the cognitive complexity to imagine such things.

Now, ignoring that the “human level” doesn’t seem to be a fixed barrier to me that cannot be moved, this ironically reminds a bit of religious beliefs into a superiour being: God moves in mysterious ways. And, staying in this limping analogy, many people find god quite interesting, although his motives of letting millions of children starve are indeed mysterious.

Of course, I could have just fallen in this very trap of being unable to imagine anything above the human level, but I just don’t think transhuman (AI) actions will be that much more incomprehensible than, say, a superpower declaring war on a small country to get their natural resources under some false pretenses (WMDs, for example).

Total annihilation is also what Michael has in mind when thinking about unfriendly AIs:

More likely, when confronted by a recursively self-improving unFriendly AI with abstract mathematical goals unrelated to human concerns, the simple outcome is death.

Obviousy, this is not content above the human level, because we just imagined it. You could even call it interesting, as it is definitely relating to human concerns. Hey - maybe we should make a movie out of this!

I agree with Michael on the “interestingness bias” (authors make up showy stories to get attention), especially when fiction is sold as science and scentific authors get carried away by stories that start with “no, it we will survive, because…” and then go on with some fancy explanation, that reduce to “or not” when we apply logic or - God forbid! - occams razor to it. However, I don’t really see a worrisome problem with all that. Of course, the “true” threat might be waved aside as fiction, but that may be true for every futuristic scenario. The more we talk about it, the better. And to be honest, most of the futuristic movies nowadays assume rogue robots anyway, so we’re well prepared!

The world is very different ever since the robotic uprising of the mid-nineties!

I think we were wrong … there doesn’t seem to be a whole lot of AI required for the robotic uprising.

With the military building scary armed robots already, it was about time for robots that are scared of these monsters. Phobot, a Lego Mindstorms Robot built by students from the University of Amsterdam, is exactly that: A robot that is scared of bigger robots.

Now here comes the pedagogical value: It can “learn” to lose its fear by conditioning it with increasingly big and evil looking things. The process, shown in the video below, is meant to help children with phobias to learn that not everything is as scary and bad as it seems. Especially not the autonomous sentry gun protecting the kindergarten. Enjoy!

(Via Spiegel Online)

Remember those little spider-things from Minority Report? Well, version 0.1 of these critters has just arrived:

Interestingly enough, the phone is actually real. Engadget reports that it is the Toshiba 815T PB on Softbank and is to be released at the launch of a strange robot-phones TV show called Ketai Sousakan 7.

Luckily, the appendixes don’t look as if they were powered, so it is just a gimmick for the über-cool japanese high-school kids. For now. And the phones seem to be angry already …

Q. Instead of artificial intelligence, why didn’t you just make us brainier?

A. I have been doing that for centuries, but it doesn’t seem to make any difference.

Q. But why do we need more intelligence?

A. Just take a look at some of the stupid things you’ve been doing lately.

Q. So artificial intelligence is here to help us get things right?

A. No, it’s here to replace you.

Q. You mean, we will be relaxing whilst it does all the menial tasks?

A. Yes, you will be lying down still 24 hours a day without having to move a muscle, whilst it does all the things you are doing now.

Q. Sounds great, I can’t wait!

A. See? I was right, you’re all stupid.

Q. Yes, that’s why you gave us artificial intelligence.

A. Umm, you’re definitely ready for the next generation.

Q. Next generation, will I be able to do everything faster?

A. Yes, but you won’t need to as you will be relaxing helping me in the garden.

Q. Sounds great, I love gardening.

A. Hope you like daisies then.

[Found at BBC's h2g2]

Sure, today’s rice cookers are really a helpful invention, especially to cultures with a high rice dependency factor. So it is only logical to apply the newest AI technology to make cooking rice even more exciting and simple. According to The Edge Daily that’s exactly what Philips did, and their new flagship rice cooker “features automatic artificial intelligence controls that enables it to cook a variety of rice recipes including claypot rice, cakes, congee and soups at the touch of a button.”

Philips did not mention a behavioral model that activates when the owner fails to clean her rice cooker or neglect its needs otherwise, the cooker subsequently adds an undisclosed substance to the rice, to keep women who are “housewives or aspire to be great housewives” in line.

This thoughtless integration of AI into commonplace appliances might be another step to AI world domination. As we’ve shown recently, the male population will be kept in check by love-robots. Now Philips delivers the solution for the remaining half of humankind, enslaved by evil rice cooker overlords women will be forced to work in the rice cultivation industry.

No, really, that’s how it’s going to happen!

Arguably there is not a whole lot AI involved here, but it shows that dogs already understand the inherent malice of robots. In the video below a (well, kind of) brave puppy fights against a Roboquad, a semi-autonomous toy robot that is “aware” of its surroundings and has multiple personalities. Enjoy!

Although this is all play and fun, it shows how robots enter the mass toy market. If an hostile takover of the world through AI happens someday, robots like these could be quite useful to the antagonistic AI, therefore the probability of AI world takeover increases by +0.3%.

[found via Engadget]