The Null Device
Posts matching tags 'ai'
A General Technique for Automating NES Games; a programmer in the US has created a system for automatically learning how to play some NES games, by monitoring control inputs, finding increasing sequences of bytes in the NES's 2Kb of RAM (which look like scores or level indicators, i.e., things to be increased) and taking it from there. It works better on some games than others (he has it playing Super Mario Brothers moderately well, and exploiting quirks no human player would stumble across, though it's hopeless at Tetris). There is a paper here.
Today in algorithmic content creation: Philip Parker, a professor of marketing in the US, has created an algorithm that automatically generates books on a variety of subjects, gathering information on the internet in the way a human author would. The article suggests that the result is of a somewhat higher quality than the usual spam ebooks harvested from Wikipedia articles:
To be clear, this isn’t just software alone but a computer system designated to write for a specific genre. The system’s database is filled with genre-relevant content and specific templates coded to reflect domain knowledge, that is, to be written according to an expert in that particular field/genre. To avoid copyright infringement, the system is designed to avoid plagiarism, but the patent aims to create original but not necessarily creative works. In other words, if any kind of content can be broken down into a formula, then the system could package related, but different content in that same formula repeatedly ad infinitum.The hundreds of thousands of books generated by this system range from the fairly generalist and relatively cheap (Webster’s English to Haitian Creole Crossword Puzzles: Level 1, which can be yours for $14.95; incidentally, “Webster's” is not a trademark) to the more specialised and pricy (The 2007-2012 World Outlook for Wood Toilet Seats for $795). As the system works on demand, it is even possible to fill the catalogue with books that could exist, and generate the books when someone buys one; it's Borges' Infinite Library as a money-making scheme.
In truth, many nonfiction books — like news articles — often fall into formulas that cover the who, what, where, when, and why of a topic, perhaps the history or projected future, and some insight. Regardless of how topical information is presented or what comes with it, the core data must be present, even for incredibly obscure topics. And Parker is not alone in automating content either. The Chicago-based Narrative Science has been producing sport news and financial articles for Forbes for a while.And following on his success with auto-distilled technical and factual tracts, Parker is next applying his system to the potentially even more lucrative field of romance novels (which have the advantage of both being defined by a formula, not requiring a huge amount of originality, and being the largest share of the consumer book market).
And if romance novels fall next, followed in short order by other functionally formulaic genres (techno-thrillers, for example, or police procedurals), we may soon find ourselves entertained by machines of loving grace. Though there's no reason why it should stop at books; given that the scripts of mass-market films (with the amounts of money invested in their production and the bottom-line-oriented conservatism of the corporations holding the purse strings) are already produced by a highly formulaised process (scriptwriters use special software to define the skeletons of their plots, making sure it fits in the formal constraints of the genre), going further and writing software that will make the plot to the next action blockbuster or quirky indie comedy would be relatively easy. Of course, today, it makes little sense to replace the scriptwriters with a piece of software whilst keeping all the actors, cameramen, lighters, gaffers and best boys on the payroll, though this may change as computer graphics technologies improve:
Using 3D animation and avatars, a variety of audio and video formats can be generated, and Parker indicates that these are being explored. Avatars that read compiled news stories might become preferred, especially if viewers were allowed to customize who reads the news to them and how in-depth those stories need to be.Then, eventually, the software will be miniaturised and commodified, becoming more widely available. Rather than belonging to content barons who fill the stores with algorithmically generated pulp fiction and technical literature, it'll live in your phone, tablet or e-reader, and will tell you stories, sing you songs and show you movies tailored to entertain you, based on your previous selections.
(via David Gerard)
Yet another titan of computing has died; this time, John McCarthy, the artificial intelligence pioneer and inventor of the Lisp programming language. He was 84.
And more on the subject of Siri; while the technology is available only on Apple's iOS platform (and currently only on the latest and greatest iPhone), an Android software company have taken it upon themselves to make their own version, in an 8-hour hackathon. It's named Iris (see what they did there?), and it sort of works:
Me: Remind me at 9pm to go and buy milk
It Recognised: remindme at 9 pm to go in hawaii
It Replied: I have two pets.
Me: Where is siberia
Replied: Wherever you make it I guess
Q: Where can I get a recipe for cheesecake?If one views this as a competitor to Siri, it falls well short (even without the bizarre voice-recognition results, it doesn't seem to contain the sort of evolving model of the user, their relationships and preferences, and the current context that makes a system like Siri work), though one could hardly expect this from an 8-hour hacking session. (If one views it as a publicity stunt to promote Dexetra's other apps, it'll probably be far more successful.) However, as a surrealist tool for injecting chaos into the lives of those who use it, it looks to be far superior, escaping the shackles of bourgeois practicality that constrain Apple's more polished product. Iris looks to be a virtual assistant André Breton could love.
A: En la esquina, con minifalda.
("In the corner, wearing a miniskirt.")
Apple's latest iPhone, the 4S, comes with a feature named Siri, an intelligent agent (based on technology from a US military AI research programme) which answers spoken questions in natural English, using web services, the current environment and a constantly evolving profile of the user and their preferences to make sense of ambiguous queries like "will I need an umbrella tomorrow?", and speaks the results back to the user—in a female voice in the US and Australia, but a male one in the UK. Apple haven't explained the reasons for the difference, but there are theories:
Jeremy Wagstaff, who runs technology consultancy Loose Wire Organisation, says: "Americans speak loudly and clearly and are usually in a hurry, so it makes sense for them to have a female voice because it has the pitch and range. British people mumble and obey authority, so they need someone authoritative." Which, apparently, still means male.There's more historical context here (which talks about disembodied machine voices having been female for a long time, since telephone operators* and WW2-era navigation systems, female voices being used in railway station announcement systems because their higher frequencies carry better against the train noise, evil computers in films being presented as male, and BMW having to recall a female-voiced navigation system in the 1990s because of complaints from German men who refused to take direction from a woman).
There's also a piece in the Atlantic about why many electronic devices designed to assist have female voices. It looks predominantly at systems in the US, and concludes that, in America at least, female voices are perceived to go better with the role of assistant—competent, level-headed, and unthreateningly loyal. Or, in other words, everybody wants to be Don Draper.
Which doesn't answer the question of why (according to Apple's in-house cultural anthropologists, anyway) British users feel more comfortable with male-voiced virtual assistants. Could it be the lack of the famous 100-watt smiles of the American service industry (as per the US psychologist who categorised British smiles as grimaces of acquiescence)? An ingrained sense of social hierarchy and/or traditional acceptance of class privilege which makes authoritative male voices more acceptable in Britain? (I wonder whether refined-sounding male British voices would be popular with American users; after all, I imagine that quite a few people wouldn't mind their virtual assistant to have a British butler persona.) Or perhaps the residual trauma of Thatcherism makes female voices with any hint of authority a hard sell in Britain? And why does Australia get the female voice option by default? Is Australia more "American" than "British" in this sense? Or is the preference for male voices some peculiarly British anomaly among the English-speaking nations?
* If I recall correctly, the very first telephone operators in the late 19th century were boys, of the same background who would have been employed in clerical tasks. They tended to horse around and play pranks too much, though, so they were replaced with female operators after a few years. Throughout living memory, the typical telephone operator (where those still existed) has been a woman.
The London Review of Books looks at various books recently published about Google, an essay on Google's data-collecting and machine-learning operations; it appears that a lot of the services Google provide are
In 2007, Google told the New York Times that it was now using more than 200 signals in its ranking algorithm, and the number must now be higher. What every one of those signals is and how they are weighted is Google’s most precious trade secret, but the most useful signal of all is the least predictable: the behaviour of the person who types their query into the search box. A click on the third result counts as a vote that it ought to come higher. A ‘long click’ – when you select one of the results and don’t come back – is a stronger vote. To test a new version of its algorithm, Google releases it to a small subset of its users and measures its effectiveness through the pattern of their clicks: more happy surfers and it’s just got cleverer. We teach it while we think it’s teaching us. Levy tells the story of a new recruit with a long managerial background who asked Google’s senior vice-president of engineering, Alan Eustace, what systems Google had in place to improve its products. ‘He expected to hear about quality assurance teams and focus groups’ – the sort of set-up he was used to. ‘Instead Eustace explained that Google’s brain was like a baby’s, an omnivorous sponge that was always getting smarter from the information it soaked up.’ Like a baby, Google uses what it hears to learn about the workings of human language. The large number of people who search for ‘pictures of dogs’ and also ‘pictures of puppies’ tells Google that ‘puppy’ and ‘dog’ mean similar things, yet it also knows that people searching for ‘hot dogs’ get cross if they’re given instructions for ‘boiling puppies’. If Google misunderstands you, and delivers the wrong results, the fact that you’ll go back and rephrase your query, explaining what you mean, will help it get it right next time. Every search for information is itself a piece of information Google can learn from.
By 2007, Google knew enough about the structure of queries to be able to release a US-only directory inquiry service called GOOG-411. You dialled 1-800-4664-411 and spoke your question to the robot operator, which parsed it and spoke you back the top eight results, while offering to connect your call. It was free, nifty and widely used, especially because – unprecedentedly for a company that had never spent much on marketing – Google chose to promote it on billboards across California and New York State. People thought it was weird that Google was paying to advertise a product it couldn’t possibly make money from, but by then Google had become known for doing weird and pleasing things. ... What was it getting with GOOG-411? It soon became clear that what it was getting were demands for pizza spoken in every accent in the continental United States, along with questions about plumbers in Detroit and countless variations on the pronunciations of ‘Schenectady’, ‘Okefenokee’ and ‘Boca Raton’. GOOG-411, a Google researcher later wrote, was a phoneme-gathering operation, a way of improving voice recognition technology through massive data collection. Three years later, the service was dropped, but by then Google had launched its Android operating system and had released into the wild an improved search-by-voice service that didn’t require a phone call.One takeaway from the article is that, while it may be said that "if you don't know what the product is, you are the product", Google don't really give that much personal information to advertisers, or even allow advertisers to target ads very precisely (as they can, for example, on Facebook). Google collect a wealth of information, though the bulk of it remains in the machine:
It isn’t possible, using Google’s tools, to target an ad to 32-year-old single heterosexual men living in London who work at Goldman Sachs and like skiing, especially at Courchevel. You can do exactly that using Facebook, but the options Google gives advertisers are, by comparison, limited: the closest it gets is to allow them to target display ads to people who may be interested in the category of ‘skiing and snowboarding’ – and advertisers were always able to do that anyway by buying space in Ski & Snowboard magazine. The rest of the time, Google decides the placement of ads itself, using its proprietary algorithms to display them wherever it knows they will get the most clicks. The advertisers are left out of the loop.
Researchers at CMU write a program that learns facts by reading the web; hilarity ensues:
NELL also judges the facts it finds, promoting some of them to the higher category of “beliefs” if they come from a single trusted source, or if they come from multiple sources that are less reliable. According to the researchers, “More than half of the beliefs were promoted based on evidence from multiple [i.e., less reliable] sources,” making NELL more of a rumor mill than a trusted source. And once NELL promotes a fact to a belief, it stays a belief: “In our current implementation, once a candidate fact is promoted as a belief, it is never demoted,” a process that sounds more like religion than science.
Sometimes NELL makes mistakes: the computer incorrectly labeled “right posterior” as a body part. NELL proved smart enough to call ketchup a condiment, not a vegetable, a mislabeling that we owe to the “great communicator,” Ronald Regan. But its human handlers had to tell NELL that Klingon is not an ethnic group, despite the fact that many earthlings think it is. Alex Trebek would be happy to know that, unlike Sean Connery, NELL has no trouble classifying therapists as a “profession,” but the computer trips up on the rapists, which it thinks could possibly be “awardtrophytournament” (confidence level, 50%).
Told by its programmers that Risk is a board game, NELL predicts with 91.4% confidence that security risk is also a board game. NELL knows that a number is a character, but then incorrectly classifies the plural, numbers, as a character trait (93.8% confidence). The computer is also 99.9% confident that business is an academic field, which may be reassuring to those in the b-school if not to those small business owners worrying about the continuation of the Bush tax cuts.
Three Israeli computer scientists have developed an algorithm for detecting sarcasm in online comments. Named SASI (Semi-Supervised Algorithm for Sarcasm Identification), the algorithm looks at linguistic features of the sentences to guess whether or not they're sarcastic. It was developed with Amazon product reviews as training data. Potential applications of such algorithms could include systems that gauge the positivity or negativity of public opinion about a subject from online discussions.
A US company is developing a system that models and replicates the styles of famous musicians. Details of how Zenph Sound Innovations' system works are scant (apparently "complex software" is used, which simulates the musicians' styles, and the resulting high-resolution MIDI files are played on robotic musical instruments; currently pianos, though a double bass and saxophone are in the works).
Currently, it is capable of reconstructing a performer's style of playing a specific work, from a recording of the work, and can be used to rebuild flawed recordings. It cannot yet play a new piece in a performer's style, though the developers are planning to work on that next.
“It introduces a whole bunch of interesting intellectual-property issues, but eventually, you ought to be able to, in essence, cast your own band,” said Frey. “You should be able to write a piece of music and for the drum piece, have Keith Moon, and for the guitar piece, you can have Eric Clapton — that is a derivation of understanding each of those artists’ styles as a digital signature. That’s further down the road, but initially, you’re going to have the ability for artist to create music and have the listener manipulate how they want to hear it — [for example] sadder.”The intellectual-property implications alluded to are interesting; the prospect is raised of a new type of copyright, over an artist's style, being created, with the artist or their estate collecting royalties from replication of their style. While this is perfectly consistent with the copyright-maximalist ideology of the corporate-dominated, post-industrial present day, it ignores the fact that artists emulate other artists all the time. While initially, courts would exercise "common sense" and leave non-software-based copyists alone (i.e., Oasis wouldn't owe licensing fees to the Beatles), sooner or later, once the technology becomes the norm, this original intent would be forgotten and, after a few strategic court cases, a new precedent would be set, declaring styles, and the elements of them, to be licensable, much in the way that patents are, and requiring anyone taking them off to license them, much as anyone sampling even a split-second of a recording has to license it. (In the age of powerful rights-licensing corporations with political clout, intellectual-property law is a ratchet that turns only one way.) Soon, the different elements of musical style would end up aggregated in the hands of a few gigantic rightsholders with well-resourced legal teams, and musicians would be routinely slugged with heavy bills, itemised by stylistic elements.
Jazari is essentially an automated, electromechanical percussion ensemble, controlled using two Nintendo Wii controllers. It consists of a MacBook, a bunch of Arduino boards and a room full of drums fitted with solenoids and motors, and software written in MAX and Java which parses input from the Wii controls and plays the drums. The software is also capable of improvising with the human operator, by imitating, riffing off and mutating what he plays.
Jazari was developed by a guy named Patrick Flanagan, who had been playing around with algorithmic composition, only to discover that people don't want to hear about algorithms, but do want to see a good live show. Anyway, here there are two videos: one of a Jazari performance (think robot samba float, conducted by a guy waving Wiimotes around; the music has a distinctly Afro-Brazilian feel to it), and one of Flanagan explaining how it works.
Artificial Flight and Other Myths (a reasoned examination of A.F. by top birds), a satire of arguments against the possibility of strong artificial intelligence:
Current A.F. is limited to unpowered gliding; a technical marvel, but nowhere near the sophistication of a bird. Gliding simplifies our lives, and no bird (including myself) would discourage advancing this field, but it is a far cry from synthesizing the millions of cells within the wing alone to achieve Strong A.F. Strong A.F., as it is defined by researchers, is any artificial flier that is capable of passing the Tern Test (developed by A.F. pioneer Alan Tern), which involves convincing an average bird that the artificial flier is in fact a flying bird.
There are religious birds who believe God made Bird in His own image, and while I do not share in most of these beliefs, I do think there’s something to be said about the motivation behind creating Strong A.F. Perhaps, as we are the only creatures on Earth capable of flight, we want to push forward past our current capabilities, perhaps even augmenting our own flying capacities if independent A.F. is an impossibility. This could be interpreted as noble, but I would argue that there’s very little utility in replicating what nature has essentially perfected. Why spend millions on an artificial flier when there are so many birds out of work?
(via Boing Boing)
A Google engineer writes about how Google's search engine attempts to understand synonyms:
We use many techniques to extract synonyms, that we've blogged about before. Our systems analyze petabytes of web documents and historical search data to build an intricate understanding of what words can mean in different contexts. In the above example "photos" was an obvious synonym for "pictures," but it's not always a good synonym. For example, it's important for us to recognize that in a search like [history of motion pictures], "motion pictures" means something special (movies), and "motion photos" doesn't make any sense. Another example is the term "GM." Most people know the most prominent meaning: "General Motors." For the search [gm cars], you can see that Google bolds the phrase "General Motors" in the search results. This is an indication that for that search we thought "General Motors" meant the same thing as "GM." Are there any other meanings? Many people can think of the second meaning, "genetically modified," which is bolded when GM is used in queries about crops and food, like in the search results for [gm wheat]. It turns out that there are more than 20 other possible meanings of the term "GM" that our synonyms system knows something about. GM can mean George Mason in [gm university], gamemaster in [gm screen star wars], Gangadhar Meher in [gm college], general manager in [nba gm] and even gunners mate in [navy gm].
The latest computer science research being funded by the US military includes a bot that will impersonate parents in the service and talk to their children when they're off fighting wars:
The challenge is to design an application that would would allow a child to receive comfort from being able to have simple, virtual conversations with a parent who is not aivailable "in-person". We are looking for innovative applications that explore and harness the power of “advanced” interactive multimedia computer technologies to produce compelling interactive dialogue between a Service member and their families via a pc- or web-based application using video footage or high-resolution 3-D rendering. The child should be able to have a simulated conversation with a parent about generic, everyday topics. For instance, a child may get a response from saying "I love you", or "I miss you", or "Good night mommy/daddy." This is a technologically challenging application because it relies on the ability to have convincing voice-recognition, artificial intelligence, and the ability to easily and inexpensively develop a customized application tailored to a specific parent.
(via Boing Boing)
A graduate arts student named Drew Burrows has created a holographic virtual sleeping partner. Titled "Inbed", the installation consists of a bed with an infrared camera and projector positioned above it, and a computer which recognises the sleeper's position and projects one of several images of a sleeping woman onto the bed, so as to interact with the sleeper. Burrows says that the piece aims to "speak on the feelings of loneliness, affection, and intimacy", a point lost on the New York Magazine article which beat this up as "weirdo student builds a virtual girlfriend because he's `too busy' to find a real one".
(via Boing Boing)
Science News has an article about recent advances in computer music processing. There has been success in creating software which understands recorded music, to the point of being able to extract note information from a (polyphonic, multitimbral, acoustically imperfect) recording. This has been achieved not by programming in rules of musical theory but by using machine learning techniques, setting up a learning system and training it from examples to infer its own rules of music:
He started with a program that had no information about how music works. He then fed into his computer 92 recordings of piano music and their scores. Each recording and score had been broken into 100-millisecond bits so that the computer program could associate the sounds with the written notes. Within those selections, the computer would receive an A note, for example, in the varying contexts in which it occurred in the music. The software could then search out the statistical similarities among all the provided examples of A.
In the process, the system indirectly figured out rules of music. For example, it found that an A is often played simultaneously with an E but seldom with an A-sharp, even though the researchers themselves never programmed in that information. Ellis says that his program can take advantage of that subtle pattern and many others, including some that people may not be aware of.The software thus developed got impressively good results in music transcription tests (68% accuracy, with the runner-up, a traditional rule-based system, getting 47%). There are numerous applications of such a technology, from automated accompanyists to "musical spellcheckers" to ways of "decompiling" recordings to a score:
Score-alignment programs could be used after a musician records a piece of music to do the kind of fine-tuning that's now performed painstakingly by recording studios, fixing such problems as notes that are slightly off pitch or come in late. "It'll be kind of like a spell-check for music," says Roger Dannenberg, a computer scientist at Carnegie Mellon University in Pittsburgh who is developing the technology.
Christopher Raphael begins the third movement of a Mozart oboe quartet. As his oboe sounds its second note, his three fellow musicians come in right on cue. Later, he slows down and embellishes with a trill, and the other players stay right with him. His accompanists don't complain or tire when he practices a passage over and over. And when he's done, he switches them off.Not everybody's happy with this, though; musicians' unions, which have opposed "virtual orchestras", are about as keen on it as buggy-whip manufacturers were on the automobile.
(via Boing Boing)
Artificial intelligence/cognitive science pioneer Marvin Minsky, who has recently written a book on the mechanisms behind emotions, gives an interview, weighing in on intelligence, cognition, the nature of self and the ineffable mysteries of life:
Q What, in your view, is love?
A There's short-term infatuation, where someone gets strongly attracted to someone else, and that's probably very often a turning-off of certain things rather than something extra: It's a mental state where you remove your criticism. So to say someone is beautiful is not necessarily positive, it may be something happening so you can't see anything wrong with this person. And then there are long-term attachments, where you adopt the goals of the other person and somehow make serious changes in what you're going to do.
Q And what is the self?
A We often imagine that there's a little person inside ourselves who makes our important decisions for us. However, a more useful idea is that you build many different models of yourself for dealing with different situations -- and each of those self-images can add to your resourcefulness.
UnSuggester is a book recommendation engine in reverse; enter a book you liked, and it'll give you a list of books you probably won't like. Apparently, fans of William Gibson's Neuromancer and Michael Moore's Stupid White Men would least want to read books on theology, the opposites of Marx & Engels' Communist Manifesto look like erotica novels, Bulgakov's The Master and Margarita is the least like an array of romance novels, Star Wars novelisations and theological texts, and the opposites of Design Patterns are mostly chick-lit, whereas The Little Prince finds itself to be the antithesis of thrillers and scifi novels. Meanwhile, people who read Illuminatus! are unlikely to read Freakonomics, and the opposite of The Da Vinci Code, with its simplistic structure and grand revelations, appears to be, naturally enough, French postmodernist philosophy.
There is a rather good hard-scifi story at Salon: "The Perfect Man" by Lauren McLaughlin. It's about a woman who has a virtual AI boyfriend made to order, who then transforms from adorably bumbling Hugh Grant-esque Hollywood Englishman stereotype to sinister, inscrutably calculating Hollywood Englishman stereotype:
The design process is easy. First step: Pick a physical template. A youth squandered on Monty Python reruns left me with a full-blown kink for English guys, so I chose a template called "Nigel" -- think Michael Palin crossed with Laurence Olivier. Then, to assure he didn't look overdesigned, I clicked the "random factor" option to introduce "lifelike imperfections."
If you want to know anything about the "human" rights travesty currently under way courtesy of draconian anti-AI laws, there's a whole subculture of liberationists ready to lecture you on it. They've got the skinny on behavioral inhibitors, recursive self-teaching limiters and other artifacts of AI "slavery." For my purposes, what it all boiled down was this: snip Pritchard's inhibitors or resign myself to dating a functionary. Do you want to date a functionary? Me neither. Thankfully, for every Webcop dutifully guarding the behavioral inhibitors of the thousands of AIs cropping up on the Web, there's a black market geek with the tools to snip.
Now that I have my sanity back, I must dive deep into the black waters of her soul, excavate her most primal desires, and do what no human male has been able to do: keep her interested in me. Thankfully, I have one freedom human males do not -- the freedom to redesign myself. I can make myself so fascinated by Lucy that all I want to do is watch her, study her. A nip here, a tuck there, and voilà, I'm in love with the girl. Well, not in love, exactly. Love is still an alien concept. But I have made myself a bit of a stalker. And the more information I gather about my lovely little monkey, the more I can adjust my personality to suit her needs. Heck, I could turn myself into Prince Charming if I wanted. Something tells me that would not tickle Lucy's fancy. In fact, the more I learn about Lucy, the more I realize she doesn't know what she wants at all. She only thinks she knows. No, Lucy's desires are my nut to crack. And crack it I will. Or she'll crack me. Oh, I don't mean to sound morbid. I'm incapable of morbid thoughts. To mitigate the persistent fear of being snuffed, I've given myself a devil-may-care attitude about death. That way I can focus my energies more intensely on Lucy.
Of course she doesn't know the contents of her subconscious. She lacks the processing power to unravel it. It's a number-crunching job, that's all. Humans, with your lovely little wet brains, will never achieve the self-knowledge you so desire.
(via Boing Boing)
Retrievr is a web-toy that lets you search images on Flickr by drawing things that look like what you're looking for on a Flash applet. It seems to go mostly by colour, rather than feature recognition, and seems to only search a small pre-cached subset of images (which tends towards the prettier end of the scale and includes a lot of sunsets, cats and arty black-and-white photos). As far as finding pictures of objects one draws, it's rather unsuccessful; though, nonetheless, it does yield interesting results.
(via The Fix)
The latest advance in Windows worms is a worm which takes over people's instant-messaging accounts and chats to their friends, attempting to talk them into downloading it; in short, an automated form of social engineering:
According to IMlogic, the worm, dubbed IM.Myspace04.AIM, has arrived in instant messages that state: "lol thats cool" and included a URL to a malicious file "clarissa17.pif." When unsuspecting users have responded, perhaps asking if the attachment contained a virus, the worm has replied: "lol no its not its a virus", IMlogic said.Which suggests that the Turing test may be easier to pass in an environment where people start messages with "lol". If your friends suddenly turn into giggling prepubescents and start trying to convince you to download a file, you know what's happening.
I wonder whether this will lead to an arms race in worm conversational abilities. Perhaps the next one will trawl message logs and pick out phrases/words used by that contact (or use them to change its own writing style)?
Patent of the day: Automatic Detection of Pornographic Images:
If the pixel color is determined to be "skin" 46, the image is sent to a first shape detection process indicated for example as "face detection" of block 48 wherein steps similar to blocks 26 and 28 of FIG. 1 are performed. If the image is detected as a "face" 50, the image is classified as "portrait" and a manual check/inspection is done only infrequently (block 52). If the image is not a "face" 54, the image is analyzed to determine if it is a body part (block 56) i.e., other than a face. If it is not a body part (58), the image is classified as a "landscape", and this type is only inspected occasionally (block 60) i.e. only a small percentage of these images are inspected manually. If the image is a body part (62), a pose detection is done to determine if there is an erotic position (block 64). If it is determined that the pose is not erotic (66), this image is classified as a "swim suit picture" and the result of the detection may be a "parental guidance" notice attached (block 68).
(via bOING bOING)
An interesting item cribbed from Jim: WordsEye is an experimental system that takes sentences in English (such as "The marmalade sky. The zebra skin ground is red. The sun is large. The hippo is next to the horse.") and renders them as images (such as this one). There's a wealth of example images here. It can represent abstract ideas (such as "John believes the cat is not blue", and automatically infer how to symbolise that something happened at a certain time, and map adjectives to textures. Of course, sometimes it gets amusing results, such as its rendering of "John looks at the cat on the skateboard", or, indeed, "the devil is in the details".
Researchers at MIT are developing software which paraphrases English-language text; for example, their software is able to take the sentence "The surprise bombing injured 20 people, 5 of them seriously," and rewrite it as "Twenty people were wounded in the explosion, among them five in serious condition." The system uses techniques adapted from computational biology to match fragments of sentences to others, skirting the entire area of semantics altogether; online translation systems like Babelfish/Systran use similar techniques. (via Techdirt)
(This is quite distinct from automatic summarisation software, which takes a text and delivers the "gist" of it. Some years ago, a large chunk of various intelligence agencies' research budgets was spent on this area, in an attempt to more easily cope with the flood of signal intelligence. And, unless the CIA and such have such systems in use, chances are it still is.)
Anyway, back to paraphrasing; when I read about this, the first thought that came to me was that it would be very useful to student plagiarists seeking to avoid detection (say, by Google searches on key sentences, or matches against other submitted assignments). Which made me wonder: have any plagiarists ever tried covering their tracks by passing an essay through several passes of Babelfish? (Given some of the grammar I've seen in student essays, I'm sure it wouldn't look too amiss.)
Computer scientists in Britain are tackling one of the hard problems in speech recognition: developing software which understands Scottish accents. The Glaswegian accent is one of the hardest on current speech-recognition software (which tends to be rather London-centric, if not American). The team from Birmingham University will be paying locals to say some phrases in the "Glesca patter", which will be analysed to develop regionally-correct voice-recognition software for use in office computers and mobile phones. (via bOING bOING)
A Scottish artificial intelligence lab has developed
a robot with an eye for the ladies. The robot, nicknamed Doki, can identify male and female faces by appearance, and as a side effect, can calculate how physically attractive female faces are. (Mind you, this assumes that attractiveness is exclusively a factor of femininity of appearance, a somewhat simplistic model.) I wonder if (a) it would be fooled by transvestites, and (b) whether it gives a readout in
Helens millihelens (the unit of beauty required to launch one ship).
The next wave in marketing is here: chatroom bots or "buddies" with virtual personalities, which befriend people, make conversation and gently encourage them to consume lifestyle products -- and potentially provide marketing analysts with a lot of customer-profile data in the form of conversations.
Most buddies are programmed with personalities that appeal to their target audiences. ELLEgirlBuddy, the Internet ego of teen magazine ELLEgirl, is a redheaded 16-year-old who likes kickboxing, the color periwinkle and French class. GooglyMinotaur, a buddy for the British progressive rock band Radiohead, affected a British demeanor with words like "mate." The Austin Powers buddy, which promotes the summer film "Goldmember," interjects the movie character's favorite phrases - "yeah, baby" and "grrr" - into conversation.
Perhaps surprisingly, thanks to improvements in natural-language technology and extensive customer databases, the bots give the illusion of being sentient. People know they're machines, but choose to suspend disbelief.
ActiveBuddy's bots save details about each user - names, birth dates, even instances when the person used offensive language. When the buddy recalls these facts, it could appear to the user that it is taking a genuine interest in him or her. "We're programmed to respond to certain signals as though in the presence of a life form," said MIT's Turkle. "These objects are pushing our buttons."
Lobsters, a pretty doovy scifiesque short story by Charles Stross. Go read.
An interesting programming contest, which involves writing a program which automatically summarises news items in haiku form. Given the complexity of the task, any satisfactory solutions are going to have to be quite interesting... (via Slashdot)
Computer-generated creativity: a program which designs product advertisements, often more creatively than human ad executives. (BBC News)