Where he at now?

« previous post | next post »

That's the question on a t-shirt designed by John Allison,  the author of the Bad Machinëry comic series:

Remember that dude? Always poppin' up in the corner? Wonder what he doin' now? Where he at now?

For those who are too young (or too old, or too fortunate in some other way) to have encountered the Microsoft's Office Assistant "Clippit", nicknamed "Clippy", the Wikipedia page may be helpful.

The idea of using natural language processing and artificial intelligence to create an automated assistant is a very old one. Perhaps our well-informed readers will be able to trace this theme to its specific fictional roots in some white-collar spin-off of the golem stories, Frankenstein's monster, R.U.R., etc.; or to determine which AI researcher (in the 1960s?) first suggested such applications — including intelligent tutoring systems — as a realistic goal.

In any case, like Clippy and his call-center relatives, most fielded applications in this space have so far gotten a rather negative response from the public.

But getting back to the question on  John Allison's t-shirt, I happen to know the answer: Clippy had a sex-change operation, got a job at SRI, signed on for some re-training in DARPA's CALO project, and changed his name to Siri.

And this time, the public response has been largely, even overwhelmingly, positive.

There are several possible factors in SIRI's apparent success, in addition to maturity of the various speech, NLP, and AI technologies involved: integration into lots of user- and usage-specific information, and into many everyday applications, in a setting that encourages voice I/O; an interface that depends on user initiative and makes it easy to start and end interactions; a well-designed and elaborately implemented "happy secretary" personality, which benefits from the Eliza effect; general Apple fannishness.

As long as the last factor is not the dominant one, we can expect to see many other  "cognitive assistant" applications that are similarly successful. If so, the next question will be, what lies ahead on Clippy/Siri's career path? Must (s)he remain an "assistant" forever — now praised and appreciated, rather than despised and rejected — or will the silicon ceiling be broken?



14 Comments

  1. Chris said,

    October 30, 2011 @ 10:44 am

    A related question: will Siri's popularity encourage spending on the "wrong" kind of AI and drain spending from the "right" kind? You're free to interpret right/wrong here any way you wish, but I'm thinking in terms of MIT's Mind Machine Project which seeks to fix the AI mistakes of the past (the actual web page seems dead, so maybe is the project, not sure). But the critical take-away for me was this: "The field of artificial-intelligence research (AI), founded more than 50 years ago, seems to many researchers to have spent much of that time wandering in the wilderness, swapping hugely ambitious goals for a relatively modest set of actual accomplishments."

    Is Siri the latest swap? Let's trade ambitious AI goals for a phone you can chat with. But in order to make that chat-phone, we'll use relatively uninteresting techniques to fake the user experience. Meanwhile, the next generation of ambitious AI researchers will see their funding dwindle because they can't show direct-to-market applications for their work.

    I am reminded of a quote that floated around the internet recently: "The best minds of my generation are thinking about how to make people click ads."

  2. Nick Lamb said,

    October 30, 2011 @ 10:51 am

    I believe the critical difference between Clippy and Siri is that Clippy offers advice unbidden. Finding out that other people can't help us is disappointing, but normal. Getting advice which seems well meant, but useless, is also not an unexpected result of asking. But when someone interrupts you (as Clippy did I believe) to offer unsolicited advice which is also useless or worse than useless then that's very frustrating. I haven't used Siri, but every description of it I've seen begins with the user explicitly opting to use Siri, not being interrupted mid-thought.

    You could probably model the interruption as a debt, the interrupting agent owes the user advice which is at least good enough to justify this interruption. Any time it interrupts with advice that's less good than that it's actually worse than an agent which offers no help even when explicitly asked.

  3. Dan Lufkin said,

    October 30, 2011 @ 11:04 am

    We'll know Siri's fate when and if a meme arises that claims that Siri is a hoax — actually a roomful of young ladies in India or the Philippines pretending to be a server farm in Tampa. Turing's criterion will apply.

  4. Rosie Redfield said,

    October 30, 2011 @ 11:53 am

    My interactions with Siri have already crossed the critical threshold – her voice recognition is more reliable than my typing.

  5. Janice Byer said,

    October 30, 2011 @ 12:46 pm

    "A related question: will Siri's popularity encourage spending on the "wrong" kind of AI and drain spending from the "right" kind?"

    Isn't there some kind of natural law that, barring human intervention, spending always drains counter-productively in both hemispheres from the "right" kind toward the "wrong" kind?

  6. Mr Fnortner said,

    October 30, 2011 @ 12:51 pm

    My experience with Clippy was much like the mail room staff's experience with the guy who knew everything: "You are always wrong." http://www.youtube.com/watch?v=fCrqy5S-H_A

  7. Mark Mandel said,

    October 30, 2011 @ 1:41 pm

    "its specific fictional roots in some white-color spin-off of the golem stories, Frankenstein's monster, R.U.R., etc.;"

    At first I thought this a typo or unintentional eggcorn for "white-collar", referring to an academic development of a fictional [figuratively blue-collar] idea. But then I thought it might refer to the reversal from the dark side (evil) of these fantasy roots to the intended light side of the tools … but oh, how much Clippy has in common with the Monster! Or at least his younger kin: "Share and enjoy!"

    And having looked at both sides now, I'm not sure. So nu?

    [(myl) Typo for "white collar". Fixed now.]

  8. Ken Brown said,

    October 31, 2011 @ 8:10 am

    @Rosie Redfield – If voice recognition is going to replace or even heavily supplement keyboard input in office work we will have to reverse the century-old trend towards open-plan offices and other sorts of shared workspace.

  9. Ginger Yellow said,

    October 31, 2011 @ 8:50 am

    Perhaps our well-informed readers will be able to trace this theme to its specific fictional roots in some white-collar spin-off of the golem stories, Frankenstein's monster, R.U.R., etc

    I'm sorry, Mark. I can't do that.

    As for why Siri is popular and Clippy was hated, surely the main reason is that unlike Clippy, Siri doesn't interrupt you uninvited and treat you like a two year old. You control the timing and subject of the interaction.

    [(myl) Yes, that's what "an interface that depends on user initiative and makes it easy to start and end interactions" means.]

  10. Jerry Friedman said,

    October 31, 2011 @ 4:19 pm

    The idea of using natural language processing and artificial intelligence to create an automated assistant is a very old one. Perhaps our well-informed readers will be able to trace this theme to its specific fictional roots in some white-collar spin-off of the golem stories, Frankenstein's monster, R.U.R., etc.;

    Not well-informed, but there is a robot secretary named Sulla in R. U. R. When the "central director" of the company orders her to chat, she tells the ingenue to take a stronger ship because the barometer is falling to 27.7, and when he asks the speed and tonnage of the recommended ship, she recites it.

    I didn't look for anything older.

    While searching for this, I found that "robot secretary" was an early name (c. 1950) for an answering machine (which younger readers may want to know was an early form of voice mail :-).

    [(myl) Thanks! Of course, I should have checked R.U.R., which I dimly recall once have read, but clearly not well, because I don't have any memory of Sulla at all.]

  11. Mary Kuhner said,

    November 1, 2011 @ 7:55 pm

    I concur that not interrupting is crucial. My first and last meeting with Clippy came about when I was trying to write a script for a religious ritual. I was just getting into the creative mindset when Clippy said, "I see you're writing a business letter. Can I help?" and proceeded to try to reformat my script. Shortly thereafter Clippy was lying bleeding on the electronic floor.

    I am not in the market for Siri, though. I work in the aforementioned open-plan space. Being able to do things quietly is key; I would never want to replace that with spoken commands.

  12. Read Weaver said,

    November 3, 2011 @ 12:27 pm

    Another Clippy reference, "Clippy Must Die." The Clippy stuff starts around 1:40.
    http://www.youtube.com/watch?v=NxcmoLKVd60

  13. Ian Wright said,

    November 5, 2011 @ 1:09 pm

    @Chris: The "right" kind of AI research never really paid off. The big, ambitious tasks those researchers set themselves turned out to be either trivially easy (Make machines better at math and binary logic) or utterly beyond their capabilities (Natural language processing).

    Meanwhile, the "wrong" kind of AI research has shown consistent returns in business, daily life, and in R&D. People doing the "wrong" kind of AI research have consistently advanced the field of AI. Motion control, expert systems, facial recognition, metadata processing…

    There's feedback between the groups of researchers, but the overall results suggest that the "right" kind of Strong AI academics have the wrong end of the stick.

  14. ENKI-][ said,

    November 5, 2011 @ 5:01 pm

    […]or to determine which AI researcher (in the 1960s?) first suggested such applications — including intelligent tutoring systems — as a realistic goal.
    That would be Lick in Man Computer Symbiosis

RSS feed for comments on this post