• http://jobs.thinkaloud.in (unregistered)

    So this Thinking Machine came outta a fairy tale ? Is this what J K Rowling wrote first ?

  • Anon. (unregistered)

    Why the interviews aren't run in parallel?

  • (cs)

    WOPR in War Games was a Connection Machine too, wasn't it?

    Someone once told me that he and his colleagues at Thinking Machines actually came up with a use for their computer. They loaded the entire Wall Street Journal news archive into it. Text searches came up instantly. "We always get the right answer for the daily WSJ trivia quiz."

    Unfortunately, Alta Vista, Google and others figured out how to do the same thing on inexpensive standard computers.

  • Captaffy (unregistered) in reply to Anon
    Anon:
    So did they actually tell the candidate before hand that the interview process would (or at least could) last 2 days? The article seemed to suggest that they hadn't. What if you had something else to do the next day?

    Seems like their interview process would actually take at least five days. Five interviews a day, and you have to be interviewed by the entire development staff of 25, plus the CEO.

  • Jim (unregistered)

    It's just not true that it didn't support any language but LISP. As a grad student I programmed one in C* (a parallelized version of C). The project was about light scattering through ice crystals.

  • Jim (unregistered) in reply to Jim

    Replying to myself, I see other's have commented on the C* support. (Why does the blog drop you into page 2 of the comments?)

  • NE (unregistered)

    It sounds a lot like a recent article I read about google actually: http://money.cnn.com/2008/05/09/technology/where_does_google_go.fortune/

    Except google is making money hand over fist, and can turn some of it's zanny ideas into money making technologies.

  • KG (unregistered)

    I think I would have loved to work at a company like that. It sounds like true computer science was occurring there, rather than this strictly business-oriented, dilbertesque, enterprisy wtf that most IT organizations are today.

  • (cs)

    End note: After the bankruptcy of Thinking Machines, some of these guys evidently did end up asking themselves the question, "What's the business use of parallel computing?" I'm a developer in a 4GL called Ab Initio, which basically uses parallelism in a big information / data warehousing setting. The Ab Initio suite is the sole product of Ab Initio Software Corporation, which was founded by Sheryl Handler and various other former employees of TM. I don't know who slapped those guys upside the head with the compentency stick, but AI is an excellent tool, even if it has its own WTF or two.

    Addendum (2008-05-13 13:29): Dang -- "competency". I guess I'm the one that needs to be slapped.

  • Salvador G. Jr. (unregistered)

    The real wtf is that the picture caption says "Connection Machines." Isn't it supposed to be "Thinking Machines?"

  • (cs) in reply to jpers36

    Oh, and the interesting work environment still applies, I hear. Interviews are still very involved in order to determine whether a potential applicant "meshes with the company culture". Their IP protection is extremely secretive, and their PR strategy is counter-intuitive. But again, their software is excellent.

  • Steve (unregistered)

    Following up on my own comment above, a couple of things occurred to me while I was walking across the parking lot to grab a snack just a few minutes ago.

    While, as I said, there were probably a few haphazard things about the interview process, it struck me that most of the interviews I've been on since I've been in the research business have been similar; that is, the scientists and true hackers usually talk about their research projects.

    The whole notion, other than probably a bit of egotism crossed with a moderate lack of social skills, is to draw you out about your interests and spark a discussion of the science involved. If you just sit there expecting a cross-examination, then you're going to be disappointed because scientists just don't work that way -- at least the ones I've been privileged to work beside.

    They probably don't care too much about the programming languages you know or the database systems you've worked on -- they assume that you can pick up whatever you need along the way, more or less the way that they have, and they're more interested in someone with an active and inquiring mind rather than a particular skill set.

    One of the telling moments in the story was the incredulousness of the interviewee regarding the hacked soft drink machine, as if this was an egregious waste of company time. Maybe it was, but you have to remember that the hackers at TMC were true hackers (not what the media have perverted the word into) -- they loved computing and computers and technology and what you could do with it. They probably worked 80 hours a week not because some boss told them that they had to work mandatory overtime but because it was fun. It was probably a weekend hack.

    It's completely different from the standard corporate mindset.

    Which is not to necessarily criticize the corporate mindset per se. We need both the process, product-oriented mind and the more scatterbrained, will-of-the-wisp scientist/hacker mind -- they complement and enhance one another.

    I happen to love being in research, if for no other reason than I get to screw off in the middle of the day writing blog comments and nobody's going to say "boo" to me, but mostly because when I get off the shuttle bus every morning I look around myself and realize that I have a rare privilege to work with some of the smartest people in the world on crazy projects that may or may not produce a result but always produce new ideas to try next and they pay me to top it all off.

  • (cs) in reply to Rune
    Rune:
    Well, I should expect some real world relevance in there, else there would be no research labs whatsoever, cause their duty is to research stuff that can be used in 'the real world'.
    Patently false; you short-sighted cats don't know the joy of non-saecular DARPA projects before 9/11: the whole point was being bereft of the market's impious quarterly demands.

    Focusing on the ten-year horizon gives birth to violently creative, disruptive ideas.

  • Farmie (unregistered) in reply to real_aardvark
    real_aardvark:
    How palpably insane do you have to be in order for the DoD to turn your application down? Would you have to pull a gun on ... no, wait a minute, that would be an instant "Funding approved," wouldn't it?
    No, for instant "Funding approved", you have to pull plans for a gun on them, and threaten them with the prospect of its future existence.
  • Tyler (unregistered)

    So what happened to the big blinky-light server? It seems like, even with its monolingual capabilities, someone could find a good use for it?

  • (cs) in reply to klutometis
    klutometis:

    Focusing on the ten-year horizon gives birth to violently creative, disruptive ideas.

    Like what?

  • A. Cube (unregistered) in reply to shadowman
    How did they ever manage to hire anyone?

    Well, the fewer employees there were at the time of the candidate's interview(s), the fewer people had to approve and the easier it would be to pass. :) Seriously, though, it sounds like hiring new talent was the least of their problems, as it would avail them little of they were incapable of selling it--and it sure sounds like they were.

  • Farmie (unregistered) in reply to Salami
    Salami:
    klutometis:
    Focusing on the ten-year horizon gives birth to violently creative, disruptive ideas.
    Like what?
    Like thinking outside the box is too scary, a better solution is to come up with a new box and think inside of that. Now if you're looking for a concrete example, you've obviously missed the point of research funding.
  • (cs) in reply to Salami
    Salami:
    klutometis:
    Focusing on the ten-year horizon gives birth to violently creative, disruptive ideas.
    Like what?
    Big bang?
  • Steve-o (unregistered) in reply to CoderDevo
    CoderDevo:
    Steve-o:
    Those same hardware assets later resurfaced and are now sold under the new "eMachines" brand name.
    I don't see how that is possible as either accurate history or even as a bad joke. Thinking Machines had nothing to do with commodoty PCs or even Intel x86 architecture. Brilliant!

    I'm sorry, I didn't catch your title... was it official historian or humor cop? Constipated!

  • Steve Dekorte (unregistered)

    They could do floating point ops and they did have a FORTRAN for it - FORTRAN 90, which was quite nice. The problem I found with using the machine was the same problem all supercomputers had - timesharing. Workstations had gotten pretty fast and it was often faster to run a job on a personal workstation than to wait for a batch job to run on a supercomputer - particularly when debugging.

  • Zygo (unregistered)

    It does seem ironic to me that this candidate asked everyone he interviewed with about business applications for technology, but doesn't notice the Coke machine.

    Hello? Point of sale operations? Database systems? Enterprise networking? Business processes? That Coke machine screams "business applications for technology."

  • Zygo (unregistered)

    I'm kind of wondering what their firing process is like.

    If it's anything like their hiring process, all it would take is one wanker who always votes "no" to bring HR to a dead halt.

  • (cs)

    The real WTF is that they make their developers pay for SODA out of pocket.

  • Manic Mailman (unregistered) in reply to Saladin
    Saladin:
    The real WTF is that they make their developers pay for SODA out of pocket.

    Best "real WTF" I've seen in a long time...

  • Anon Fred (unregistered) in reply to Saladin
    Saladin:
    The real WTF is that they make their developers pay for SODA out of pocket.
    My boss has always paid for my SODA licenses.

    http://en.wikipedia.org/wiki/Service-Oriented_Development_of_Applications

  • Sejanus (unregistered) in reply to Steve-o
    Steve-o:
    It is my understanding that, as part of the agreement, the administration sold off the assets to recoup a portion of the losses. Those same hardware assets later resurfaced and are now sold under the new "eMachines" brand name. Brilliant!

    Wait, what? I call bullshit.

  • (cs) in reply to CoderDevo
    CoderDevo:
    Steve-o:
    Those same hardware assets later resurfaced and are now sold under the new "eMachines" brand name.

    I don't see how that is possible as either accurate history or even as a bad joke. Thinking Machines had nothing to do with commodoty PCs or even Intel x86 architecture. Brilliant!

    Surely you mean "Brillant".

  • biziclop (unregistered)

    I love these stories. They make my life feel kinda dull.

  • CoderDevo (unregistered) in reply to Salvador G. Jr.
    Salvador G. Jr.:
    The real wtf is that the picture caption says "Connection Machine's." Isn't it supposed to be "Thinking Machines?"

    The company was Thinking Machines, but their product line was called Connection Machine. CM-1, CM-2 and CM-5.

  • lcrl (unregistered)

    Maybe the original CM didn't have fortran, but certainly the later CM-2 and CM-5 did. I used them. Granted it was a wacky dialect of fortran 90, but it wasn't hard to port.

    Of course a lot of scientific code was written specifically for vector machines like the cray, and sucked badly on a CM, but that's another WTF altogether.

  • CoderDevo (unregistered) in reply to Steve-o
    Steve-o:
    CoderDevo:
    Steve-o:
    Those same hardware assets later resurfaced and are now sold under the new "eMachines" brand name.
    I don't see how that is possible as either accurate history or even as a bad joke. Thinking Machines had nothing to do with commodoty PCs or even Intel x86 architecture. Brilliant!

    I'm sorry, I didn't catch your title... was it official historian or humor cop? Constipated!

    Sorry. Yes, self-designated historian here. If you had written humor I wouldn't have responded. I'm no comedian, but I would have given you a pass if you had said,

    After their bankruptcy, Thinking Machine's assets were liquidated and later resurfaced as a Russian botnet.

  • Steve (unregistered) in reply to Steve
    Steve:
    Beats the hell out of the Windows/*NIX/Mac world in which we're living today.

    No it doesn't. You could run a more powerful computer on the sort of hardware inside the average games console controller these days, so in addition to everyone you could do back on that piece of crap you can do all the funky modern stuff, like realtime 3d games, multi-channel audio, massively multiplayer games with people from around the world etc etc. If you were back then again, and were offered the choice of working on that, or a modern Windows PC you'd have to be some sort of retard to choose the former.

  • (cs)

    That place sounds like one of the SBC-spinoff Community Cults... err.. Churches

  • UnHolyGuy (unregistered)

    The main sin of Thinking Machines was not being an abstract research shop, but was being dishonest about the results of their research. They were pretending that they had an actual product that they were selling, when in fact almost all the sales were government subsidized.

    They cooked their books, pure and simple.

    Also, the paranoia and the insistence on a corporate mono culture are both negatives that in the end discourage creativity and innovation in my mind. You need some dissent to be a healthy organization, you need different kinds of people.

    Just look at the way Ab Initio has failed to capture the ETL market space even though they started with an immense lead over their competition. The one time I tried o buy Ab initio was a combination of a legal engagement on the order of the OJ Simpson trial and applying for a Top Secret security clearance. I think they would have polygraphed me if they could have.

  • fmobus (unregistered) in reply to Andrew
    Andrew:
    ratis:
    (print "Who in their right mind would use LISP?")
    [...]

    The vector "dot product" is another good parallel example. For vectors a & b, a "dot" b = SUM(a[k]*b[k]). Each product term can happen at the same time. a = (1 4) b = (2 5)

    LISP: (+ (* 1 2) (* 4 5)) CPU1: (* 1 2) => 2 CPU2: (* 4 5) => 20 CPU3: (+ 2 20) => 22

    Of course, modern Fortan has the "dot product" vector function built-in. On multi-core hardware, Fortan should run in parallel with no code changes.

    Your example is wrong. While instruction 1 and 2 can be run in parallel, instruction 3 will have to wait for the first two to complete.

    WRT "no code changes"... at least in C, to use hyper-threading capabilities, at least with OpenMP, one has to use #pragma directives to tell the compiler where exactly he can "paralellize" stuff. Wouldn't that be necessary in fortran too?

  • (cs) in reply to jpers36
    jpers36:
    End note: After the bankruptcy of Thinking Machines, some of these guys evidently did end up asking themselves the question, "What's the business use of parallel computing?" I'm a developer in a 4GL called Ab Initio, which basically uses parallelism in a big information / data warehousing setting. The Ab Initio suite is the sole product of Ab Initio Software Corporation, which was founded by Sheryl Handler and various other former employees of TM. I don't know who slapped those guys upside the head with the compentency stick, but AI is an excellent tool, even if it has its own WTF or two.

    Addendum (2008-05-13 13:29): Dang -- "competency". I guess I'm the one that needs to be slapped.

    To quote Rowan Atkinson in his pre-Mr Bean days: "My body ... is my tool."

  • JL (unregistered) in reply to FredSaw
    FredSaw:
    JimM:
    So, what's the actual WTF here?
    The fact that the article doesn't give us any of the recipes from the cookbook.
    Which is a pity because, as a company doing cutting-edge AI research, they probably had some revolutionary new recipes for cake.
  • Barf 4eva (unregistered) in reply to Rune
    Rune:
    Well, I should expect some real world relevance in there, else there would be no research labs whatsoever, cause their duty is to research stuff that can be used in 'the real world'.

    Furthermore, I would decline, I like cool technology, but playing around with cool stuff gets boring after a while when you don't have any appliance for it.

    Picture yourself in a really cool car, a big lambo or something, and you sit there, flashing the lights, honking the horn, perhaps rev up a bit. But then, after 5 minutes or so, you get bored. Without your drivers license it's really no fun at all.

    See, that's when you grow some balls and drive the beast anyways, driver's license be damned.

  • (cs) in reply to FredSaw
    FredSaw:
    Salami:
    klutometis:
    Focusing on the ten-year horizon gives birth to violently creative, disruptive ideas.
    Like what?
    Big bang?
    Whatever happened to the Little Bang, incidentally? Not that I'm desperate: I just haappen to like short girls.
  • (cs) in reply to fmobus
    fmobus:
    Andrew:
    ratis:
    (print "Who in their right mind would use LISP?")
    [...]

    The vector "dot product" is another good parallel example. For vectors a & b, a "dot" b = SUM(a[k]*b[k]). Each product term can happen at the same time. a = (1 4) b = (2 5)

    LISP: (+ (* 1 2) (* 4 5)) CPU1: (* 1 2) => 2 CPU2: (* 4 5) => 20 CPU3: (+ 2 20) => 22

    Of course, modern Fortan has the "dot product" vector function built-in. On multi-core hardware, Fortan should run in parallel with no code changes.

    Your example is wrong. While instruction 1 and 2 can be run in parallel, instruction 3 will have to wait for the first two to complete.

    WRT "no code changes"... at least in C, to use hyper-threading capabilities, at least with OpenMP, one has to use #pragma directives to tell the compiler where exactly he can "paralellize" stuff. Wouldn't that be necessary in fortran too?

    Oh Jeez, more "code examples."

    Sit down, kids, and let me tell you a story. This is a story about computer hardware. (Well, actually, it's about the nincompoops who built a company that built the hardware.)

    Actually, I don't need to tell you this story. Read the OP: it may or may not improve your reading comprehension, depending upon how many grey cells you have to bash together.

    What's with this OpenMP crap? Why not bring in Intel's TBB?

    And how many bits are there in a CM-2 register, anyhow?

  • Sigivald (unregistered)

    Newf: Given that WarGames came out the year TM was founded, I'd be shocked if WOPR was a CM, since none would have shipped. Plus the internet is silent about such a connection, and that is exactly the sort of useless geek trivia the internet best preserves.

    (The WOPR did have a lot of blinkenlights, but that's not unique to the CM, and the arrangement is radically different, if you look at the pictures)

    Regarding the lack of an FPU: http://www.longnow.org/views/essays/articles/ArtFeynman.php

    Evidently Richard Feynman (!) worked with the TM people, and didn't find the lack of dedicated FP or Vector units a limitation, when working with many thousand CPUs, for at least some scientific computation (Quantum Crap Of Some Sort I Won't Pretend To Understand).

    ("According to Feynman's calculations, the Connection Machine, even without any special hardware for floating point arithmetic, would outperform a machine that CalTech was building for doing QCD calculations. From that point on, Richard pushed us more and more toward looking at numerical applications of the machine.")

  • Theresa (unregistered) in reply to dpm

    Well, that absurd interview procedure was used all over the place in the 1990s. I got my first job after two days of interviews with just about everyone at a successful, money-making company with really utterly cool technology. Severely jet-lagged, I actually fell asleep in the last one.

    Oh, and I stayed for close to ten years.

    The full story is at:

    http://comediehumaine.blogspot.com/2006/04/sleep-makes-you-successful.html

  • (cs) in reply to CoderDevo
    CoderDevo:
    And hey, what's wrong with Lisp?
    What'th wrong with Lithp? It maketh you thound tho thilly, thatth what! After all, can you take a language theriouthly when it has a thyntakth bathed on Eth-Ekthpreththionth?
  • (cs) in reply to Salami
    Salami:
    klutometis:

    Focusing on the ten-year horizon gives birth to violently creative, disruptive ideas.

    Like what?

    Kama Sutra during compile time. This will bear the fruit that will harvest the next generation of widgets.

    The names Salami and Klutometis automatically sent my mind to this for some reason.

  • Djn (unregistered) in reply to Steve
    Steve:
    (...)

    People were trying new things, often totally off the wall, wacky things, with respect to architecture and (gasp!!) operating systems (anyone remember Very Long Instruction Word architectures?).

    Oh, it's still kicking around - Itanium is more or less VLIW.

  • ps (unregistered) in reply to CoderDevo
    CoderDevo:
    (Neat mice had two wheels under them instead of a ball or laser.)

    I'm currently using one of them. Made by Honeywell (keytronic, really) around 15 years ago, still going strong.

  • Steve (unregistered) in reply to Steve
    Steve:
    Steve:
    Beats the hell out of the Windows/*NIX/Mac world in which we're living today.

    No it doesn't. You could run a more powerful computer on the sort of hardware inside the average games console controller these days, . . .

    You miss my point.

    Sure, the hardware today is much better, faster, cheaper, etc., than it was in the 1980s. That's obvious.

    Yes, an iPod has several orders of magnitude more memory and faster clock cycle than the Cray Y-MP that I programmed back then.

    The point is that there was more diversity in operating systems and a lot of them were much more interesting than Windows, *NIX, or Mac OS X (which is really essentially FreeBSD under the covers, and so is a flavor of *NIX).

    There were a lot of interesting ideas floating around and being tried. Some of them were, indeed, crap, simply different for the sake of being different, and some of them were absolute gems (Symbolics Lisp Machine, anyone?).

    The MFELAB Cray CTSS (Cray Time Sharing System, not to be confused with the MIT CTSS or Compatible Time Sharing System) had a lot of features that I miss today, such as drop files, which essentially contained the entire state of the machine, including I/O, all in one convenient package.

    If you wanted to (and sometimes you did) you could pause the execution of a job, copy the drop file to another completely different machine, and, under certain limited circumstances, restart the job as if nothing had happened.

    And, of course, the MULTICS system had security features and privilege separation which are missing from any "modern" operating system.

    There's no question that the hardware is objectively more powerful today but by its very ubiquity, it's not as interesting. The operating systems? Yawn.

  • Warthog (unregistered) in reply to Captaffy

    So, how was anyone supposed to deal with this if they already have a real job?

    First day: "I have a dentist appointment tomorrow, can't come in today."

    Next day: "Uh, I gotta go back to the dentist all day tomorrow too."

    Next, next day: "Um...teeth exploded...gotta go back"

    etc....

    Were they only hiring recent graduates?

  • Kuba (unregistered) in reply to jpers36
    jpers36:
    The Ab Initio suite is the sole product of Ab Initio Software Corporation, which was founded by Sheryl Handler and various other former employees of TM. I don't know who slapped those guys upside the head with the compentency stick, but AI is an excellent tool, even if it has its own WTF or two.

    Oh well, I see that Sheryl's ways stay with her. Ab Initio seems to be a WTF of a company: who the heck offers a software platform that has no publicly viewable documentation to speak of, and is not really available? As in buy online / download a trial / see code samples, etc? If their approach was any good, it'd find lots of use at small businesses too. Lunacy...

Leave a comment on “Thinking Machines”

Log In or post as a guest

Replying to comment #:

Log Out

« Return to Article