23 thoughts on “Wolfram Alpha

  1. First off, I love Mathematica, I think Wolfram Alpha is going to be a terrific tool, and I’m a Stephen Wolfram fan (despite NKS). But I agree with Jordan that his computational knowledge engine is still a bit underwhelming after all the hype:

    Mathematics – A/A+ in equation solving, integration, and many other areas outside of probability/combinatorics and things like complexity theory. It’s in the B-/F range for the latter categories. Many areas covered in Mathematica are also strangely missing… for example, why doesn’t it know what a Gray Code is?!?

    Physics – B+ for relativity, classical mechanics, some particle physics… but it doesn’t know the Schrödinger equation? Near zero knowledge of polymer physics, thermodynamics (aside from simple gas law stuff), etc.

    Chemistry – “C”, nice displays for various chemicals, but the coverage is severely lacking. Also, near zero knowledge of reactions (even really textbook stuff like the Michael reaction), synthesis, and things like transition metal coordination compounds.

    General reasoning – It knows who Utada Hikaru is, and how old she is (A++), but why can’t I divide the length of Mount Everest by her age? Lots of problems with conditional statements, etc.

    I don’t want to write too long of a comment, so I’ll stop here. Like I said, W.A.’s going to be a nice tool to have around, but it’s not something I’ll use every day, and it certainly won’t be my first stop for reference material.

  2. Jordan and Robert.

    I find WolframAlpha totally amazing. It’s a bold intermediate towards AI.

    Pleeze!! Is an operating system bug free on it first release? This is 100 times more complex than an operating system.

    Star Trek predicted the cell phone and now it turns out it predicted WolframAlpha
    (Captain Kirk) Computer, give me a map of the trajectory of the Hubble satellite and pinpoint its current position.

    Just like Geordie, I also was inspired to write a blogpost extolling WolframAlpha

  3. Q: “sides on loonie”
    A: “Wolfram|Alpha isn’t sure what to do with your input. Related units: loonie” (the link to loonie just gives a few exchange rates)
    Q: “how many sides are on a loonie?”
    A: “Wolfram|Alpha isn’t sure what to do with your input.”
    Q: “how many sides are on a canadian $1 coin?”
    A: “Wolfram|Alpha isn’t sure what to do with your input. Related inputs to try: canadian coin”
    Upon clicking on “canadian coin”,
    A: a map showing the distance from Canada to a city in Spain named “Coin”
    Q: “:(”
    A: “Wolfram|Alpha isn’t sure what to do with your input.”

    Maybe this is a bad example, but I actually haven’t yet found a question that I want to ask that it understood… one spat back what appeared to be just its lexical splitting of the input (input “cycles of cpuid”), but also with no information suggesting it understood what I was asking. I’ll be sticking with Wikipedia for a while yet. 😦

  4. That’s a good point. It is just an alpha release (unless the product name is actually “Alpha”), and considering the problems I’ve had with my own stuff, I’m not in any position to complain about the quality of alpha releases.

    However, I wonder if “alpha” has become the new “beta”.😉 With great “beta” products like GMail, the excuse “it is just a beta release” doesn’t really work so well anymore, haha.

  5. In other news, N.K.S. was just referenced in a Nature News and Views article about what computation can tell us about the limitations of reductionism in the physical sciences.

  6. It is certainly very impressive, and a great demo. As some have pointed out here and elsewhere – there are large gaps of data. This leads some users to think that it is useless because it could not compute the answer they sought. The reality seems to be that Alpha has a very high capability in computing some data, but other data is completely missing. However, the user has very little idea what data Alpha actually has access to. So, the main weakness here is a user-interface problem – how to inform the user about what data is available and what is not. I assume that Alpha will continue to get better at giving clues to the user about data it has access to. And, of course, the databases will continue to grow.

    The other challenge for the user is one of syntax. It seems that Alpha can be very powerful if you phrase your query properly. But, an improper phrasing may result in nothing. I know that I will have to learn better what sort of syntax works best with Alpha.

  7. It’s complex analysis capabilities are also lacking… it doesn’t know what a contour integral is, it doesn’t know how to compute a residue (something Mathematica can certainly do), and it only gives branch point singularities…

    • it knows how to calculate a residue, you have to use the exact mathematica syntax for it:
      try Residue[Cos[m x]/m,{m,0}]

  8. Filippo,

    Well alright then. I could swear that wasn’t there before… but I could certainly have made a mistake formatting the input, and in any case, I left them a note about it. Thanks🙂

    I don’t mean to be so picky. My problem with W.A. is that, since it’s come out, it’s failed me every time I’ve had a real question or calculation that the Google calculator feature couldn’t handle. And I’ve heard the same story from everyone I’ve talked with inside and outside my lab group.

    • Robert,
      on that I agree. Plus, they should post a tutorial on the syntax that one can use, because I think that many of the failed attempts are due to a poor knowledge on how to write queries and a yet small amount of data. (Well, the total is already huge, but since it has to cover so many fields, it’s not so deep, that’s what I mean)

      • Filippo,

        I think what really bothers me is the fact that W.A. a nice framework, Mathematica, and a large number of highly competent computer scientists and mathematicians behind it. But instead of focusing on, say, making a system to reason about and solve (less-than-simple) problems in probability/combinatorics/
        analysis/etc., it’s (literally) trying take on the world.

        What you’re left with is a collection of (sometimes flawed) apps and lookup tables for a few toy problems in many diverse areas… that are largely useless to ‘experts’.

        -Knowing a few formulas for black hole surface area/entropy/etc. is a long ways from knowing astrophysics (and I certainly mean to imply that *I* know anything more than this).

        -Not knowing the twenty common amino acids (it’s missing isoleucine)… is a really bad sign.

        -Exact sequence lookup in a mystery human genome (i.e. edit distance can only equal zero)… is almost worthless. You’re going to compete for traffic with the NCBI site with this?

        -Non-rotatable ball-and-stick/space-filling models?

        -No knowledge of chemical reactions or reaction mechanisms whatsoever?

        Alright, well I think I’m going to stop commenting on this because I’m beginning to feel bad. I can tell a lot of effort went into W.A., and I wish them the best. But I also really really wish they would focus on something… anything.

  9. Don’t worry, the blog’s not over. (I think)
    He’s been pretty busy recently. There should be interesting stuff to report soon, as there often is; how soon just varies.🙂

Leave a Reply

Please log in using one of these methods to post your comment:

WordPress.com Logo

You are commenting using your WordPress.com account. Log Out / Change )

Twitter picture

You are commenting using your Twitter account. Log Out / Change )

Facebook photo

You are commenting using your Facebook account. Log Out / Change )

Google+ photo

You are commenting using your Google+ account. Log Out / Change )

Connecting to %s