Wolfram Alpha Needs an Open Source Version | Hacker News Reader

Wolfram Alpha Needs an Open Source Version

pointlessrants.com

79 pointsphreeza36 comments

In the very long term, Mathematica needs an open-source equivalent. Someday Wolfram might open it; it would be a gift to humanity.

It depends which parts of Mathematica you need.

Python is an alternative, especially within the Sage package [0].

Then there is also Maxima [1].

[0] http://www.sagemath.org/ [1] http://maxima.sourceforge.net/

Indeed, there are plenty of open source solutions that can solve your particular problem, and many problems don't need any sort of symbolic tools. There's also Octave with a nice collection of packages on Octave-Forge, there's PARI (included in Sage), and there's LAPACK, and I'm sure there are others I'm either forgetting or haven't heard of.

I use Octave daily (in the last hour, will return to it when I finish my soup), but have never seen anything that matched Mathematica's symbolic/pure math functionality.

Will check out Sage, haven't used Maxima since 2004. Thanks!

Edit: Sage looks nifty, Maxima looks unchanged, at least at screenshot-level.

SymPy (sympy.org) is a well-developed Python package that works with symbolic/analytical mathematics on a level comparable to Mathematica's.

Octave is more of an open source Matlab equivalent. That is, mostly linear algebra.

I too can attest to the extensive abilities of the Sage package. It's extremely powerful and easy to use given that it's based in Python. Even if you don't want to download it, give the notebook a try: http://www.sagenb.org/.

It stretches the definition of equivalent but folks interested in an open-source version of the Mathematica language should see or contribute to http://www.mathics.org/

I'm not sure what you can do in Mathematica that you couldn't do in R http://www.r-project.org

There's "can do" and then there's "is like."

It's been a long time since I used Mathematica, but I always used it for symbolic computations (e.g. "derivative of x^2 = 2x"), especially when I was learning calculus and differential equations. As far as I'm aware R doesn't even natively have a way of storing symbolic equations, let alone solving them (at the least, it's not its main use-case).

But I could be wrong; I used mathematica and R for very different things so I may just be unaware of their similarities.

https://code.google.com/p/ryacas/ to name one. It's not native but yet so aren't a lot of libraries in other languages.

Hasn't been updated since 2007... math doesn't really change that often but that does not look promising...

In principle R language is equally potent than Mathematica's one (functional+self-modifying), so Mathematica clone can be done in R (and in fact there is a built-in function "D" which calculates derivatives of native expressions symbolically).

Of course this won't ever happen because of the paradigm -- R is for data crunching and does this way better than Mathematica.

Mathematica provides symbolic computation, R does not.

Hats of if you can do http://blog.wolfram.com/2012/10/23/calculating-the-energy-be... in R or, for that matter, about half of the other examples given in that blog.

> Mathematica needs an open-source equivalent.

Mathematica _has_ open source equivalents. It has equivalents that predate Mathematica. Their libraries just aren't as extensive.

Sympy (written in python) also provides a number of features.

Maybe we need an open-source Mathematica or Wolfram Alpha, but I don't think Wolfram would benefit of either being open-source.

He'd be remembered as having given a gift to mankind, which he might like very much.

His very large book was the "gift" to the public. Now we just need to find the "rule" that explains the universe.

Wasn't that NKS?

Seriously though, he did seem open to the idea of at least partially open-sourcing Mathematica during a recent AmA: http://sx.reddit.com/r/IAmA/comments/qisot/im_stephen_wolfra...

I think there are at least three different ideas being discussed here:

1/ A web-calculator which does more advanced functions. See http://live.sympy.org/#example and Sage Web Notebooks -- hell even google does quite a few things.

2/ A natural language science search thingie -- i.e. a smart algorithm which understands what data to shoot at you given your query. (query recognition ML? NLP?)

3/ An open source / community-editable platform for editing scripts for 2/.

I can imagine there are some interesting lessons about machine learning and NLP to be learned from such a project. In particular 3/ seems like a good thing to have in general: functional modules (transformations?) which take some input data and produce another type of data.

The components 1/, 2/ and 3/ above would work really well if restricted to a specific application domain. This way we would have a chance of competing with Wolfram and understand verbs from that domain + use structured data from dbpedia et al.

Very cool. pm me if you are working on any of the above. I would like to help.

I've actually have been working on a site that does this exact same thing for the past few months. I could definitely use some help from developers - feel free to drop me a line if anyone wants to talk.

I personally feel that calculator software should be open source and easy to modify, it not only allows people to learn the background processes by inspecting the code, but it has numerous other educational benefits.

Are you planning on making it modular? If you made it something similar to what DuckDuckGo (http://duckduckhack.com/) does that would be very interesting. I would also contribute if that was the case.

I've also been working on something similar. I will get in touch.

I'm also thinking of something similar as a potential part of my 'tiny-chunks-of-reference-info' project. That is, some of the chunks would be parameterizable with active computation.

There might be a bunch usable from the desktop 'Qalculate!' for formula/units interpretation:

http://qalculate.sourceforge.net/

(Converting 'Qalculate!' into an all-Javascript, single-web-page application might be an interesting project, too...)

It will be rather hard to compete with 15 million lines of Mathematica and over 25 years of research but you have to start somewhere http://en.wikipedia.org/wiki/Wolfram_Alpha#Technology

This blog post is from 2009, I think Wolfram Alpha was pretty new back then. It would be interesting to know what he was looking for and if it still isn't there.

His example of a query that doesn't work seems to return data now, when typed slightly differently:

Not working link taken from blog post:

http://www.wolframalpha.com/input/?i=graph+u.s.+unemployment...

Working result:

http://www.wolframalpha.com/input/?i=graph+u.s.+unemployment...

I think it would be great to have a nice open source symbolic math package with a good UI, but for the 3 times a year that I need to know how to factor some equation like (x-1)/(sqrt(x)-1) , wolfram alpha does a pretty good job!

I need to explore sage and mathics more, they both look cool

While I get the point, wikipedia works because people with weird hobbies can get away with just text if they want to.

getting data into the right shape for gnu|alpha or wolfram|alpha|community is probably not as trivial

Check out the dbpedia graph, they're doing much of this already. For example, for structured info about San Francisco: http://dbpedia.org/page/San_Francisco

Here are their datasets: http://wiki.dbpedia.org/Datasets

He doesn't really mean 'open source'. He means "user contributed content". Which is an entirely different thing. You could run wikipedia on closed-source/proprietary software if you wanted to.

But how else would you preserve the myth of Stephen Wolfram's unique genius than by keeping Wolfram stuff proprietary?

Well, if Sergey Brin had finished his intern project we would have an open source version of Mathematica :-).

Say, what?

Sergey interned at Wolfram Research: http://www.wolfram.com/company/scrapbook/page07.html#1993_ML...

wolfram doesn't need to be open source. you want/need and open source version of wolfram.

Reply on news.ycombinator.com