Hacker News new | past | comments | ask | show | jobs | submit login
The Story of Spikey (2018) (stephenwolfram.com)
39 points by bookofjoe 32 days ago | hide | past | web | favorite | 13 comments



What's with the surge of Wolfram stories on Hacker News lately?


Might be connected to them opening up for free access to the Wolfram language, letting a wider audience play around with it.


Great question. I've posted several over the years and most never make it to the front page.

There are some physics, engineering, and math types on HN that can really use the functionality of Mathematica (and it is awesome), but the usual crowd here is more on the software engineering side (building systems and not simulations with differential equations and such) abd thus not as excited about Mathmatica related posts. Yes, I use Python/Numpy/Sage and Julia too, but they have a ways to go.

Recently Wolfram made some strides by making free some of their stack (arguably the less important parts) albeit with a crazy license. That caused a lot of discussion as to why they did it and how useful it was. This was of interest to this crowd (the people who would maybe use the software engineering part of their stack).


> making free some of their stack (arguably the less important parts)

How is the core of their product the less important part?


Because Mathematica is the part that actually solves equations and does file parsing, genetic algorithms, 3D visualizations, blockchain...etc.

It looks like you still need an expensive and closed source Mathematica license to go with the engine part.

That is why there was so much discussion...most HN commenters we're trying to figure out what Stephen's announcement actually meant. The fine print also gives them the right to audit your use and you have to respond with what they need in 10 days, so I'm not sure how many people would want to deal with that if they don't have to.


I'm not sure you understand the power of the engine.

I can run the following in a jupyter notebook

  image = CurrentImage[]
  faces = FindFaces[image]
  Show[{image, 
    Graphics[{EdgeForm[{Red, Thick}], FaceForm[], 
      Rectangle @@@ faces}]}]
and it will take a photo using the computer's camera and draw a box around my face.

Executing

  DSolve[{y'[x] == y[x], y[0] == 1}, y[x], x]
returns

  y[x] -> e^x
Edit: The following demos also work:

https://www.wolfram.com/mathematica/new-in-10/enhanced-graph...

https://www.wolfram.com/mathematica/new-in-10/enhanced-3d-im...

https://www.wolfram.com/mathematica/new-in-10/symbolic-geome...


The engine is quite powerful and allows most of the math and knowledge features from Mathematica, true. But playing with it via Jupyter I see how they've purposely crippled some of the more useful features. For example, Manipulate (the interactive feature where parameters can be adjusted via sliders) doesn't work, but Jupyter itself supports such interactions in Python, so it's not Jupyter's fault. Also, in Mathematica entities can be clicked on and explored, but in the Jupyter interface an entity is displayed as a static image (that deceptively looks like a button but can't be interacted with).


I don't think many of us on the main thread earlier this week did. Please explain why they still sell expensive Mathematica licenses if it is all now free (note that I'm not saying it isn't worth it). What are we missing?

Regardless, thanks for correcting my error!


Because of the "crazy" license (I would say it's quite reasonable, essentially it disallows using it for anything else than development or in the context of non-commercial personal projects), the user interface [1], the support... I don't use Mathematica myself, I don't know what functionality is missing.

[1] the free wolfram engine is a command line (or API) application and the Jupyter interface is much more primitive than the real thing: https://github.com/WolframResearch/WolframLanguageForJupyter


I would guess that over half of their current customers use Mathematica for development and prototyping (but don't put anything into production). This would effectively mean that they would let go of ~50% of their revenue over night (we're this to be true). So that line of thought is why I'm so confused. Does that make sense?


My impression was that they did it to gain new customers. People who didn't use it because they could only use it for prototyping but up to now weren't able to directly use their solution in production. Where production includes internal not-customer-facing tools, for example a custom UI or webpage for engineers or financial analysts where the backend calls into Mathematica as opposed to for example Python and its math libraries.


> up to now weren't able to directly use their solution in production.

That’s exactly what you __cannot__ do with their new “free wolfram engine for developers” license.


Thanks for pointing that out!

I hadn't followed the original discussion. I mistook the new license for a simplified runtime/production license. Looking at the announcement and the licensing terms again, I don't understand the benefit. At least, I don't feel more inclined to give it a try than I was before. (It's a great product but I still wouldn't base a project on it.)




Guidelines | FAQ | Support | API | Security | Lists | Bookmarklet | Legal | Apply to YC | Contact

Search: