Looks decent - I'll certainly try the demo. Though I'm no mathematician Mathematica is my favorite Maths playground software.
I'm curious about their implementation of NLP. This has enormous potential as a teaching tool - the biggest hurdle to getting more out of a package like this on first use is not knowing how to interact with it. However, I had had similar expectations of Alpha, and was greatly disappointed. though full of witty easter eggs and stocked with a rich variety of datasets, exploratory queries proved sadly frustrating.
Here's what I want to see in version 9 (laugh now, realize I'm right later): Kinect interaction. Perhaps unwittingly, Microsoft have just launched the next great peripheral and it seems intuitively popular with the public in a way I haven't seen for years and years. Now, imagine you've generated or imported a mathematically-specified 3d object in M., and imagine it inside a bounding box with handles on the vertices and local minima/maxima. Multi-point, multi-d interactivity would be both absorbing for students and potentially extremely productive for professionals, without requiring significant retooling of the core.
About a year ago, I was working on a machine learning problem. I had a lot of training data I wanted to manually classify, so I put together a Mathematica interface for the data that would let me browse the training set and assign classifications from a USB gamepad. It was pretty straightforward... only took an hour or two to implement, and it sure beat classifying with mouse clicks!
Only problem with that is the fact that Microsoft could either retool the kinect to disallow the use by mathematica. Some other third party would have to work similar to kinect for it to actually happen.
That's true, but I doubt they will - and if they do, hordes of imitators will appear. My hunch (ie I have no evidence whatsoever for this) is that MS's huffing and puffing over people's reverse engineering of the Kinect was little more than posturing designed to increase viral interest. If they really wanted to keep it limited to the Xbox they'd have forced a proprietary connector to one of the memory expansion slots or such, rather than giving it USB 2 and acting surprised when people tried plugging it into other devices.
Intentional or not, it's having a good launch - word is that it's flying off shelves, and seems to have instant mass appeal. I was at an information management technology conference a week ago, and guess which two booths were the most crowded? The ones that were trading 3 minutes of sales pitching for a go at Dance Dance Revolution or whatever the Kinect version is called. Incidentally, the other thing being used to bribe jaded conference-goers was the prospect of winning an iPad every hour - there must have been 30 exhibitors using that one.
I'm curious about their implementation of NLP. This has enormous potential as a teaching tool - the biggest hurdle to getting more out of a package like this on first use is not knowing how to interact with it. However, I had had similar expectations of Alpha, and was greatly disappointed. though full of witty easter eggs and stocked with a rich variety of datasets, exploratory queries proved sadly frustrating.
Here's what I want to see in version 9 (laugh now, realize I'm right later): Kinect interaction. Perhaps unwittingly, Microsoft have just launched the next great peripheral and it seems intuitively popular with the public in a way I haven't seen for years and years. Now, imagine you've generated or imported a mathematically-specified 3d object in M., and imagine it inside a bounding box with handles on the vertices and local minima/maxima. Multi-point, multi-d interactivity would be both absorbing for students and potentially extremely productive for professionals, without requiring significant retooling of the core.