Apture: hyperlinking made easy

I like apture because it takes something annoying like a web pop-up and turns it into a convinient, easy to use, piece of technology both for editors and readers. The idea,  incubated by Tristan Harris while working at HCI  Stanford University, was motivated to avoid loss of contextual focus when reading (for example) news in the web: following a link will fire a pop-up instead of opening a new tab (or following the hyperlink in the current pag). Videos, music, maps, … any link is viewed in its context, see a demo below:

It is also easy to use for editors. Really simple to install, it brings the content editor the opportunity to link free content pulled from many free information sources, including Wikipedia, Googe Maps, Revver, Imeem, Hulu, YouTube, and Flickr. Even PDFs of documents can be linked via Scribd! And it’s free for bloggers!

What a pitty that wordpress.com does not let us use it…

Pario: ’spheres’ to mimic reality

Carnegie Mellon University (CMU) and Intel Research have made some real progress on a thrilling set of technologies which could change the way we interact in the future. The Claytronics’s project is aimed to give tangible, interactive forms to information so that a user’s senses will experience digital environments  indistinguishable from reality (passing a kind of Turing test).

The basic idea is to build tiny spheres (catoms, nanoscale robots) that can move, store and communicate and that can be programmed to form dynamic 3D shapes of real objects. While the final solution is still years away, interesting visionary applications are spoted: 3D fax, real-presence as oposed to tele-presence (medical emergencies, business meetings), … Many technical (hardware and software) fundamental challenges need to be addressed, as well as some controversial consequences, food for tought to ethics and phylosophy.

If you happen to have 90 minutes free time, it is worth watching Todd C. Mowry’s talk at Stanford.

Efficient polynomial evaluation

How would you code a method to evaluate a n-th degree polynomial? I once needed to code this type of routine for a project I was working on. At that time I used to program fast FPU processors, so I never paid attention to performance. I went carelessly coding the straight forward approach:

void polynomial::eval(const vector<double>& coef, double x, double *res)
{
  double res=0;
  for (int i=0; i<coef.size(); i++)
    res += coef[i] * pow(x, i);
  return res;
}

The library ran smooth on our P-IV intel processor. It was used for a couple of years without problems. As time went by and product cost was pushed down, we sooner decided to move to a XScale intel processor. It was then when we noticed quite an impact on the performance (from 0.2 to 12 sec overhead). It turns out that my harmless function was called 9 million times! A mathematician friend of mine pointed me out to a more efficient algorithm: Horner’s scheme. I admire it for the elegance and simplicity of its solution: just a rewrite of the commonly used polynomial expression:

The new routine runs now 10 times faster!