Over at c|net there's a video demo of Photo Triage that is definitely worth checking out. This is a piece of software that I would love to use. In fact, it looks to me like there are some new UI idioms that might be worth extending into the OS (and GUI apps in general):

1) automatic chronological clustering... taking advantage of the fact that if a bunch of shots were taken close together, chances are the same metadata (keywords, etc) will apply to them.

2) Quick zoom of details... in this particular application you mouse over a thumbnail and a larger version flies out to give you more detail.

3) Visual transitions between views -- in the past, if you wanted to switch from a month view to a weekly view, you pretty much had to pop from view to view, with no transition. The problem is that this can be really disorienting. The use of animation in this case is way more than just window dressing... it's telling the user how it's getting from one view to the next.

4) Quick creation of aggregating "bins". This is huge, in my opinion. Most GUI apps don't deal well with creating selection sets on the fly (e.g., I want these three from this folder, two from here, and one from another folder) or referring to more than one set at a time. The notion of having these arbitrarily created stacks makes for extremely natural manipulation of objects. Cool Cool Cool.
posted on 8/26/2004 08:47:00 AM

This is a screen shot of the future.



Ironically, it was created over a decade ago, by Baecker and Marcus. What they did in their work "Human Factors and Typography for More Readable Programs" (ACM Press,
1990) was to outline a way to make programs more readable by taking advantage of typographic resources beyond the 80-column, monospace font display that has pretty much defined the representation of computer programs since the 60s. (Note: I could be lying here... I'm not certain that 80 columns was the standard in the 60s, but you get the point.)



These two screen shots are just to whet the reader's appetite for an upcoming entry in which I will babble on and on about what programming editors *could* be.

posted on 8/16/2004 12:52:00 PM

Colorcell is a great example of a popular experiment run right.

The success of a genetic algorithm depends a great deal upon the "fitness function" you assign to it (see explanation below). In this case, the fitness function is the number of folks who click on a box of four colors. The colorboxes that are clicked on the most survive and procreate, thus ensuring that their "genes" will go on. Colorboxes "mate" by combining elements of both parents, and possibly introducing a mutation

Here's how it is done well:

1. The expriment is simple enough for most people to understand, but just complex enough to be interesting; the gene representation (4 squares of color in a box) and the function are really simple.
   
2. The outcome is interesting... it's sort of like graphic design by the masses.
   
3. Everything is presented in an interesting and appealing way (so people will actually participate).

But here's how this thing could really explode... they caught the eye of a Connector (to use a term from Malcom Gladwell's The Tipping Point), a person who is connected to a lot of people. In this case, the somebody was an editor of boing boing, a wonderful weblog of pretty much random interesting stuff. When you have a reasonably large amount of folks participating, that's when things start to get really powerful (for instance, try Amazon's recommendations... sometimes they're scarily good at finding other things you might like, based upon other people's purchases).

Well, the site got "boinged" by the number of people who visited, and they had to shut things down for a bit. That's the downside of getting the attention of a Connector... you had better be ready to deal with the sudden flood of interest.

-- update... A week later (I forgot I had this post sitting in my Draft queue) I revisited the site. Things are back up and running.

I hope folks keep visiting this site; what you might see here is a good indicator of the public's taste for color schemes. I may even steal a few schemes from the survivors. In a way, this is sort of The Wisdom of Crowds, for color schemes.

Neato.

---
p.s.: A genetic algorithm is a process that seeks to create a computational model that uses bits of ideas from sexual reproduction and natural selection. Essentially what you do is to break down bits of potential solutions into "genes" that can be randomly combined, evaluated against a fitness function, and then selected according to its fitness and "mated" with other successful members of the "population" of potential solutions. The mating takes bits from each of the parent solutions, combines them, and occasionally introduces a random mutation just to keep the gene pool interesting. Wet, lather, rinse, repeat. After a number of generations pass by, you might come up with some interesting solutions.

posted on 8/06/2004 03:35:00 PM