I think the tables he chose were examples of bad table design rather than anything inherently wrong with using tables per se. It would have been more insightful had a better table example been used followed by a the new representation with the exact same data. This would have made it easier to assess the relative strengths of each approach.
I don't know the data very well, but I think the table could be improved by doing the following:
Remove the columns with only one value and make part of the header or footer.
For the columns where nearly all of the values are identical, a light highlight on the different values followed by a slight graying of the common value would have worked.
A small space could have been added to the bottom of every fifth row to make it easier to track horizontally.
The far right columns could have been aligned so that the numbers on either side of slash lined up with the ones above or below.
I think there is some promise in the circos technique, but the author would be well served to eliminate some of his rhetoric.
In particular, he should recall Tufte's advice that every visualization should serve to answer a question. Tables are very good for looking up information, in much the same way that a telephone book is good at providing the published phone number of a specific person or business. Despite reading through two examples, I am still unsure what relationships the visualization is particularly adept at highlighting.
I don't know what "agnostic to the data domain" means, but I assume that it suggests that the visualization has application regardless of the questions being asked of the data. That attitude implies that the author fails to understand what makes visualization interesting and useful in the first place.
This reminds of of some of the graphs they have in wired. While I'd love to have a table to backup every visualization, I think if you're trying to communicate the data you need to have both. I wish I knew who designs the graphs for wired, I'd follow his work any day.
This is pretty cool. The only problem is that the graphs are slightly unconventional, so people may be put off by them at first.
They may seem less intuitive than other graphs, but it's probably more a matter of familiarity. We understand other graphs so well in part because we've been exposed to them so thoroughly.
It takes a little bit of study and I was put off at first, but then it became very interesting. He's communicating quite a lot with each path, it just takes a moment to get used to. It looks to me like it will become second nature.
I don't know the data very well, but I think the table could be improved by doing the following:
Remove the columns with only one value and make part of the header or footer.
For the columns where nearly all of the values are identical, a light highlight on the different values followed by a slight graying of the common value would have worked.
A small space could have been added to the bottom of every fifth row to make it easier to track horizontally.
The far right columns could have been aligned so that the numbers on either side of slash lined up with the ones above or below.