This is all good until you meet the positional problems between the different browsers. Until all browsers conform to the same standards (proprietary or 3rd party), some problems will exist. There are workarounds, but it depends on how important the site is to you. The best sites are not the ones that look the most interesting, but are more professional looking with meaningful content.
I disagree. The most important thing that CSS provides is the ability to quickly change sites across multiple pages. It also speeds up the downloading of the material by downloading one file that is used for multiple files. This not only saves time/bandwidth for users, but servers also experience a lower transmit cost and are able to afford a lower quota for more users.
These are the most important attributes of CSS, and in my opinion account for at least 75% of why it was developed.
As for the original question. There is almost no difference b/t slicing and dicing for a div or for a table. DIVs are harder to position, but images that span over multiple rows/columns can also make tables hard to arrange.
Whomever said slicing and dicing was useless does not yet understand how images are loaded in a page and how anti-virus programs can really affect load times. Additionally, what would you rather, missing a corner of a 500x150 image or missing the whole thing? While transmission speeds and other hardware are improving, they are still not at the point that you can just write off the slice-and-dice approach altogether.