I've thought about PowerPoint-like (Keynote-like) presentations for years.  Probably because I've been around congitive psychologists quite a bit, I decided that the slide  and corresponding speech should present the material  -- maybe as cues -- in a decidedly different way, that way the audience could choose whatever works for them: aural or visual, text or graphics. This might be a totally wrong way of approaching talks, but it's worked for me.
I'm not sure why, but I've had three PowerPoint-related epiphanies at HICSS conferences: one was during lunch  talking to a guy who was adament that slides should be essentially a transcript of what is spoken.  He was a cognitive psychologist, and I was really disappointed :)
The second time was at the HICSS dinner/luau someone the same day I'd given a presentation in the Digital Documents track.  A fairly well known guy, although I can't remember his name, said that I had the best slides of the track.  This was after he'd had a few mai tais, but I still take it as a complement :)
The third time was at a HICSS plenary presentation  by Elizabeth Monk Daley called "Expanded Concepts of Literacy". She was the dean of USC School of Cinema-Television. What I remember about her slides are there were almost no words. I remember stick figures, and slides that reminded me of storyboards.
Besides the Tufte stuff that I've blogged previously, you might want to look at the following:
