I wouldn't say that I'm a good speaker, but I'm certainly a much better speaker than I used to be. It's not just about transmitting a certain number of bits of information per minute; it's also about making sure that those bits are being received at the other end. I often throw jokes (and quasi-jokes, like my "purpose of cryptography is to force the US government to torture you" line) into talks as a way to help keep the audience's attention; and I watch the audience for signs that I'm moving too fast or too slow for them.

But for all of this, I don't think the material I convey has suffered in the slightest. One audience member told me that my cryptography-in-one-hour talk was the "most densely packed hour of information" he had ever seen. If being a good speaker pushed me away from having and conveying good ideas, my talks should have been getting progressively less informative, not more so.

I posit that while PG is seeing a real effect, it's not the effect he thinks he's seeing. Rather than style detracting from substance, it seems to me that there's selection bias: In order to be invited to give talks, you must have at least one of {good ideas, good style}. As a result, those talks which are completely devoid of interesting ideas are inevitably given very well -- we never see talks which are given by poor speakers who have no interesting ideas. This in no way means that speaking well is responsible for the lack of substance.