If your looking for the next 5 years..
1. Multicore and grid are vital.
Multicore is a given. Intel and AMD will have 8 core chips in general desktops when this product is in the prime curve of its use.
Grid is more of a luxury. But making asychronus threading easier goes a long way toward that.
In a perfect world I would like to be able to just recompile and see substanital multicore improvements. You guys know the library calls that can be split up such as graphics, matrix math, distributed sorting, regex on bioinformatic scale datasets and many
Regex would be a good one because I love to show Perl snobs that I can do bioinformatics in C# and VB.NET. I took the O'Reilly books and converted them to C#. Yes I was bored.
That would show the true strength of "inheritence."
2. Take out Adobe/Macromedia .. Sorry can't sound too antitrust.
Build tools that make them moot.
3. Ajax is a hint, but still an infant technology. The core question is how work can I do with web applications and why should anything be on my hard drive? Can't it sit on my server or on the net.?
4. Video as a data source. With facial and pattern recognition your on the edge of having programs look at pictures. Then they can tell you who or what is in the picture / video. That is the next major fulcrum.
5. Same with text .. read a document and write a searchable synopsis. Use machine inference as the gateway to machine learning.
6. On a personal note.. Sneak a copy of FoxPro back onto the disk. Fox does a lot of data clean up work that VB and C# don't do well. I shouldn't have to buy it seperately. Don't Let Fox Die!
I talk more about computers at braxtonperry.blogspot.com I also challenge Marshall Brain ( a CS professor and NC State) on his blogs about AI.