As our host has pointed out, the AI community has spent about 50 years being fairly bogus, and grossly underestimating the processing power required to solve interesting problems.

Many introductory AI courses still focus on A* and formal reasoning, which basically have nothing to do with intelligence. But there's a lot of good work being done, too--computer vision is slowly becoming useful; statistical databases of common sense can support modest inference engines, and Kurzweil just demoed another vision device.

Give me another 1,000-fold increase in computing power (beyond current stream processors), and make it portable, and I'll give you some pretty mind-blowing wearable computers. And the accompanying advances in robotics would basically transform the economy--a few hundred cheap teraflops dedicated to vision and physics would make manual labor almost obsolete.

...Stirling, the technology required to support a direct neural interface would presumably involve: (1) serious processing power, (2) a rough understanding of a few major brain subsystems, and (3) mixed electric/biological engineering bordering on primitive nanotech. I could see it happening (late) in my lifetime, sure, but it would be an awfully exciting lifetime--more like 1900-to-2000 than the future you described.