Actually i don't think they make programmers dumber, but i believe they let dummies think they're programmers!

As a CS guy who studied lots of hardware out of the school, i believe EE folks are more likely to be good programmers. But understanding the underlying hardware wont be enough in many cases to make someone a good programmer. Math knowledge including combinatorics and a bit of calculus is a must have if you are working on cutting edge software and by cutting edge i mean software systems that you can't find proper existing algorithms for your needs to code them well. A programmer who is good at math, algorithm design and data structures with enough knowledge of the technology/tools he is working on/with and has hardware knowledge as a gift, if is a good team worker, can be a good programmer. And a real story: A month ago, a taxi driver asked me about my job, "Computer programmer" i said. He thought for a sec, and replied "Yeah... but... that's what everyone does, i mean what your real job?!... how do you make a living?"

I'm not sure about java, but in .net there are very rare cases that you can't find proper set of assemblies and classes to guarantee performance. Except for the cases in which you need low level access to system resources or when you're designing multi-platform applicatins, .net seems to be a perfect tool in hand to make softwares. About a year ago, i was given the task make a software to control a box-camera designed for scientific purposes(16 Megapixels of 16bit, gray scaled pixels). The software had to have lots of features and had to be built in about 2 weeks. The best tool for me was .NET to ensure that i will finish the project before the deadline, but i was just not sure if the performace will be as well as if i had coded it with c++, when the software had to do lots of calculations on each pixel before drawing. The 1st version rendered a full image in about 45second which was too slow. But using the right assemblies, classes and methods dramatically improved the performance and the 2nd version rendered the same image in less than 2 seconds. same as the test c++ application i made later. The only problem with .net is that sometimes the bare algorithm using fors and ifs may not help you getting the performance you need due to the memory management system the CLR provides. (Specially when lots of cycles should be passed). Sometimes you need to use other tools (classes inside special namespaces) to get the performance you need.

there's a book i like, "Microsoft .NET for Programmers" by "Fergal Grimes".