GPUs do support branching, though it causes a performance hit if nearby threads do not take the same branch.
They also have a form of cache now. You can switch it from a hardware-managed cache to a user-managed cache at will. Its not as large and has some limitations.
I disagree -- we don't need molecular-level detail, but we do need more detail than machine learning modelers (including Numenta) put in now.
Existing models don't spike, so there is no fine temporal coding. They don't have interesting recurrent architectures, so they don't do things like oscillate, a ubiquitous phenomenon in the brain. They don't use local synaptic plasticity rules to train themselves in an unsupervised way but rather try to do supervised learning. And they are not at scale (mostly because said training algorithms do not scale).
I don't think these are wetware implementation details. They are important. And they are more or less ignored right now, particularly in conjunction.
It might want to purchase an ads company of some sort to start competing with Google more fiercely in the lucrative ad space.
ARM seems unlikely given that their previous purchases have gotten them an ARM license and development team, and its an expensive purchase just for talent.
I would actually go with nVidia -- this would give them both a CPU and GPU firm in-house, and $7 billion is a bargain for that IP and talent, especially since the company is profitable. They also have interesting products coming out -- compute-specific processors and mature technology selling into the high-margin research market, 3D vision, PhysX, etc. -- that Apple could integrate deeply into their other products (Mac Pro, AppleTV, OS X, etc.).
So what happens when a social network starts that includes links to your friends current Facebook profiles but guarantees much stronger privacy? Think Friendfeed but more ambitious.
They're still using desktop operating systems and software while everyone else is using mobile for everything, even when an app has no place on a phone (AutoCAD, really??)
So lets go ahead and predict a renaissance of desktop operating systems and using computers to get shit done.
I think the key question is whether Microsoft will agree to support it in Internet Explorer. They have a competing format, JPEG XR. It has been standardized. Microsoft's patents on it are royalty free under their Community Promise. It is already implemented in Windows 7 and IE9 and several third-party applications on Windows. They distribute source code that can work with the format on several platforms, but that code is incompatible with GPL and similar licenses (BSD-like licenses are fine.) An independent implementation of the standard that is compatible with the GPL could be developed without legal problems though.
It may be instructive to consider how deliberate the Open Source branding and media frenzy was, and particularly the few paragraphs where the author deliberately positions himself as a spokesperson. Consider how well it worked to bring open source into the corporate world.
Perhaps more campaigns of this nature need to be fostered.
I spent the last 3 months living in a tent in the mountains of New Mexico. Between my laptop and my cell phone's tethering, there was no issue at all getting work done. And it was an incredible experience.
If you were willing to move to warmer climates during the winter you could do this all year no problem.
Young women care about manliness of the sort you describe, because they don't meet many rich guys and aren't trying to get married anyway. Women in their late twenties care about money because they interact with people with real jobs all the time and want to get married, and are probably starting to get ugly.
I always just do it recursively and go "oh did you mean efficiently?"
I mean reversing a link list is literally reverse the tail of the list recursively then append the head to the end. Base case is the empty list. Maybe 8 lines of C for reverse, a few more for appending to the end (easy) and the struct for a Node.