More on the LifedriveCF, Linux and a taste of Numenta.

Perusing through the Lifedrive Universe, I ran into a very interesting comparative video put out this weekend by PDAColombia. Here they compared -side by side- and running simultaneously the Lifedrive in its original format running with the Lifedrive CF. The most striking difference was the reboot times. There was an obvious advantage with the CF drive, taking about 1/2 the time to reboot as the harddrive version.

However, some apps appeared to run at the same speed and load at the same times as on the hard drive. That was very surprising. Overall, the Lifedrive CF appeared more robust and appeared to be faster on most counts.

For me, this video is a Godsend. I may in fact wait until October/November when the Linux based Palms hit the shelves. I am still debating whether to buy one, not wanting to be a part of the first adopter or bleeding edge candidates who purchased the Lifedrive. But if the price of the Lifedrive - used preferably - comes down, I may buy one and hack the daylights out of it.

PDAColombia.com video

Pdacolombia videocast 10
Uploaded by pdacolombia

Palm Linux



I have also spent some time this weekend over at hack&dev to see how the linux project is coming along. Those guys have made great strides. However, I am begining to wonder if they will curtail their project now that Palm will be supporting their own version of Linux. But you should pop over there to see some of the incredible strides that they have made with OPIE.




Linux Development on Palm.

Linux on Palm Screenshots.


Numenta

Okay, for the ultimate geek weekend, I spent 90 minutes listening to Jeff Hawkins, the founder of Palm giving a lecture on Hierarchical Temporal Memory: Theory and Implementation. First of all, I was really impressed by Jeff Hawkins' incredible mind and ability to comprehend not only the anatomy, but the inner workings of the brain. In addition, I truly admired his ability to stand up to some of the world's most distinguished Neuroscientists and Neurologists. The grilling that took place following the lecture reminded me of Grand Rounds given once a week, when a complete review of a case was done. It was one of the most chilling aspects of my life many years ago, particularly if I was involved in the case. If I was as knowledgable and confident as Mr. Hawkins, I probably would have saved a few shirts from the puddles of sweat that encompassed each of these grillings.

So, why was I listening to this lecture? It was not purely to learn about Hierarchical Temporal Management, which I found absolutely fascinating, but to get a handle on what's next at Palm. After hearing about Numenta, I am convinced that the project coming out in November is not what everyone has been writing about. Numenta is not a machine, but a thought process which attempts to emulate the brains' thought processes through non-static, mobile "guesswork" that dwarfs our thinking about events. It is preemptive. It is akin to thinking about a sequence of numbers and colors before they have been produced. It is akin to knowing the difference in texture between velvet and carpet. Some of the guess work appears to involve appointing memories of other objects and things and interpolating them to come up with what you are holding in your hands. A real life example would be putting out your hand and having someone tell you that you are about to feel a cup filled with hot cocoa. You already have a preconceived notion about the scent of hot cocoa, the ceramic feel of a cup, the warmth involved and the visualization of the cup, such that when it is placed in your hand you can translate it to mean: Warm cup of Cocoa in hand. All of these things are involved in making that decision. Numenta appears to be about trying to emulate these thought processes to come to these conclusions.

One can see quite readily that this has the potential for a lot of DoD type stuff: Face recognition, object identification etc. Mr. Hawkins has stated that the early project will involve food/recipe recognition to start with. There appears to be some semblance of software out on the website to experiment with Numenta for now.

I suspect that it would be possible to have some form of recipe program just placed on the next Palm Project without much fanfair or discussion just to see how this works. Given the fact that Numenta (let's call it a program for now) works on Linux and Mac OSX, I think that the new Linux device will incorporate this well.

What this lecture did for me was to give me an idea of where Jeff Hawkins is heading. I am not certain how this fits into the Palm ecosystem, but I do foresee something along the lines of Search algorithms coming out of this. Google has been the prominent search king for now, but if Hawkins' Numenta is capable of adaptive search strategies, we may be looking at a handheld device dwarfing Google's capablities. For example, say that you want to find a picture of a yellow lemon. With Google, we would have to place file tags on the picture. With Numenta, the search program would already know that it is looking for a picture and then it would have to analyze the picture, shapes etc to come up with the answer. The obvious problem - and hence the reason why I do not believe that this is something to get excited about in November (the expected launch date for the Linux Palm Series) - is that this would require tremendous computing power - way more than a Handheld can handle.

So, 90 minutes gave me some insight into the mind of Jeff Hawkins, but sheds less light on the new devices that will be running Palm Linux.




Bye for now.
LDD

(updated, thanks to Bryan :)).