It's more than that I believe. I'm currently enrolled at a masters degree level at the university, having 5 classes to do with AI (data mining, neural networks, fuzzy systems, semantic web and genetic algorithms), having also completed "basics of AI" undergraduate course.
However, I'm still taking both Machine Learning and Artificial Intelligence Stanford classes, because they tend to focus on things we aren't doing much (and vice versa). Admittedly, I mostly dropped Artificial Intelligence since I own and read the book, and since the class doesn't feel worth the time with so many calculations going on (the probability part requires you to spend 50% of time doing calculations, which isn't hard, but just takes time -> it should be a part of Probability 101 not Intro to AI).
The Machine Learning class is amazing though, in particular the extra insight professor Ng gives when describing algorithms and their usage, as well as the great learning system (review questions and most importantly programming tasks). Also Machine Learning tends to focus on giving you actual knowledge, rather than trying to test you accurately - good tests that rank students are important for classic studies where diplomas and thus scores are the end goal, which is not the case of online courses, which should be about learning.
Lately, I've been using Windows more and more, and for the simple reason of Linux still messing around with the basics - having trouble with xorg.conf/X11 again, after 3 years, is a letdown. Additional problems I didn't see mentioned are: 1) Disk/Network scheduling: Windows actually does much better disk scheduling, I don't have to fear being temporarily disabled to use the PC/open sites(listen to music) when transferring files or downloading stuff because no one still implemented/defaulted to a proper disk scheduling/network QOS that emphasizes on latency instead of bandwidth. 2) Constant fiddling with configurations I've already set. Yes, admittedly I am using Arch Linux, but that is because most other major distributions failed at what Linux should excel, which is users ability to configure things for what they need. The problem with the current state of things is some perverse desire to change old, working configuration methods (adding new mandatory files, changing syntax or semantics, adding and removing mandatory fields), which gives every update an added 15 minutes fixing period. 3) Major DEs and associated programs crash, much more often than in the case of Windows. For that reason, and because I dislike having a smartphone interface on a 23'' monitor, I'm still using Openbox, and the fact is, there has been barely any new apps (things you need, like menu bars, taskbars and similar) in that region.
But the reason I still stick with Linux is the obvious supremacy of console applications when it comes to programming. Even if I do stuff in IDEs such as Eclipse for Java programming, being able to easy monitor files, create scripts in a much more enjoyable way than for the windows batch, I still have an upper hand developing in Linux. And if they would add something as good as the Linux console and tools in Windows (and not through 3rd party stuff like cygwin), I would probably make a complete switch to Windows.
Aside from a single crypto class I had at my university, and a friend who's an expert at these things, I don't know much.
However, from what little I could grasp from the summary, they were using salts (and hashes, which is the bare minimum) to save passwords. The main idea of salts is to prevent people using rainbow tables (precalculated password -> hash mappings), and just doing reverse lookups to obtain a password from hashes. However, it still doesn't mean any real security if they didn't use at least something as good as bcrypt for hashing (bcrypt actually encrypts salts with hashes iirc), MD5 and SHA can be cracked fast enough on todays computers.
I'm much more worried about credit card information, how exactly have they been encrypting it (remember, they had to access it themselves)? They had to keep the keys for decryption somewhere, and it's worrying if those keys are compromised.
Yeah, sure, it is indeed fast, but it does feel like the aim Asimo has is to achieve all goals and solve all problems modern of modern AI and robotics.
Focusing on complex terrain walking, jumping, object manipulation, image and voice recognition and probably many more things not seen in the video is too much. And the problem of doing so many things in once is that once you fail one thing, the whole robot becomes of variable usability - if it fails walking, trips and falls, it will be of little use to people (same applies for failing to properly comprehend commands or detect objects).
In comparison to the american (from my European standpoint) projects, like the grand DARPA challenges that aim to solve only a few things at once and may be aiming at a near future practical usage : autonomous cars, I feel Asimo will remain out of any real practical usage for a long time, until every single of those things has been nearly perfected.
Slashdot Mirror
It's more than that I believe. I'm currently enrolled at a masters degree level at the university, having 5 classes to do with AI (data mining, neural networks, fuzzy systems, semantic web and genetic algorithms), having also completed "basics of AI" undergraduate course.
However, I'm still taking both Machine Learning and Artificial Intelligence Stanford classes, because they tend to focus on things we aren't doing much (and vice versa). Admittedly, I mostly dropped Artificial Intelligence since I own and read the book, and since the class doesn't feel worth the time with so many calculations going on (the probability part requires you to spend 50% of time doing calculations, which isn't hard, but just takes time -> it should be a part of Probability 101 not Intro to AI).
The Machine Learning class is amazing though, in particular the extra insight professor Ng gives when describing algorithms and their usage, as well as the great learning system (review questions and most importantly programming tasks). Also Machine Learning tends to focus on giving you actual knowledge, rather than trying to test you accurately - good tests that rank students are important for classic studies where diplomas and thus scores are the end goal, which is not the case of online courses, which should be about learning.
Lately, I've been using Windows more and more, and for the simple reason of Linux still messing around with the basics - having trouble with xorg.conf/X11 again, after 3 years, is a letdown.
Additional problems I didn't see mentioned are:
1) Disk/Network scheduling: Windows actually does much better disk scheduling, I don't have to fear being temporarily disabled to use the PC/open sites(listen to music) when transferring files or downloading stuff because no one still implemented/defaulted to a proper disk scheduling/network QOS that emphasizes on latency instead of bandwidth.
2) Constant fiddling with configurations I've already set. Yes, admittedly I am using Arch Linux, but that is because most other major distributions failed at what Linux should excel, which is users ability to configure things for what they need. The problem with the current state of things is some perverse desire to change old, working configuration methods (adding new mandatory files, changing syntax or semantics, adding and removing mandatory fields), which gives every update an added 15 minutes fixing period.
3) Major DEs and associated programs crash, much more often than in the case of Windows. For that reason, and because I dislike having a smartphone interface on a 23'' monitor, I'm still using Openbox, and the fact is, there has been barely any new apps (things you need, like menu bars, taskbars and similar) in that region.
But the reason I still stick with Linux is the obvious supremacy of console applications when it comes to programming. Even if I do stuff in IDEs such as Eclipse for Java programming, being able to easy monitor files, create scripts in a much more enjoyable way than for the windows batch, I still have an upper hand developing in Linux. And if they would add something as good as the Linux console and tools in Windows (and not through 3rd party stuff like cygwin), I would probably make a complete switch to Windows.
Aside from a single crypto class I had at my university, and a friend who's an expert at these things, I don't know much.
However, from what little I could grasp from the summary, they were using salts (and hashes, which is the bare minimum) to save passwords.
The main idea of salts is to prevent people using rainbow tables (precalculated password -> hash mappings), and just doing reverse lookups to obtain a password from hashes.
However, it still doesn't mean any real security if they didn't use at least something as good as bcrypt for hashing (bcrypt actually encrypts salts with hashes iirc), MD5 and SHA can be cracked fast enough on todays computers.
I'm much more worried about credit card information, how exactly have they been encrypting it (remember, they had to access it themselves)? They had to keep the keys for decryption somewhere, and it's worrying if those keys are compromised.
Yeah, sure, it is indeed fast, but it does feel like the aim Asimo has is to achieve all goals and solve all problems modern of modern AI and robotics. Focusing on complex terrain walking, jumping, object manipulation, image and voice recognition and probably many more things not seen in the video is too much. And the problem of doing so many things in once is that once you fail one thing, the whole robot becomes of variable usability - if it fails walking, trips and falls, it will be of little use to people (same applies for failing to properly comprehend commands or detect objects). In comparison to the american (from my European standpoint) projects, like the grand DARPA challenges that aim to solve only a few things at once and may be aiming at a near future practical usage : autonomous cars, I feel Asimo will remain out of any real practical usage for a long time, until every single of those things has been nearly perfected.