I teach Bayesian networks and machine learning, two of the topics listed in the above article, in my undergrad and graduate courses in AI. In my opinion, these are subtopics and not core topics, and I will be the first to admit that they are by no means defining topics for a CS curriculum. People list them on resumes because they are hot topics in research, good for catching the eyes of graduate school admissions people and R&D managers.
The fundamental topcis, as the author notes, should be data structures (including graphs), applied math (including the graph theory, probability, discrete math, and linear algebra to support Bayesian networks), and basic programming and OO design. Of course, the author also has his dose of "hot topics" (XML rather than "data representation principles", abstract data types and metadata, etc.).
Graphical models are just one family of many important representations for reasoning and learning, and for computation in general, and learning is just one of a number of important computational mechanisms. Certainly, the modern computer scientist has to be aware of this.
If we look at intelligent systems as a branch of CS that is concerned with "human-like (cognitively plausible) and rational (utility-maximizing) problem solving", as Russell and Norvig set forth in the leading AI textbook (Artificial Intelligence: A Modern Approach), we see where the above topics are coming from. AI is popular. Many more people have superficial interests in AI than will actually take jobs in theoretical research or applied fields of "computational intelligence", and I think that is what this commentary illustrates.
I think grad school classmate of mine, Jesse Reichler, said it very well when Windows XP came out: marketing and flexibility are all well and good, but stability should be job one, by definition.
In the intervening four years, we've seen Job One go from security, to spam, and now back towards stability. Having a working, rock-solid kernel should be job one: if you can't make it stable, how can you make it secure?
Microsoft's stated priorities are interrelated (at least security and spam, and to an extent, security and stability). Its market share has always been built on the availability of many more applications than for other operating systems, and to a lesser degree on transparency, developer tools, and a pretty-looking interface (the "butterfly"). So, having one kernel with seven distributions is fine. Just think of it as SCO swallowing up Ubuntu, Red Hat, Slackware, Debian, etc. or OpenBSD, NetBSD, and FreeBSD.;-)
It doesn't necessarily bode well, though, given Microsoft's penchant for factoring a lot of nonessential features into the kernel. When a service pack (such as Win2K SP1) is 100+Mb, your customers are conditioned to expect bloatware. Everything I've read about Longhorn tells me there are slick features, but nothing I've read indicates that this trend shows any sign of abating.
... Chinese president Hu Jintao threw a chair at a PRC ambassador who announced his decision to emigrate to the United States, shouting that "America has screwed us" and vowing to "*#&$(&@ kill America".
MS has been trying to build bridges to China for more than seven years: first in founding Microsoft Research China (now MSR Asia), a pure research facility that eventually became an R&D wing. Kai-Fu Lee, the former exec being sued for going to Google, founded MSR China in 1998 and came back to Redmond in 2000.
Several years back Ballmer shook hands on a $100M outsourcing plsu $20M investment deal that senior management found it hard to live up to, and so they amended it to $55M in jobs and $60M in investments in Chinese IT.
If you look at the Seattle Times article, there are links to the court filings, including KFL's deposition, where he indicates that a major part of his work on MS's international business since coming back to the states, has been keeping MS from making blunders. These include making ill-advised promises that it hasn't been able to live up to, and (if it had) would have meant outsourcing at a rate that would strip American jobs, despite its assurances.
Mirosoft's China strategy is starting to fall apart as its hiring has slowed stateside and it becomes evident that it's trying to have its cake and eat it too.
There's a Chinese saying about learning: it is like a leaf in the stream; if it does not move forward, it will be swept back. Google's managing to hold on to market share is a nontrivial achievement.
That said, there's a "them that has, gets" mechanism at work here. Just as we will see when Vista's out, user inertia is strong. Baidu may have its work cut out for it.
I shouldn't be surprised any more at trash talking after the CEO of the Dance and "Developers, Developers, Developers", but that's all it is: another boorish remark.
Ballmer seems to be one of those CEOs who equates being overwrought with being enthusiastic.
Discouraging employees from leaving is just one side effect that Microsoft is trying to achieve using this case as a precedent-setter.
By making it less worth while for companies such as Google to hire away its best talent, Microsoft is also decreasing the outward osmotic gradient.
As the founder of MSR Asia, Kai-Fu Lee brings unique relationship with Chinese researchers with him, but your typical R&D tech staffer will have a harder time "being worth it" if they have to cross the iron curtain, so to speak.
Slashdot Mirror
I teach Bayesian networks and machine learning, two of the topics listed in the above article, in my undergrad and graduate courses in AI. In my opinion, these are subtopics and not core topics, and I will be the first to admit that they are by no means defining topics for a CS curriculum. People list them on resumes because they are hot topics in research, good for catching the eyes of graduate school admissions people and R&D managers.
The fundamental topcis, as the author notes, should be data structures (including graphs), applied math (including the graph theory, probability, discrete math, and linear algebra to support Bayesian networks), and basic programming and OO design. Of course, the author also has his dose of "hot topics" (XML rather than "data representation principles", abstract data types and metadata, etc.).
Graphical models are just one family of many important representations for reasoning and learning, and for computation in general, and learning is just one of a number of important computational mechanisms. Certainly, the modern computer scientist has to be aware of this.
If we look at intelligent systems as a branch of CS that is concerned with "human-like (cognitively plausible) and rational (utility-maximizing) problem solving", as Russell and Norvig set forth in the leading AI textbook ( Artificial Intelligence: A Modern Approach ), we see where the above topics are coming from. AI is popular. Many more people have superficial interests in AI than will actually take jobs in theoretical research or applied fields of "computational intelligence", and I think that is what this commentary illustrates.
I think grad school classmate of mine, Jesse Reichler, said it very well when Windows XP came out: marketing and flexibility are all well and good, but stability should be job one, by definition.
;-)
In the intervening four years, we've seen Job One go from security, to spam, and now back towards stability. Having a working, rock-solid kernel should be job one: if you can't make it stable, how can you make it secure?
Microsoft's stated priorities are interrelated (at least security and spam, and to an extent, security and stability). Its market share has always been built on the availability of many more applications than for other operating systems, and to a lesser degree on transparency, developer tools, and a pretty-looking interface (the "butterfly"). So, having one kernel with seven distributions is fine. Just think of it as SCO swallowing up Ubuntu, Red Hat, Slackware, Debian, etc. or OpenBSD, NetBSD, and FreeBSD.
It doesn't necessarily bode well, though, given Microsoft's penchant for factoring a lot of nonessential features into the kernel. When a service pack (such as Win2K SP1) is 100+Mb, your customers are conditioned to expect bloatware. Everything I've read about Longhorn tells me there are slick features, but nothing I've read indicates that this trend shows any sign of abating.
... Chinese president Hu Jintao threw a chair at a PRC ambassador who announced his decision to emigrate to the United States, shouting that "America has screwed us" and vowing to "*#&$(&@ kill America".
MS has been trying to build bridges to China for more than seven years: first in founding Microsoft Research China (now MSR Asia), a pure research facility that eventually became an R&D wing. Kai-Fu Lee, the former exec being sued for going to Google, founded MSR China in 1998 and came back to Redmond in 2000.
Several years back Ballmer shook hands on a $100M outsourcing plsu $20M investment deal that senior management found it hard to live up to, and so they amended it to $55M in jobs and $60M in investments in Chinese IT.
If you look at the Seattle Times article, there are links to the court filings, including KFL's deposition, where he indicates that a major part of his work on MS's international business since coming back to the states, has been keeping MS from making blunders. These include making ill-advised promises that it hasn't been able to live up to, and (if it had) would have meant outsourcing at a rate that would strip American jobs, despite its assurances.
Mirosoft's China strategy is starting to fall apart as its hiring has slowed stateside and it becomes evident that it's trying to have its cake and eat it too.
There's a Chinese saying about learning: it is like a leaf in the stream; if it does not move forward, it will be swept back. Google's managing to hold on to market share is a nontrivial achievement.
That said, there's a "them that has, gets" mechanism at work here. Just as we will see when Vista's out, user inertia is strong. Baidu may have its work cut out for it.
I shouldn't be surprised any more at trash talking after the CEO of the Dance and "Developers, Developers, Developers", but that's all it is: another boorish remark.
Ballmer seems to be one of those CEOs who equates being overwrought with being enthusiastic.
Discouraging employees from leaving is just one side effect that Microsoft is trying to achieve using this case as a precedent-setter.
By making it less worth while for companies such as Google to hire away its best talent, Microsoft is also decreasing the outward osmotic gradient.
As the founder of MSR Asia, Kai-Fu Lee brings unique relationship with Chinese researchers with him, but your typical R&D tech staffer will have a harder time "being worth it" if they have to cross the iron curtain, so to speak.
Strange attractors, n'est-ce pas?
(Sorry, had to say that. What do I owe the pun fund?