Where to start with this? Ignoring the inaccurate and unhelpful analogy with blacksmiths and bookkeepers (does the author really believe that nowadays we don't have people whose job it is to work with metal?), the thesis seems to be that, over time, coding is becoming more and more removed from the basic instruction set of the machine. This is hardly a newsworthy observation. No, I haven't looked at Knuth in a while, nor have I had the need for an assembler lately. Yes, over time, we are using increasingly higher level languages, but I don't think that this anything other than blindingly obvious. What next for Slashdot? "Hey look, this interweb thingy seems to be getting quite popular"?
The additional thesis seems to be that, in some vague and unspecified way, AI will take over and do the coding for us. As an AI researcher myself, this is, given the state of the art (by which I mean not what AI can actually do now but what the current philosophical limits of AI are), a complete red herring. Moore's Law is no answer to the Chinese Room - a challenge that the AI community has not yet satisfactorily answered.
Is it me or does this report seem rather unscientific? I quote: "we looked at data from a recent study of 44 users attempting 620 tasks." But no mention of the conditions under which these tasks were set. It's obvious that variations in the experimental conditions will produce variations in the results. For example, someone trying to find a product on a particular website may be inclined to give up after 3 clicks if they know they can just click over to the Walmart site to look. On the other hand, if you say to someone "here's a task, achieve it using this website" it's likely that they will persevere a bit more.
The cynic in me suggests that the main purpose of the article is to publicise their roadshow. But then, the report does have graphs. Who am I to argue with graphs?
This Register article discusses one reason why outsourcing may be so much cheaper (although paying peanuts is probably a much bigger influence.) Seems to me that one way to address this problem may be to charge more for software to companies that outsource to countries where stealing software is endemic.
Slashdot Mirror
Where to start with this? Ignoring the inaccurate and unhelpful analogy with blacksmiths and bookkeepers (does the author really believe that nowadays we don't have people whose job it is to work with metal?), the thesis seems to be that, over time, coding is becoming more and more removed from the basic instruction set of the machine. This is hardly a newsworthy observation. No, I haven't looked at Knuth in a while, nor have I had the need for an assembler lately. Yes, over time, we are using increasingly higher level languages, but I don't think that this anything other than blindingly obvious. What next for Slashdot? "Hey look, this interweb thingy seems to be getting quite popular"? The additional thesis seems to be that, in some vague and unspecified way, AI will take over and do the coding for us. As an AI researcher myself, this is, given the state of the art (by which I mean not what AI can actually do now but what the current philosophical limits of AI are), a complete red herring. Moore's Law is no answer to the Chinese Room - a challenge that the AI community has not yet satisfactorily answered.
Is it me or does this report seem rather unscientific? I quote: "we looked at data from a recent study of 44 users attempting 620 tasks." But no mention of the conditions under which these tasks were set. It's obvious that variations in the experimental conditions will produce variations in the results. For example, someone trying to find a product on a particular website may be inclined to give up after 3 clicks if they know they can just click over to the Walmart site to look. On the other hand, if you say to someone "here's a task, achieve it using this website" it's likely that they will persevere a bit more. The cynic in me suggests that the main purpose of the article is to publicise their roadshow. But then, the report does have graphs. Who am I to argue with graphs?
This Register article discusses one reason why outsourcing may be so much cheaper (although paying peanuts is probably a much bigger influence.) Seems to me that one way to address this problem may be to charge more for software to companies that outsource to countries where stealing software is endemic.