I am sorry, but this is analogous to saying that frequent window-smashing by neighborhood thugs protects your local bank from bank-robbers. Industrial espionage exists because there is a profit to be made from stealing comptetitive secrets or data; that wouldnot change, no matter how secure the systems would be. There would always be black-hats around that would get paid enough dough to break in.
I think the real gain is in the rates of white-hat and black-hat consultants alike:-) --yeah, I am jealous...
Agreed. I also buy the analogy the Rutan brothers and the X-Prize folks are using for their efforts to the time of the Wright brothers: even while heavier-than-air-aircraft would have been a huge military deal just before WWI, most of the true research and development happened by individuals and private companies --and that way of thinking and excelling continued straight into WWII.
In those days NACA (NASA's pre-cursor) did not compete with aviation companies, like NASA does today. NACA would conduct extensive (and expensive) research (into airfoils for example) and release massive amounts of data into the public domain. Hell, the NACA airfoil series are still very, very useful today.
All that changed post-WWII and into the Cold War when aerospace superiority begame a Huge Deal and massive amounts of government spending into defense products created/re-inforced the military-industrial complex that has taken over aerospace in the last few decades. Today, NASA is a cartel of the big US aerospace companies to live off huge government contracts while achieving very little and avoiding competition and raising cost-of-entry into exorbitant levels...
Just one thing I'd like to add to/. so that it shows up in future Google searches: I purchased some cheap IKEA-branded CFLs for my living room. I loved them color-wise and consumption-wise but they created way too much IR interference (infrared) with my TV remote control. The TV is a Sony Wega, so I doubt it's the TV set that's at fault.
My remote wouldn't work, or only work intermittently and it drove me nuts until I realized that this was only happening at night, at which point I started suspecting the CFLs and hit Google/Usenet... my advice: buy brand-name CFLs and if you get stuck with some noisy ones, take care to position the core of the bulb away from any IR receiver --and if you have a multi-bulb lamp like I do, replace the bulb facing your IR receiver with an incadescent or better-quality CFL.
Well, if you're excited about user-rankings and feedback, check out the newsbot in my.sig. It focuses on user interaction with the code/algorithm to build not just page rankings but also relationships --between pages, and between users. Try it out, I am guessing you'll like it...
I hope one day browser coders remember that the proper term for a browser is "user agent". A true agent should be more than a plain old HTTP downloader and renderer, and there are things that we could right now with existing technology.
E.g.: Firebird has added a "view by most visited" feature to the history toolbar. That's cool but what about the biggest waste of a user's time? bookmarks! Why can't modern browsers automatically organize bookmarks? I'll give you one clue: Bayesian filters. Thunderbird has them, and bookmarks are corpi of a sort, as long as you are indexing visited pages for keywords (which of course you can do as you have a full HTML/DOM parser!).
I wish I was a better C++ coder; I have a lot of context and collab filtering working in the newsbot in my sig (all in Python), but users will trust a centralized server so far; a client-side user agent can take these concepts much further...
The other side
on
RFID Explained
·
· Score: 2, Interesting
I am on the other side of this argument: RFIDs are actually good for the consumer, and there is little financial incentive for retailers to do anyting too big brotherly with RFID data; here's my older/. post on the matter.
However, I've had yet another thought recently, one that I haven't heard in any RFID discussion; I am currently in Hong Kong, home of the wondrous Octopus Card an RFID-based smart debit card. Octopus is used for every transit system in the HK metro area, and is increasingly used by retailers to pay for small transactions. Now, actual use of the Octopus rocks: you don't have to take it out of your wallet/bag/briefcase, just swap the whole thing over the reader; you can get an Octopus chip implanted in things other than a card, e.g. the back cover of a Nokia phone, etc.
But one other feature is very cool: an Octopus is anonymous. Anonymous as in cash: you can buy an Octopus and charge it with cash and it does not get traced back to you. There's the potential of RFIDs to actually enhance your privacy by reducing the overhead of certain transactions, and that's pretty big in my book.
I guess it's kind of the same thing as GSM SIM cards: yes they can be used to trace you --both phone-record-wise and location-wise via E911 services-- but you can also go to a shop and pay cash for a cell and a pre-paid SIM and you're online anonymously. There are two sides to every coin...
I agree; Microsoft cares. Here's my take: Microsoft is aware that every major computer (or any technology) platform has been replaced ultimately by a lower-tier, commodity replacement. Big iron by mainframes, then minicomputers, then workstations and PCs. Their entire future and fortune depends on their control of one commodity platform, the Wintel PC. However, there is a huge emergent platform out there, and that is mobile phones and devices that talk to them, not the PC.
I travel globally, mostly Europe and Asia; although PCs are common outside the US, mobiles are ubiquitious. Fifteen year olds SMS while waiting for a bus in Hong Kong and Paris and London, while they may just occasionally double-click on an IE icon on a desktop at home or a net cafe.
Most of the/. crowd extrapolates from the US life-style which is car- and PC-centric while most of the world is public-transport- and mobile-centric; mobiles are much more useful and much more attractive to an average user, because they have much higher network value: think of it like this: PCs became a killer app when they started talking to each other, either via common standard or common networks. Power law of network value and all that... well, how can you compete for network nodes with cells?
Microsoft is feeling the hit of an emerging, cost-competitive commodity platform and they want to hedge their bets. However, history is against them; Nokia makes much better phone UIs and they are riding GSM whereas MS has to deal with the fragmented and underdeveloped US cell market...
Sounds very interesting... are you by any chance using IPython as the shell? if not, it might worth a look, it basically makes the Python CLI a super-shell, so it sounds like you could insert your distributed code under the IPython layer and get a distributed super-shell:-)
I think it boils down to one reason and one reason only: calling party pays. The rest of the world has it, the US does not. In the US, if you get called on your cell, you pay the extra part of the bill (the diff between local and cell rates, in essence). This goes back to the inane decision of having cell phones share area codes with fixed-line (why? the US had call-forwarding for a long time, which is essentially the equivalent of what's in place today... never figured that one out) and the fact that most US consumers have flat-rate, all-you-can-talk plans for local calls, unlike the rest of the world.
So, everything evens out; the US has much worse mobile usage patterns, but they do have unlimited local calls, dirt cheap long-distance and also pretty cheap mobile calling plans. What's missing is the utter ubiquity of cell phones that exists in europe (where back home 12 buys you a GSM SIM card in the corner kiosk w/ prepaid calling time...) OTOH europeans are still getting ripped off from getting charged for fixed line service in the same way as in the pre-digital switching times. Which one's worse?
I'll bite: hard-drive space costs $1/GB if you buy cheap, non-redundant drives that you will manage and administer yourself --i.e. a PC. Enterprise data warehouses cost much, much more than that, and that's just storage, disk array controllers and UPS systems... never mind IT staff and most importantly store-HQ infrastructure to send all that data HQ-side. And then you're forgetting the little matter of big iron to actually *crunch* these numbers into meaningful results, which will multiply costs a few times...
Now, having said all that, big retailers do that now anyway, for product "classes". Say that you're right and they spent a few hundred million in expanding their capabilities by several orders of magnitude. They can track each one of their customers and each purchase they have ever made, at an enormous cost that will only decrease their profit margins. What makes you think that they will then turn around and sell these data to their competitors? even if they make a profit on the data (acquisition and management costs vs. sale), they lose an enormous competitive advantage which will impact them directly. In fact, there would be shareholder revolt.
The internet fridge stuff is show-floor bait, nothing more. If there was any value in replenishing customer fridges, groceries could do that *now* --if you use a "club" or "loyalty" card they can find out both what you have bought and your usual rates of consumption and they could warn you when you're close to running low (say, via SMS). So why don't they do it? no profit: if you run out of a necessity --say milk-- you're gonna go and buy it anyway, and the profit margins on milk are non-existent, so why spend the extra dough for practically zero return? OTOH, if it's not a necessity --say egg-nog-- they cannot tell that you want to buy it anyway and they would probably just annoy you... and besides, didn't Webvan offer something like that:-) ?
Well, every retailer has access to your CC information if you pay with it. And they can track your buying habits with one. And they can share that information with other retailers. So the only thing missing is the "find out who you are when you walk in" part. OK, that's not currently possible (well, there's that face recognition tech...) So what?
Why does a store need to know who you are when you walk in? what possible profit motive is there? Targeted advertising? You mean to say that there retailers will pay money (big money; think of the IT infrastructure) to advertise their wares to a customer that has *already walked in the door*? Maybe they can profile you so that they can mail you special coupons at your home... that's a novel idea indeed... I wonder where all them catalogs came from...
The other paranoid fantasy implicitly mentioned here is that retailers will magically forget all about competition and start sharing customer profiles and sales information --which is a potential gold mine for competitors trying to squeeze into new territory or market segments. That's just not going to happen --one big reason ASPs failed in the '90s, BTW... the tech is already here, there's just no reason to go that way...
As I mentioned elsewhere, the true privacy concerns are with governments attempting to consolidate private data warehouses for their own purposes. The commercial value of these data alone guarantees a corrupt and badly-overseen system...
1. Right now, big global-class (Fortune Global 500) retailers have enterprise systems that rival those on the Supercomputer Top 100. A big retail chain, say a 400-500 store one generates easily hundreds of GBs of data per day. Now, that may not sound like much compared to some scientific number-crunching, but keep in mind these data have to be stored, audit-trailed and managed. And that's just "classes", not instances. Tracking instances will blow things up by orders of magnitude. Most importantly, it will require a "thick" *store* infrastructure to keep track of the massive data flow back to HQ. This is the antithesis of recent industry trends (can you see your neighborhood grocery store having a server room and IT staff on-hand?).
Second of all, perishables or non, most retail items have very low manufacturing (physical) cost, compared to their management cost, and there are two different forces at play: a) perishable items like milk are already sold at or below cost (where ever that is still legal). Adding existing infrastructure to track such items will only decrease profit margins, which will not happen. It's far simpler/cheaper to assume some loss and try to slim down your costs otherwise (via better inventory mgt). b) non-perishable items like the shirts you mention are already at extraordinary profit margins (30% or so for clothing to 70-80% for things like perfumes, etc). Yet retailers still struggle; why? because the individual profit margin for an item is less significant than the aggregate; tracking (or losing) one shirt or one bottle of perfume is not very important for a department store; what is important is making sure that most shirts or perfumes are sold at a profitable margin. RFIDs are *not* going to change any of these equations, except the inventory management one, which will benefit consumers *directly* (via lower prices).
2. Stores that care about shrinkage use RFIDs already --what do you think those inserts in books are, or the tags in clothing? Stores that care less about shrinkage or cannot implement it with current technology (because of cost) will benefit from RFIDs --grocery store self-checkouts come to mind. However, this will mean de-activating the RFIDs on checkout; tracking specific instances (such as flagging them as 'sold') will increase store-based IT (see pt 1).
3. True, it will probably be illegal in some jurisdictions, but that's not certain. Competitive shopping is already happening and is legal. As to the problems with range, I agree, but you can still extract good statistical results from small samples. Further, individually-coded RFID tags are a dead-end, just like proprietary UPC was (if not more so).
One last point: I do retail data-mining; there's little of all the "nightmare" scenarios discussed on this thread that cannot be done already with existing technology. There is a good reason why it doesn't happen: there is no benefit to the companies and there will be no benefit until IT costs come way down. The real nightmares will not come from better tracking technology; they will come from consolidation of data warehouses by governments or companies and from the lack of legal precautions against such consolidation and exploitation. That's what should worry y'all, not a new bar-code...
Sorry, I am not trying to mis-inform. To your points:
1. This happens already in a smaller scale. It's called competitive shopping and there's nothing illegal about it. Now, querying individual RFIDs outside a store entrance will be most likely illegal in some jurisdictions (EU for one), but not others. If RFIDs are still live post-checkout, it will happen, legally or not.
2. How? An RFID tracker could be very small. And you would not need to scan everybody. Just one car querying a particular parking lot row could extract enough statistically significant information in an afternoon. Two or three such cars could determine your sales flow exactly. And that's all you need, if e.g. you're planning to open a competitive store a block down the street.
3. Which is inevitable. The article mentions EPC; whether it's EPC or some other standard, a new, super-UPC will have to come about as there are way too many companies involved in supply chains. There will *not* be proprietary RFID codes...
Same thing can happen now with UPC barcodes. The reason that most retailers don't do it (and they don't) are the infrastructure costs I mentioned earlier, and the simple fact that there is no inherent value in individual customer behavior. None. Only aggregates are interesting to a retailer (such-and-such zip code is buying more Gap than Banana Republic, etc). And they do that right now, without RFIDs.
You're assuming that it's to the interest of the retailer to leave the RFID active after you leave the store. That's just not the case. Let me clear up some RFID myths (and I am a retail systems consultant, BTW, this is my bread-and-butter):
RFIDs allow stores to track instances of products; i.e. a specific can of Coke (serial # so-and-so, part of case such-and-such that was shipped by Joe Q. Supplier) instead of the current UPC "class identifier", i.e. "this is just a can of Coke". Now, people read this and they just jump ahead and assume that a retailer, however big, is ready to pay millions and millions of dollars for infrastructure in their warehouses, distribution centers and ultimately stores, to track trillions of product instances. Wake-up call: no they're not, and no they will never be. At most they might track some informative 'class-attributes' to borrow an OOP term: things like supplier or lot number. The whole RFID-allows-instance-tracking is only useful for items whose management cost is much higher than its physical cost: think auto or airplane parts, drugs, etc; not Gap shirts. This will not go away for decades, even allowing for Moore's law to keep going and for lower associated IT costs following that same trend.
One added side-benefit of RFIDs is controling shrinkage, i.e. shoplifting. For that to actually work, and assuming instance-tracking is out of the question (see above) paid-off items have to be de-activated by the store itself upon checkout. Your questions are thus moot.
Besides shrinkage, lemme tell you a little secret: retailers are very, very, very competitive. Suppose they don't de-activate paid-off RFIDs and let the chips keep on responding to query signals. You know what will happen within a week of that being rolled out by someone like Wal-Mart? Target will set up a truck in a Wal-Mart parking lot and start measuring their sales. Do you think Wal-Mart will let that happen? And AFAIK there's no way to stop that from happening unless RFIDs come with built-in Public/Private Key infrastructure, which will only increase their managerial costs (a lot; just think of all the suppliers Wal-Mart and Target share!) re-inforcing my first point.
Now, instead of paranoid worries, I hope people start focusing on the promise of RFIDs: instant checkouts, instant inventories, instant customer feedback to the retailer (meaning better product choices by the stores) and much better inventory management (meaning lower prices!). Never mind trackable warranties, potential theft prevention/insurance, etc, etc, etc...
You got it upside down; this is about building the *index* faster, not serving pages. Google AFAIK updates their index at the first of its month, so we can only assume it takes =30 days to build.
In my view, personal recommendations from a search engine are mostly valuable for topical content --i.e. news items. However, the optimizations from these papers don't sound to me like they can do much for this case --news items pop up in a news site, and re-indexing the news source itself (say, the front page of CNN) won't tell you much about a particular CNN story.
At any rate, personal news recommendations is a favorite topic of mine: this is why I built Memigo: to create a bot that finds news I am more likely to like. Memigo learns from its users collectively and each user individually --and BTW, it predates Google News by a good 6 months, IIRC. The memigo codebase (all in Python) is now up to the point where it can start learning what content each user likes... If you like Google News you'll love Memigo.
And BTW, I did RTFA when it was on memigo's front page this morning:-)...
If you like Google news you may like the newsbot in my.sig. Pretty much everything GN does (albeit some features are not as refined as GN's) with many, many more features added on.
I completely agreed with Graham's article. I know his tool of choice is LISP, but as I was reading the text, my epiphany with Python kept coming to my mind. I used to hack PHP too, but I don't any more. PHP is better than the alternatives in the Web-building niche, but it's not either as general-purpose or as scalable (in terms of project size) as Python --my toy newsbot in my.sig is 100% pure Python, all 20kLOC of it... I could go on and on about Python, but that will hardly be as persuasive as going over to python.org, reading the tutorial and writing some simple short project in py. You will be converted in no time...
You have to remember that the US government has subsidized infrastructure in the past. Namely, the Interstate system and the US airline industry (the latter both directly and indirectly via funding R&D thru defense spending). Both initiatives were rationalized using defense reasons and purposes (troop movement, leadership in aerospace). However, both have resulted in better infrastructure, lower costs, and generally improved social conditions.
My point is, subsidizing isn't inherently wrong. It should be considered as investment writ large: society as a whole expects a return on this investment and they're willing to pay with taxes up front. Don't mix smart and forward-looking subsidizing with dumb subsidizing (as in say, the EU countries are doing by protecting their state telcos or state power companies or state airlines --although that's going away as well).
The American mentality against any and all subsidizing is hurting the US economy more than helping it lately, as you do need large infrastructure for some things: witness the 2G cell licenses (GSM has won anyway, but now the US is 2-3 yrs behind everyone else), the sorry state of the railway system or lack of city public transport (as in subways, light trams, or anything else remotely environmentally- and people-friendly).
There are easy ways around this: you can age songs (guaranteeing that new songs get priority) or you can use web-of-trust ratings to see what high-trusted peers recommend currently.
BTW, I know this as my newsbot does the same sortof filtering for news articles, and although there is some feedback reinforcement for recent news, it does work suprisingly well.
Re:Think of the back-end info this gives them
on
Google buys Pyra Labs
·
· Score: 2, Interesting
I completely agree. I've been building up my newsbot to do just that: sniff out interesting new articles/web memes based on weblogs and inter-linking. I don't know if I could ever compete with Google+Blogger though, as you need a lot of users for the results to get more and more interesting --and noone has more users than Blogger. I think this buyout fits the model.
Slashdot Mirror
I am sorry, but this is analogous to saying that frequent window-smashing by neighborhood thugs protects your local bank from bank-robbers. Industrial espionage exists because there is a profit to be made from stealing comptetitive secrets or data; that wouldnot change, no matter how secure the systems would be. There would always be black-hats around that would get paid enough dough to break in.
:-) --yeah, I am jealous...
I think the real gain is in the rates of white-hat and black-hat consultants alike
Agreed. I also buy the analogy the Rutan brothers and the X-Prize folks are using for their efforts to the time of the Wright brothers: even while heavier-than-air-aircraft would have been a huge military deal just before WWI, most of the true research and development happened by individuals and private companies --and that way of thinking and excelling continued straight into WWII.
In those days NACA (NASA's pre-cursor) did not compete with aviation companies, like NASA does today. NACA would conduct extensive (and expensive) research (into airfoils for example) and release massive amounts of data into the public domain. Hell, the NACA airfoil series are still very, very useful today.
All that changed post-WWII and into the Cold War when aerospace superiority begame a Huge Deal and massive amounts of government spending into defense products created/re-inforced the military-industrial complex that has taken over aerospace in the last few decades. Today, NASA is a cartel of the big US aerospace companies to live off huge government contracts while achieving very little and avoiding competition and raising cost-of-entry into exorbitant levels...
Just one thing I'd like to add to /. so that it shows up in future Google searches: I purchased some cheap IKEA-branded CFLs for my living room. I loved them color-wise and consumption-wise but they created way too much IR interference (infrared) with my TV remote control. The TV is a Sony Wega, so I doubt it's the TV set that's at fault.
My remote wouldn't work, or only work intermittently and it drove me nuts until I realized that this was only happening at night, at which point I started suspecting the CFLs and hit Google/Usenet... my advice: buy brand-name CFLs and if you get stuck with some noisy ones, take care to position the core of the bulb away from any IR receiver --and if you have a multi-bulb lamp like I do, replace the bulb facing your IR receiver with an incadescent or better-quality CFL.
Well, if you're excited about user-rankings and feedback, check out the newsbot in my .sig. It focuses on user interaction with the code/algorithm to build not just page rankings but also relationships --between pages, and between users. Try it out, I am guessing you'll like it...
Google for IPython, it does some of what you want and a lot more.
I hope one day browser coders remember that the proper term for a browser is "user agent". A true agent should be more than a plain old HTTP downloader and renderer, and there are things that we could right now with existing technology.
E.g.: Firebird has added a "view by most visited" feature to the history toolbar. That's cool but what about the biggest waste of a user's time? bookmarks! Why can't modern browsers automatically organize bookmarks? I'll give you one clue: Bayesian filters. Thunderbird has them, and bookmarks are corpi of a sort, as long as you are indexing visited pages for keywords (which of course you can do as you have a full HTML/DOM parser!).
I wish I was a better C++ coder; I have a lot of context and collab filtering working in the newsbot in my sig (all in Python), but users will trust a centralized server so far; a client-side user agent can take these concepts much further...
I am on the other side of this argument: RFIDs are actually good for the consumer, and there is little financial incentive for retailers to do anyting too big brotherly with RFID data; here's my older /. post on the matter.
However, I've had yet another thought recently, one that I haven't heard in any RFID discussion; I am currently in Hong Kong, home of the wondrous Octopus Card an RFID-based smart debit card. Octopus is used for every transit system in the HK metro area, and is increasingly used by retailers to pay for small transactions. Now, actual use of the Octopus rocks: you don't have to take it out of your wallet/bag/briefcase, just swap the whole thing over the reader; you can get an Octopus chip implanted in things other than a card, e.g. the back cover of a Nokia phone, etc.
But one other feature is very cool: an Octopus is anonymous. Anonymous as in cash: you can buy an Octopus and charge it with cash and it does not get traced back to you. There's the potential of RFIDs to actually enhance your privacy by reducing the overhead of certain transactions, and that's pretty big in my book.
I guess it's kind of the same thing as GSM SIM cards: yes they can be used to trace you --both phone-record-wise and location-wise via E911 services-- but you can also go to a shop and pay cash for a cell and a pre-paid SIM and you're online anonymously. There are two sides to every coin...
I agree; Microsoft cares. Here's my take: Microsoft is aware that every major computer (or any technology) platform has been replaced ultimately by a lower-tier, commodity replacement. Big iron by mainframes, then minicomputers, then workstations and PCs. Their entire future and fortune depends on their control of one commodity platform, the Wintel PC. However, there is a huge emergent platform out there, and that is mobile phones and devices that talk to them, not the PC.
/. crowd extrapolates from the US life-style which is car- and PC-centric while most of the world is public-transport- and mobile-centric; mobiles are much more useful and much more attractive to an average user, because they have much higher network value: think of it like this: PCs became a killer app when they started talking to each other, either via common standard or common networks. Power law of network value and all that... well, how can you compete for network nodes with cells?
I travel globally, mostly Europe and Asia; although PCs are common outside the US, mobiles are ubiquitious. Fifteen year olds SMS while waiting for a bus in Hong Kong and Paris and London, while they may just occasionally double-click on an IE icon on a desktop at home or a net cafe.
Most of the
Microsoft is feeling the hit of an emerging, cost-competitive commodity platform and they want to hedge their bets. However, history is against them; Nokia makes much better phone UIs and they are riding GSM whereas MS has to deal with the fragmented and underdeveloped US cell market...
Sounds very interesting... are you by any chance using IPython as the shell? if not, it might worth a look, it basically makes the Python CLI a super-shell, so it sounds like you could insert your distributed code under the IPython layer and get a distributed super-shell :-)
I think it boils down to one reason and one reason only: calling party pays. The rest of the world has it, the US does not. In the US, if you get called on your cell, you pay the extra part of the bill (the diff between local and cell rates, in essence). This goes back to the inane decision of having cell phones share area codes with fixed-line (why? the US had call-forwarding for a long time, which is essentially the equivalent of what's in place today... never figured that one out) and the fact that most US consumers have flat-rate, all-you-can-talk plans for local calls, unlike the rest of the world.
So, everything evens out; the US has much worse mobile usage patterns, but they do have unlimited local calls, dirt cheap long-distance and also pretty cheap mobile calling plans. What's missing is the utter ubiquity of cell phones that exists in europe (where back home 12 buys you a GSM SIM card in the corner kiosk w/ prepaid calling time...) OTOH europeans are still getting ripped off from getting charged for fixed line service in the same way as in the pre-digital switching times. Which one's worse?
I'll bite: hard-drive space costs $1/GB if you buy cheap, non-redundant drives that you will manage and administer yourself --i.e. a PC. Enterprise data warehouses cost much, much more than that, and that's just storage, disk array controllers and UPS systems... never mind IT staff and most importantly store-HQ infrastructure to send all that data HQ-side. And then you're forgetting the little matter of big iron to actually *crunch* these numbers into meaningful results, which will multiply costs a few times...
:-) ?
Now, having said all that, big retailers do that now anyway, for product "classes". Say that you're right and they spent a few hundred million in expanding their capabilities by several orders of magnitude. They can track each one of their customers and each purchase they have ever made, at an enormous cost that will only decrease their profit margins. What makes you think that they will then turn around and sell these data to their competitors? even if they make a profit on the data (acquisition and management costs vs. sale), they lose an enormous competitive advantage which will impact them directly. In fact, there would be shareholder revolt.
The internet fridge stuff is show-floor bait, nothing more. If there was any value in replenishing customer fridges, groceries could do that *now* --if you use a "club" or "loyalty" card they can find out both what you have bought and your usual rates of consumption and they could warn you when you're close to running low (say, via SMS). So why don't they do it? no profit: if you run out of a necessity --say milk-- you're gonna go and buy it anyway, and the profit margins on milk are non-existent, so why spend the extra dough for practically zero return? OTOH, if it's not a necessity --say egg-nog-- they cannot tell that you want to buy it anyway and they would probably just annoy you... and besides, didn't Webvan offer something like that
Well, every retailer has access to your CC information if you pay with it. And they can track your buying habits with one. And they can share that information with other retailers. So the only thing missing is the "find out who you are when you walk in" part. OK, that's not currently possible (well, there's that face recognition tech...) So what?
Why does a store need to know who you are when you walk in? what possible profit motive is there? Targeted advertising? You mean to say that there retailers will pay money (big money; think of the IT infrastructure) to advertise their wares to a customer that has *already walked in the door*? Maybe they can profile you so that they can mail you special coupons at your home... that's a novel idea indeed... I wonder where all them catalogs came from...
The other paranoid fantasy implicitly mentioned here is that retailers will magically forget all about competition and start sharing customer profiles and sales information --which is a potential gold mine for competitors trying to squeeze into new territory or market segments. That's just not going to happen --one big reason ASPs failed in the '90s, BTW... the tech is already here, there's just no reason to go that way...
As I mentioned elsewhere, the true privacy concerns are with governments attempting to consolidate private data warehouses for their own purposes. The commercial value of these data alone guarantees a corrupt and badly-overseen system...
OK, counter-point:
1. Right now, big global-class (Fortune Global 500) retailers have enterprise systems that rival those on the Supercomputer Top 100. A big retail chain, say a 400-500 store one generates easily hundreds of GBs of data per day. Now, that may not sound like much compared to some scientific number-crunching, but keep in mind these data have to be stored, audit-trailed and managed. And that's just "classes", not instances. Tracking instances will blow things up by orders of magnitude. Most importantly, it will require a "thick" *store* infrastructure to keep track of the massive data flow back to HQ. This is the antithesis of recent industry trends (can you see your neighborhood grocery store having a server room and IT staff on-hand?).
Second of all, perishables or non, most retail items have very low manufacturing (physical) cost, compared to their management cost, and there are two different forces at play: a) perishable items like milk are already sold at or below cost (where ever that is still legal). Adding existing infrastructure to track such items will only decrease profit margins, which will not happen. It's far simpler/cheaper to assume some loss and try to slim down your costs otherwise (via better inventory mgt). b) non-perishable items like the shirts you mention are already at extraordinary profit margins (30% or so for clothing to 70-80% for things like perfumes, etc). Yet retailers still struggle; why? because the individual profit margin for an item is less significant than the aggregate; tracking (or losing) one shirt or one bottle of perfume is not very important for a department store; what is important is making sure that most shirts or perfumes are sold at a profitable margin. RFIDs are *not* going to change any of these equations, except the inventory management one, which will benefit consumers *directly* (via lower prices).
2. Stores that care about shrinkage use RFIDs already --what do you think those inserts in books are, or the tags in clothing? Stores that care less about shrinkage or cannot implement it with current technology (because of cost) will benefit from RFIDs --grocery store self-checkouts come to mind. However, this will mean de-activating the RFIDs on checkout; tracking specific instances (such as flagging them as 'sold') will increase store-based IT (see pt 1).
3. True, it will probably be illegal in some jurisdictions, but that's not certain. Competitive shopping is already happening and is legal. As to the problems with range, I agree, but you can still extract good statistical results from small samples. Further, individually-coded RFID tags are a dead-end, just like proprietary UPC was (if not more so).
One last point: I do retail data-mining; there's little of all the "nightmare" scenarios discussed on this thread that cannot be done already with existing technology. There is a good reason why it doesn't happen: there is no benefit to the companies and there will be no benefit until IT costs come way down. The real nightmares will not come from better tracking technology; they will come from consolidation of data warehouses by governments or companies and from the lack of legal precautions against such consolidation and exploitation. That's what should worry y'all, not a new bar-code...
Sorry, I am not trying to mis-inform. To your points:
1. This happens already in a smaller scale. It's called competitive shopping and there's nothing illegal about it. Now, querying individual RFIDs outside a store entrance will be most likely illegal in some jurisdictions (EU for one), but not others. If RFIDs are still live post-checkout, it will happen, legally or not.
2. How? An RFID tracker could be very small. And you would not need to scan everybody. Just one car querying a particular parking lot row could extract enough statistically significant information in an afternoon. Two or three such cars could determine your sales flow exactly. And that's all you need, if e.g. you're planning to open a competitive store a block down the street.
3. Which is inevitable. The article mentions EPC; whether it's EPC or some other standard, a new, super-UPC will have to come about as there are way too many companies involved in supply chains. There will *not* be proprietary RFID codes...
Same thing can happen now with UPC barcodes. The reason that most retailers don't do it (and they don't) are the infrastructure costs I mentioned earlier, and the simple fact that there is no inherent value in individual customer behavior. None. Only aggregates are interesting to a retailer (such-and-such zip code is buying more Gap than Banana Republic, etc). And they do that right now, without RFIDs.
Now, instead of paranoid worries, I hope people start focusing on the promise of RFIDs: instant checkouts, instant inventories, instant customer feedback to the retailer (meaning better product choices by the stores) and much better inventory management (meaning lower prices!). Never mind trackable warranties, potential theft prevention/insurance, etc, etc, etc...
You got it upside down; this is about building the *index* faster, not serving pages. Google AFAIK updates their index at the first of its month, so we can only assume it takes =30 days to build.
In my view, personal recommendations from a search engine are mostly valuable for topical content --i.e. news items. However, the optimizations from these papers don't sound to me like they can do much for this case --news items pop up in a news site, and re-indexing the news source itself (say, the front page of CNN) won't tell you much about a particular CNN story.
:-)...
At any rate, personal news recommendations is a favorite topic of mine: this is why I built Memigo: to create a bot that finds news I am more likely to like. Memigo learns from its users collectively and each user individually --and BTW, it predates Google News by a good 6 months, IIRC. The memigo codebase (all in Python) is now up to the point where it can start learning what content each user likes... If you like Google News you'll love Memigo.
And BTW, I did RTFA when it was on memigo's front page this morning
If you like Google news you may like the newsbot in my .sig. Pretty much everything GN does (albeit some features are not as refined as GN's) with many, many more features added on.
I completely agreed with Graham's article. I know his tool of choice is LISP, but as I was reading the text, my epiphany with Python kept coming to my mind. I used to hack PHP too, but I don't any more. PHP is better than the alternatives in the Web-building niche, but it's not either as general-purpose or as scalable (in terms of project size) as Python --my toy newsbot in my .sig is 100% pure Python, all 20kLOC of it... I could go on and on about Python, but that will hardly be as persuasive as going over to python.org, reading the tutorial and writing some simple short project in py. You will be converted in no time...
You have to remember that the US government has subsidized infrastructure in the past. Namely, the Interstate system and the US airline industry (the latter both directly and indirectly via funding R&D thru defense spending). Both initiatives were rationalized using defense reasons and purposes (troop movement, leadership in aerospace). However, both have resulted in better infrastructure, lower costs, and generally improved social conditions.
My point is, subsidizing isn't inherently wrong. It should be considered as investment writ large: society as a whole expects a return on this investment and they're willing to pay with taxes up front. Don't mix smart and forward-looking subsidizing with dumb subsidizing (as in say, the EU countries are doing by protecting their state telcos or state power companies or state airlines --although that's going away as well).
The American mentality against any and all subsidizing is hurting the US economy more than helping it lately, as you do need large infrastructure for some things: witness the 2G cell licenses (GSM has won anyway, but now the US is 2-3 yrs behind everyone else), the sorry state of the railway system or lack of city public transport (as in subways, light trams, or anything else remotely environmentally- and people-friendly).
Well, AFAIK, 2.3 (or maybe 2.2) has finally taken this issue away. EOL conventions donot matter for source code any more.
There are easy ways around this: you can age songs (guaranteeing that new songs get priority) or you can use web-of-trust ratings to see what high-trusted peers recommend currently.
BTW, I know this as my newsbot does the same sortof filtering for news articles, and although there is some feedback reinforcement for recent news, it does work suprisingly well.
I completely agree. I've been building up my newsbot to do just that: sniff out interesting new articles/web memes based on weblogs and inter-linking. I don't know if I could ever compete with Google+Blogger though, as you need a lot of users for the results to get more and more interesting --and noone has more users than Blogger. I think this buyout fits the model.
!= != opposite