Imagine having a pair of glasses with a camera/microphone at the outside corner of each lens. You wear these everywhere you go, everything you see/hear gets recorded, you drop them in their holder next to your bed every night and they dump their contents to your personal memory backup. I'd start using something like this in a heartbeat, with appropriate protections (encrypted, password based on my biometrics, Fifth Amendment protected).
This kind of thing will be feasible in ten to twenty years if Moore's law continues to hold.
And a few years after that, it may be possible to have something like this, without the glasses - the microphones are implanted in your earlobes, the camera sits inside your eye on your blind spot, and you can't take it off...
I doubt I'm going to convince you, but I feel I ought to try:
For example, you say Science cares about reality. You fail in the next point by not crediting Law with the exact same concern. You say all they're concerned with is a believable story to satisfy the audience (jury). But the exact same thing as true of Science (the audience being their peers).
There is a difference. The scientific audience says "show me how I can do it", and unreplicable phenomena get weeded out. What killed cold fusion was grad students in labs all over the world failing to see neutrons, not the pronouncements of respected professors.
The legal audience says "if those twelve people agree, that's what happened". Shitloads of peer-reviewed studies said breast implants didn't affect women's immune systems, but several juries said otherwise, and Dow Corning went bankrupt.
They have fundamentally different ways of looking at the world, so naturally they interact badly.
Science cares about external consistency. Scientists build models of the world, test them, and throw them away when they are inconsistent with observation.
Law cares about internal consistency. One of the most important considerations is precedent - "we did it this way last time". When the world changes, precedent gets overturned - eventually.
Science cares about reality. The gold standard in science is the published, reproducible procedure.
Law cares about verisimilitude - believable stories. The gold standard in law is getting twelve members of the community to believe your story, and not just any twelve people - if a person has any expertise related to the matter in court, they will be filtered out of the jury pool.
Science is never the last word. Observation can always make you change your model. Newton was the last word for centuries, now he's an approximation to Einstein.
Law is supposed to be final, and it defends its finality fiercely - witness the resistance to checking old decisions with new DNA techmology, whereas in science the first thing you do with a new tool is compare it with your old measurements.
Yeah, yeah, I know, Thomas Kuhn, postmodernism, yadda yadda... the above is the idealized way science works - reality is more complex and slow, but by and large peer review works.
Almost all of the information on Javascript you can find on the web is aimed at web designers who aren't programmers. They pass around little snippets of code that "worked for me", which 99% of the time really means "worked when I tried it on IE". I personally would kill for a document entitled "Javascript for people who know what a first-class function is", because such a document might be written by someone who knows Javascript from MS DOM from ECMA DOM, and might explain it to me.
Umm...not trying to say the EFF is in any way bad, but are there any right reponses that don't involve the EFF immediately?
Yeah, yeah:
If you are named in one of these lawsuits, and you did it, and the damages are not completely out to lunch, go pay for your own damn lawyer and negotiate, rather than sucking up EFF time and money. And stop screwing legitimate copyright holders, OK?
The use of this option will be rather rare, granted...
If you are named in one of these lawsuits, and you didn't do it, call the EFF, now. A few expensive countersuits will keep the RIAA from using this as scare tactics. Extra funding for the EFF from the RIAA would be nice, too.
If you are named in one of these lawsuits, and you did it, but the damages against you are ridiculously high, call the EFF, now. Don't settle out-of-court for your life savings without getting some decent advice first.
If you aren't named in one of these lawsuits, but the idea of an industry group beating up indiscriminantly on thousands of individuals makes you mad, call the EFF, now, and make a donation!
The fact that the (almost) top person at Apple has made this clarification shows how much importance they're putting against these claims.
What's the #1 argument used against Macs in corporate workplaces? "They're not compatible - we don't want to support more than one platform."
What's the best way to show that's wrong? Have a sysadmin type open up his Powerbook and say "Look! It sees the file servers, it opens Word/Excel/PowerPoint files, it sees the printers, and it does all that with no support."
How do you get that kind of support? Make your laptops attractive to geeks. Powerbooks are becoming the Swiss army chainsaw of choice for sysadmin types - they let you run MS apps and Un*x development/diagnostic tools.
How do you keep that kind of support? Pay attention when the geeks question your claims. Treat their objections seriously, and answer them with minimal corporate spin applied.
Apple has actually handled this flap pretty well. Their artificial benchmarks are at least defensible, and the application demo wall-clock times are really impressive - beating the competition by a factor of two is definitely past the point where you should sit up and take notice.
I want to see the chapter on "The FUD of War"...
on
Platform Evangelism
·
· Score: 1
... in which he shows how lying can be a strategic tool. Subheadings should include:
The art of the non sequitur - saying "it's better" when at best it's just different
Words mean what I say they mean - saying "non-standard" when you mean "compatible with the published standard, but not our extensions"
Exaggeration for fun and profit - saying "we ship next quarter" when you mean "next year" "or maybe never"
Of particular interest will be the conclusion in which he shows how this is all ultimately for the customer's benefit...
... in 1987. I worked for Xerox AI Systems at the time and was one of the developers for the Lyric release of LOOPS.
LOOPS had:
A single-inheritance object system with GUI support (class, method, and object browsers).
Editing with structure editors that manipulated the parse tree directly. The structure editor was also used as the inspector in the debugger.
refactoring support in the browsers (select a method and move it to another class, etc).
automated global refactoring based on code analysis - this being 1987, it had a pseudo-natural language interface (EDIT ALL METHODS CALLING FOO AND REFERENCING *BAR*...).
LOOPS is one of the primary predecessors of CLOS (the other being Flavors).
Parse-tree-based editing has been around for awhile - Google for "syntax directed editing". Paradoxically, it works much better in Lisp than in other, more syntax-heavy domains; when your editor insists that everything be syntactically and semantically well-formed all the time, it's best that there be very few, very general syntax rules. This is why it also works in Smalltalk.
The Constitution says that copyright is "for a limited time". DRM is typically forever, so it goes beyond copyright. With that in mind, the law we really need is:
Any work that is only published under DRM that will last beyond the copyright in force at the time of its publication, loses its copyright protection.
If creators feel the need for mechanical protection beyond the terms of copyright, they should not be able to use the legal system in addition to their fair-use-defeating technology.
But, it is possible to go below the minimum. The most egregious Apple example of this recently was the original G3 tower "hockey-puck" mouse, which sacrificed function nearly completely on the altar of form. I have one of these on a hand-me-down computer, and I can't use it without curling my fingers down over the front edge to find the cord - you can't orient it otherwise. ARGH!
The no-button pro mouse, on the other hand, is a work of art that works.
Here's a privacy policy I'd like to hear from a major corporation, just once:
We view our collection of data about our customer base as both an asset and a liability.
It is an asset in that we use it to generate aggregated data about you, and sell that data to people who want to sell you things.
It is also a liability in that we got this data from you with a promise to never release it in detail, allowing others to see your individual habits and preferences.
With that promise in mind, we repeat that we will never sell our data about you in a manner that allows you to be individually identified. If we ever go out of business, our last act as a corporation will be to destroy any indices that would enable that.
Google, for one, started as a Stanford university project,...
The core Google page ranking algorithms, yes. The Google clustering infrastructure, no. According to this, Google didn't scale beyond what would fit into a Stanford dorm room before it went commercial in 1998.
Raw speed will always be useful for problems that are hard to parallelize. Right now those problems (parts of crypto, some quantum physics calculations, etc.) are important scientifically, but away from the money.
Industry will spend R&D money on clustering for storage and reliability, without major government subsidy, because there's a crying need for it. How much government money went into Google/eBay/Amazon?
Government research is supposed to complement industry R&D - to be aimed at fields where the results are still important, but maybe not as profitable. This is why government should not abandon raw speed as a research goal.
The difference between Apple "Inoperable Failure" machines and Dell "Inoperable Failure"machines looks to be very small - the real difference between them is in machines that were "Broken but Still Operable".
Maybe they're seeing people who got their computers working, then discovered they were running XP...
By quite a bit, actually. Look here down around 1989.
Also, on the same history page, in the Mid 80's section, you'll see an entry for an expert system named Pride developed at PARC. Pride helped Xerox design their first line of desktop copiers, and is quite famous within the company.
I worked for (long lamented) Xerox AI systems from 1986-88, and consulted for them off and on through 1994, which is how I know about this.
Park it at L1 or L2. Space is roomy, so building big things is easier. Aiming is easier - you do have to be more accurate and have a better collimated beam, but you only have to track it across a degree or two to cover the whole earth, and you could aim by tracking the whole ring, so you'd need less powerful deflector magnets. You can power it with solar energy. And, the vacuum is free!
There is the little problem of getting there, setting up shop, and building a 1000 km structure, of course...
What Minsky is actually flaming about here is the damage done to the field by the original "AI winter" and the real possibility of a new AI winter starting in a few years.
"AI winter" is the name given to the collapse of strong AI as a business model in the mid 80's - expert systems and symbolic AI in general didn't deliver on their promises, and so the money went away. As a guy who got his doctorate in AI in 1985, I can tell you all about it.;-)
One of the major causes of AI winter was researcher hubris - lots of people hacked up systems that appeared to solve 80 percent of certain complex problems and then said "all that stands between us and a complete solution is money and time". For many of those systems, solving the last 20 percent would have taken 2000 percent of the time, if it could have been done at all. The tragedy of AI winter, though, is that basically all of symbolic AI was abandoned, though some of it is creeping back out into the light with obfuscated syntax (see my.sig below).
What Minsky sees here is a lot of people heading down the same path, but with neural nets and small robots instead of expert systems. The new systems are doing some interesting things relative to the old symbolic AI systems (though they do have the advantage of 20 years of Moore's law to help them). But, will they scale up? Right now, nobody knows. If they don't, the last thing the field needs is another cycle of overpromise/underdeliver/abandon.
Maybe AI is just plain hard, and cracking it will take longer than one or two computer industry business cycles.
Why are desktop computing systems fragile? Because in the markteplace, they are judged on exactly two criteria:
How big is the check I'm writing right now?
How fast is it?
With these as your evaluation function, you are guaranteed to get systems with little redundancy and little or no internal safety checks.
One regrettable example of this is the market for personal finance programs. The feature that sells Quicken is quick-fill - the heuristic automatic data entry that makes entering transactions fast. Never mind that Quicken's register file is fragile - it frequently loses track of balances (requiring the moral equivalent of fsck), and every few years the accumulated unfixed cruft causes a major failure, requiring insane fixes like exporting all your data as QIF files and reimporting it into a new register.
If Quicken were back-ended into a real database, with real transactions, real consistency checks, and real crash recovery, all this would go away. But it would make Quicken slower and require more hardware horespower to run it - the marketplace would punish them for improving their lives.
What the original article is proposing is:
We accept that systems will always suck
Therefore, we should build multi-level suckage damage control into them
Another possibility is:
We accept that there is a tradeoff between system speed and safety
Therefore, we take the speed hit where safety is important
"Maclike" means Apple makes and controls both the hardware and the software, which means:
They fit together better
They can make radical changes which require support from both sides, without going outside the company.
"Microsoft-like" means MS has a monopoly on the software, so they can tell the hardware guys what to make to maximize MS profits. This means:
They fit together better, because MS has to implement fewer options
There will be less radical innovation, because the hardware guys will have to get MS' blessing to sell anything really new to the lucrative corporate market that is the main target of this controlled architecture.
Slashdot Mirror
Imagine having a pair of glasses with a camera/microphone at the outside corner of each lens. You wear these everywhere you go, everything you see/hear gets recorded, you drop them in their holder next to your bed every night and they dump their contents to your personal memory backup. I'd start using something like this in a heartbeat, with appropriate protections (encrypted, password based on my biometrics, Fifth Amendment protected).
This kind of thing will be feasible in ten to twenty years if Moore's law continues to hold.
And a few years after that, it may be possible to have something like this, without the glasses - the microphones are implanted in your earlobes, the camera sits inside your eye on your blind spot, and you can't take it off...
There is a difference. The scientific audience says "show me how I can do it", and unreplicable phenomena get weeded out. What killed cold fusion was grad students in labs all over the world failing to see neutrons, not the pronouncements of respected professors.
The legal audience says "if those twelve people agree, that's what happened". Shitloads of peer-reviewed studies said breast implants didn't affect women's immune systems, but several juries said otherwise, and Dow Corning went bankrupt.
Science cares about external consistency. Scientists build models of the world, test them, and throw them away when they are inconsistent with observation.
Law cares about internal consistency. One of the most important considerations is precedent - "we did it this way last time". When the world changes, precedent gets overturned - eventually.
Science cares about reality. The gold standard in science is the published, reproducible procedure.
Law cares about verisimilitude - believable stories. The gold standard in law is getting twelve members of the community to believe your story, and not just any twelve people - if a person has any expertise related to the matter in court, they will be filtered out of the jury pool.
Science is never the last word. Observation can always make you change your model. Newton was the last word for centuries, now he's an approximation to Einstein.
Law is supposed to be final, and it defends its finality fiercely - witness the resistance to checking old decisions with new DNA techmology, whereas in science the first thing you do with a new tool is compare it with your old measurements.
Yeah, yeah, I know, Thomas Kuhn, postmodernism, yadda yadda... the above is the idealized way science works - reality is more complex and slow, but by and large peer review works.
Almost all of the information on Javascript you can find on the web is aimed at web designers who aren't programmers. They pass around little snippets of code that "worked for me", which 99% of the time really means "worked when I tried it on IE". I personally would kill for a document entitled "Javascript for people who know what a first-class function is", because such a document might be written by someone who knows Javascript from MS DOM from ECMA DOM, and might explain it to me.
While I'm waiting for that document, my favorite Javascript reference is http://www.xs4all.nl/~ppk/js/version5.html.
Yeah, yeah:
If you are named in one of these lawsuits, and you did it, and the damages are not completely out to lunch, go pay for your own damn lawyer and negotiate, rather than sucking up EFF time and money. And stop screwing legitimate copyright holders, OK?
The use of this option will be rather rare, granted...
And I'm an idiot: it's the Electronic Frontier Foundation.
If you are named in one of these lawsuits, and you didn't do it, call the EFF, now. A few expensive countersuits will keep the RIAA from using this as scare tactics. Extra funding for the EFF from the RIAA would be nice, too.
If you are named in one of these lawsuits, and you did it, but the damages against you are ridiculously high, call the EFF, now. Don't settle out-of-court for your life savings without getting some decent advice first.
If you aren't named in one of these lawsuits, but the idea of an industry group beating up indiscriminantly on thousands of individuals makes you mad, call the EFF, now, and make a donation!
That's the Electronic Freedom Foundation, folks...
What's the #1 argument used against Macs in corporate workplaces? "They're not compatible - we don't want to support more than one platform."
What's the best way to show that's wrong? Have a sysadmin type open up his Powerbook and say "Look! It sees the file servers, it opens Word/Excel/PowerPoint files, it sees the printers, and it does all that with no support."
How do you get that kind of support? Make your laptops attractive to geeks. Powerbooks are becoming the Swiss army chainsaw of choice for sysadmin types - they let you run MS apps and Un*x development/diagnostic tools.
How do you keep that kind of support? Pay attention when the geeks question your claims. Treat their objections seriously, and answer them with minimal corporate spin applied.
Apple has actually handled this flap pretty well. Their artificial benchmarks are at least defensible, and the application demo wall-clock times are really impressive - beating the competition by a factor of two is definitely past the point where you should sit up and take notice.
The art of the non sequitur - saying "it's better" when at best it's just different
Words mean what I say they mean - saying "non-standard" when you mean "compatible with the published standard, but not our extensions"
Exaggeration for fun and profit - saying "we ship next quarter" when you mean "next year" "or maybe never"
Of particular interest will be the conclusion in which he shows how this is all ultimately for the customer's benefit...
LOOPS had:
A single-inheritance object system with GUI support (class, method, and object browsers).
Editing with structure editors that manipulated the parse tree directly. The structure editor was also used as the inspector in the debugger.
refactoring support in the browsers (select a method and move it to another class, etc).
automated global refactoring based on code analysis - this being 1987, it had a pseudo-natural language interface (EDIT ALL METHODS CALLING FOO AND REFERENCING *BAR*...).
LOOPS is one of the primary predecessors of CLOS (the other being Flavors).
Parse-tree-based editing has been around for awhile - Google for "syntax directed editing". Paradoxically, it works much better in Lisp than in other, more syntax-heavy domains; when your editor insists that everything be syntactically and semantically well-formed all the time, it's best that there be very few, very general syntax rules. This is why it also works in Smalltalk.
Unlike Godwin's Law, though, the probability that the claimant is correct also approaches 1.
If you're going to mention As The Apple Turns, you must also mention its evil twin site Crazy Apple Rumors.
I visit them both daily.
If creators feel the need for mechanical protection beyond the terms of copyright, they should not be able to use the legal system in addition to their fair-use-defeating technology.
But, it is possible to go below the minimum. The most egregious Apple example of this recently was the original G3 tower "hockey-puck" mouse, which sacrificed function nearly completely on the altar of form. I have one of these on a hand-me-down computer, and I can't use it without curling my fingers down over the front edge to find the cord - you can't orient it otherwise. ARGH!
The no-button pro mouse, on the other hand, is a work of art that works.
Here's a privacy policy I'd like to hear from a major corporation, just once:
We view our collection of data about our customer base as both an asset and a liability.
It is an asset in that we use it to generate aggregated data about you, and sell that data to people who want to sell you things.
It is also a liability in that we got this data from you with a promise to never release it in detail, allowing others to see your individual habits and preferences.
With that promise in mind, we repeat that we will never sell our data about you in a manner that allows you to be individually identified. If we ever go out of business, our last act as a corporation will be to destroy any indices that would enable that.
The core Google page ranking algorithms, yes. The Google clustering infrastructure, no. According to this, Google didn't scale beyond what would fit into a Stanford dorm room before it went commercial in 1998.
Raw speed will always be useful for problems that are hard to parallelize. Right now those problems (parts of crypto, some quantum physics calculations, etc.) are important scientifically, but away from the money.
Industry will spend R&D money on clustering for storage and reliability, without major government subsidy, because there's a crying need for it. How much government money went into Google/eBay/Amazon?
Government research is supposed to complement industry R&D - to be aimed at fields where the results are still important, but maybe not as profitable. This is why government should not abandon raw speed as a research goal.
Followed by:
2.1 China claims U.S. never went to Moon in 1969.
2.2 China is supported by moronic moon-landing denyers worldwide.
2.3 A century from now, history books say China landed on moon first in 21st century.
Scary enough for you?
The difference between Apple "Inoperable Failure" machines and Dell "Inoperable Failure"machines looks to be very small - the real difference between them is in machines that were "Broken but Still Operable".
Maybe they're seeing people who got their computers working, then discovered they were running XP...
[ducks, runs for cover...]
By quite a bit, actually. Look here down around 1989.
Also, on the same history page, in the Mid 80's section, you'll see an entry for an expert system named Pride developed at PARC. Pride helped Xerox design their first line of desktop copiers, and is quite famous within the company.
I worked for (long lamented) Xerox AI systems from 1986-88, and consulted for them off and on through 1994, which is how I know about this.
... and steal one that Just Works.
Sorry, I couldn't help myself...
These type of leaks traditionally occur around Halloween, don't they?
Park it at L1 or L2. Space is roomy, so building big things is easier. Aiming is easier - you do have to be more accurate and have a better collimated beam, but you only have to track it across a degree or two to cover the whole earth, and you could aim by tracking the whole ring, so you'd need less powerful deflector magnets. You can power it with solar energy. And, the vacuum is free!
There is the little problem of getting there, setting up shop, and building a 1000 km structure, of course...
What Minsky is actually flaming about here is the damage done to the field by the original "AI winter" and the real possibility of a new AI winter starting in a few years.
;-)
.sig below).
"AI winter" is the name given to the collapse of strong AI as a business model in the mid 80's - expert systems and symbolic AI in general didn't deliver on their promises, and so the money went away. As a guy who got his doctorate in AI in 1985, I can tell you all about it.
One of the major causes of AI winter was researcher hubris - lots of people hacked up systems that appeared to solve 80 percent of certain complex problems and then said "all that stands between us and a complete solution is money and time". For many of those systems, solving the last 20 percent would have taken 2000 percent of the time, if it could have been done at all. The tragedy of AI winter, though, is that basically all of symbolic AI was abandoned, though some of it is creeping back out into the light with obfuscated syntax (see my
What Minsky sees here is a lot of people heading down the same path, but with neural nets and small robots instead of expert systems. The new systems are doing some interesting things relative to the old symbolic AI systems (though they do have the advantage of 20 years of Moore's law to help them). But, will they scale up? Right now, nobody knows. If they don't, the last thing the field needs is another cycle of overpromise/underdeliver/abandon.
Maybe AI is just plain hard, and cracking it will take longer than one or two computer industry business cycles.
How big is the check I'm writing right now?
How fast is it?
With these as your evaluation function, you are guaranteed to get systems with little redundancy and little or no internal safety checks.
One regrettable example of this is the market for personal finance programs. The feature that sells Quicken is quick-fill - the heuristic automatic data entry that makes entering transactions fast. Never mind that Quicken's register file is fragile - it frequently loses track of balances (requiring the moral equivalent of fsck), and every few years the accumulated unfixed cruft causes a major failure, requiring insane fixes like exporting all your data as QIF files and reimporting it into a new register.
If Quicken were back-ended into a real database, with real transactions, real consistency checks, and real crash recovery, all this would go away. But it would make Quicken slower and require more hardware horespower to run it - the marketplace would punish them for improving their lives.
What the original article is proposing is:
We accept that systems will always suck
Therefore, we should build multi-level suckage damage control into them
Another possibility is:
We accept that there is a tradeoff between system speed and safety
Therefore, we take the speed hit where safety is important
They fit together better
They can make radical changes which require support from both sides, without going outside the company.
"Microsoft-like" means MS has a monopoly on the software, so they can tell the hardware guys what to make to maximize MS profits. This means:
They fit together better, because MS has to implement fewer options
There will be less radical innovation, because the hardware guys will have to get MS' blessing to sell anything really new to the lucrative corporate market that is the main target of this controlled architecture.