that old bum had he napoleans head he surely woulda sung
effegies for fleacers and pre raphaelites
a long stream of sneezes
among the uptights
anactoria's embrace
although lewd made haste
as sappho caught her taste
and history was undone
a screaching parlance a mischevious dance
Poe's passing said he launched in greatness
the last of the enchanced
what meat begins with the letter G?
Ask not old man swineburn
his lady, or the tree
Wednesday, June 26, 2013
Wednesday, June 19, 2013
Google Goofs With Ray Kurzweil
Ray Kurzweil wrote a book on "How to Create a Mind" and somehow Sergey Brin of Google was so dumb in these areas as to think that Kurzweil had said something significant. He hadn't. All Kurzweil did was rehash 1980s research and ideas by others and add a bit of modern brain imaging. Kurzweil is no more the messenger of how the brain works than Gould was for evolution. Both were huxsters and nothing more.
Anyone who knows the history of A.I. will recognize that the basic theory (and even the diagrams that are used to illustrate it) is very much in the spirit of a textbook model of vision that was introduced in 1980, known as neocognitron.
Now that isn't to say that Kurzweil didn't start a music revolution with his synthesizer, and his blind reading machine, but that was 1960s technology and theres a big what have you done for me lately.
Sergey specifically hired Kurz to work on natural language processing, an area Ray knows little to nothing about.
I'm reminded of Hinton giving a ted talk about neural network recognition of individual letters. And as I watched I thought HEY this looks familiar. Nothing had changed since the 1980s. These old FRUMPS are not the leaders. Who are? Well people like me who are actually there in the details pushing on the real leading edge and testing mecanisms of neural design. Edelman's younger protege Olaf Sporns -Networks of the Brain
Even more disappointing is the fact that Kurzweil never bothers to do what any scientist, especially one trained in computer science, would immediately want to do, which is to build a computer model that instantiated his theory, and then compare the predictions of the model with real human behavior. Does the P.R.T.M. predict anything about human behavior that no other theory has predicted before? Does it give novel insight into any long-standing puzzles in human nature? Kurzweil never tries to find out.
Anyone who knows the history of A.I. will recognize that the basic theory (and even the diagrams that are used to illustrate it) is very much in the spirit of a textbook model of vision that was introduced in 1980, known as neocognitron.
Now that isn't to say that Kurzweil didn't start a music revolution with his synthesizer, and his blind reading machine, but that was 1960s technology and theres a big what have you done for me lately.
Sergey specifically hired Kurz to work on natural language processing, an area Ray knows little to nothing about.
I'm reminded of Hinton giving a ted talk about neural network recognition of individual letters. And as I watched I thought HEY this looks familiar. Nothing had changed since the 1980s. These old FRUMPS are not the leaders. Who are? Well people like me who are actually there in the details pushing on the real leading edge and testing mecanisms of neural design. Edelman's younger protege Olaf Sporns -Networks of the Brain
Even more disappointing is the fact that Kurzweil never bothers to do what any scientist, especially one trained in computer science, would immediately want to do, which is to build a computer model that instantiated his theory, and then compare the predictions of the model with real human behavior. Does the P.R.T.M. predict anything about human behavior that no other theory has predicted before? Does it give novel insight into any long-standing puzzles in human nature? Kurzweil never tries to find out.
Tuesday, June 18, 2013
From Windows 7 To Linux Mint - The Lenovo Dragon Breathes Peacefully
I've been struggling to do development on my big lenovo dual pentium 3.4 gHz machine using windows 7. When pushed, this box goes into full dragon mode and huge fans scream to life as the CPUs peg 100% usage. The problem is, in Windows 7, it seems like doing almost ANYTHING pegs the CPU at 100% all part of the NSA spying software they've installed no doubt.
When the system would come back from sleep it would animate through all the background images I had missed locking my system for ten minutes.
Countless times the whole thing would just freeze up. Gool ol Indian H-1B engineering no doubt. So, while Windows 7 runs great on a i7 8 core CPU it runs like crap on a brute of a box that it should be screamingly fast on. What gives? Obviously all the developers hadn't bothered to test it with older boxes.
Well it got so bad I couldn't run a database, a browser, and eclipse at the same time. So I gave up and installed MINT. Unlike PeppermintOS which is ugly MINT is a full 1 Giga distribution so you can't just slap it on a CD. So I put in a DVD and tried to burn it BUT turns out my refurb has only DVD reading capabilities. That sucks. OK so I slapped it on my smart key and used Auto-Load partition to get it recognized as a proper drive. It ran. But unlike when you boot with it and are allowed to test it all out it just asked "do you want to install it into your windows.." and feeling terrified I was about to blow away my boot I paused.
Then I got angry. Hell yeah blow away windows. I don't care I cant take it any more!
So I clicked. And in a few minutes MINT came up. And my windows 7 boot was nicely preserved.
And then it dawned on me. The beautiful silence. Both CPUs were running at 15%. Which was as it should be. In fact I cannot ever get the CPU to run at 100% dragon breath mode on Linux.
American engineers. God bless em.
When the system would come back from sleep it would animate through all the background images I had missed locking my system for ten minutes.
Countless times the whole thing would just freeze up. Gool ol Indian H-1B engineering no doubt. So, while Windows 7 runs great on a i7 8 core CPU it runs like crap on a brute of a box that it should be screamingly fast on. What gives? Obviously all the developers hadn't bothered to test it with older boxes.
Well it got so bad I couldn't run a database, a browser, and eclipse at the same time. So I gave up and installed MINT. Unlike PeppermintOS which is ugly MINT is a full 1 Giga distribution so you can't just slap it on a CD. So I put in a DVD and tried to burn it BUT turns out my refurb has only DVD reading capabilities. That sucks. OK so I slapped it on my smart key and used Auto-Load partition to get it recognized as a proper drive. It ran. But unlike when you boot with it and are allowed to test it all out it just asked "do you want to install it into your windows.." and feeling terrified I was about to blow away my boot I paused.
Then I got angry. Hell yeah blow away windows. I don't care I cant take it any more!
So I clicked. And in a few minutes MINT came up. And my windows 7 boot was nicely preserved.
And then it dawned on me. The beautiful silence. Both CPUs were running at 15%. Which was as it should be. In fact I cannot ever get the CPU to run at 100% dragon breath mode on Linux.
American engineers. God bless em.
Friday, May 17, 2013
JSON Exposed Repositories & EclipseLink vs. Linq
There is a devilishly difficult pattern question in java architecture which is emerging as new deviceology throws us to the wolves of mobile devices and cloud beastments which is the ubiquitous need for json-able ORMs and AJAX feeders for data. After three years pondering a hand hacked and painful repository pattern using hand squaged DAOs floating over Linq For Buggy Entities 3.5 and Linq for almost fixed Entities 4.0 that can now finally expose a property used in a link rather than force you to use a table trigger - I finally came to explore the tough patterns back in the java world which had gotten somewhat better but not completely.
So the first devil sausage says use the beatiful new JPA plugin which does visualization in Eclipse Juno to blast out your db tables and your entity beans then just slam some JSF and some pretty facets and UI wonders into the front end and blammo beaty and beauty for not much cost and time. To put darning on the socks EclipseLink will now also blast out all the JSON you need over rest for your AJAX calls.
But what if you chose the pure route - ditch the JSF and go straight HTML with Javascript and a boofy subslather like Knockout or Angular. Is this heresy to the java world? Moreover will you lose all the pretties that you get with something like PrimeFAces and other nice beasts?
It's a tough one. More and more JSF client workups seem dated and old. But on the other side if you've GOT the power of JSF why not use it. tough tough.
Things are changing as cloud and mobile demand robust ajax servers and tooling to spin json at a mile a minute. the question is can the developers keep up. You would think that after 100 years of this someone would set up a model centric framework that just blasts out all the rest for you. No not Roo. And not stuff that starts with clunky UML either.
In a way I kinda like the new graphical entity modeler in Eclipse. It's def the way to go. Now if they can just include the JAXB annotation properties and intelligent defaulting so you get BOTH XML, JSON and entities all keyed out right. Still it's miles ahead of Linq and .Net.
So the first devil sausage says use the beatiful new JPA plugin which does visualization in Eclipse Juno to blast out your db tables and your entity beans then just slam some JSF and some pretty facets and UI wonders into the front end and blammo beaty and beauty for not much cost and time. To put darning on the socks EclipseLink will now also blast out all the JSON you need over rest for your AJAX calls.
But what if you chose the pure route - ditch the JSF and go straight HTML with Javascript and a boofy subslather like Knockout or Angular. Is this heresy to the java world? Moreover will you lose all the pretties that you get with something like PrimeFAces and other nice beasts?
It's a tough one. More and more JSF client workups seem dated and old. But on the other side if you've GOT the power of JSF why not use it. tough tough.
Things are changing as cloud and mobile demand robust ajax servers and tooling to spin json at a mile a minute. the question is can the developers keep up. You would think that after 100 years of this someone would set up a model centric framework that just blasts out all the rest for you. No not Roo. And not stuff that starts with clunky UML either.
In a way I kinda like the new graphical entity modeler in Eclipse. It's def the way to go. Now if they can just include the JAXB annotation properties and intelligent defaulting so you get BOTH XML, JSON and entities all keyed out right. Still it's miles ahead of Linq and .Net.
Tuesday, February 12, 2013
New Book - Foundational Principles of Cognitive Cybernetics
I've been stumbling with this one. It's my first serious book. I want something that will change the world. It's no small task.
What is cognitive cybernetics? And what are foundational principles? It attacks what must be the basis for developing synthetic thought, real thinking intelligent machines.
I got so mad reading Kurzweils childish pathetic attempt to understand all this that I just had to throw my gauntlet down in the ring.
In the meantime, "Poems for College Curmudgeons" will be coming out soon. A more light hearted affair. But this book, this big book sometimes I crumple into a paper ball just thinking of the hugeness and difficulty. I worry it will be a decade to complete but somehow it needs doing now. Onward.
What is cognitive cybernetics? And what are foundational principles? It attacks what must be the basis for developing synthetic thought, real thinking intelligent machines.
I got so mad reading Kurzweils childish pathetic attempt to understand all this that I just had to throw my gauntlet down in the ring.
In the meantime, "Poems for College Curmudgeons" will be coming out soon. A more light hearted affair. But this book, this big book sometimes I crumple into a paper ball just thinking of the hugeness and difficulty. I worry it will be a decade to complete but somehow it needs doing now. Onward.
My first computer....
It's hard to get across to newbies how far we've come. My first computer only had six letters I like to tell people. It looked like this:
OK that's a 6502 mine was a 8008. You'd punch away at that keyboard for hours entering pre-written machine codes. You'd have to calculate the jump offset of your branch or if statements. Can you imagine? Well anyways your reward for two hours of typing in low level codes was a number that looked like your expected result, something like 2718 on your calculation of the e constant. But of course it would say it in hex so it would say A9E. Sometimes it said A8E and you'd spend hours looking for a mistake. That's what computers were. Not video games. Not word processors. Not web browsers. A9E. Think about it.
Often when it said A8E you'd have to get out a soldering iron. Because the boards were early designs to support the processors and get them out for testing, but since so much was in design the boards would have jump wires where they needed a new connection. A lot would come lose or disconnect slightly. Then you'd get a program which said A9E sometimes and A8E sometimes. In the words of Tron you'd "bring out the logic probe" and track it down. Sometimes these connections gave way when it was hotter or more humid. Things we never worry about today. Can you imagine software that worked but not on rainy days? Yep that's the way it was.
We programmed a "star wars" game for our HP41CVs, a RPN programmabe uber calculator the size of a brick.
An MIT student stole mine and I was so pissed about that.
From RSKEY.org - "A stunning calculator even by today's standards, more than 20 years after its introduction, the HP-41C remains one of the best handheld calculating devices ever conceived. The HP-41C line remained in production for over 10 years, practically defining the high-end calculator industry throughout the 1980s. Numerous machines remain in use today, no doubt due in part to versatility offered by the calculator's four expansion ports. The HP-41C has been used in the most exotic places, including the Space Shuttle or the cockpit of the Concorde."
Why was it called the CV? Well because of course it had FIVE TIMES the memory of the 441 byte HP41C. TWO THOUSAND TWO HUNDRED and THIRTY THREE bytes. I mean, can you imagine the luxury of having a thousand bytes to spare? I constantly tell my programmers that I don't want to hear them whine that a 2k memory leak isn't a big deal, that leak was bigger than my whole development platform on the HP 41!
Here is what our first video game looked like: (each line of characters was displayed one at a time, one line per second. The sequence below would take six seconds:
> <
> <
> <
> <
> * <
Now what this was meant to simulate was luke firing his torpedo into the death star. You'd guess the left right position of the tube and enter a number. If you got it right you'd see:
XXXXXXXXXXX
that was a video game from the early eighties geek style. In order to see this game you'd need to load your code up with a strip reader, ten strips about four inches wide and half an inch tall. Having the extra magnetic card reader meant you were cool at a level that granted you immediate VAX access. We were geeks but the word hadn't been invented yet.
A rebuilt HP41cx with a card reader costs over 500 bucks today. That's how much people loved them. Can you imagine writing your senior thesis on a calculator with one line of text at a time?
''The HP-41CX was the same as the HP-41CV but added the Time module (stop watch plus clock with alarms), an Extended Functions / Extended Memory module, a text editor, and some additional functions."
Later the HP 71B came out. While the 41 only had 4k ram cartridges, whopping endlessly huge 32k RAM cartridges could be had for the 71. At the time it seemed like madness, who could possibly write a program that required 32,000 characters?
Later the ill-fated HP 42 would come out with TWO lines of text. TWO! Such luxury after starting with just four letters just a few years prior. This was amazing stuff. Where would it all end we wondered, where would it all end. My best friend fish told me that soon there would be portable calculators with eight lines of text but I told him that wasn't likely. I mean really four lines were more than enough for anyone.
OK that's a 6502 mine was a 8008. You'd punch away at that keyboard for hours entering pre-written machine codes. You'd have to calculate the jump offset of your branch or if statements. Can you imagine? Well anyways your reward for two hours of typing in low level codes was a number that looked like your expected result, something like 2718 on your calculation of the e constant. But of course it would say it in hex so it would say A9E. Sometimes it said A8E and you'd spend hours looking for a mistake. That's what computers were. Not video games. Not word processors. Not web browsers. A9E. Think about it.
Often when it said A8E you'd have to get out a soldering iron. Because the boards were early designs to support the processors and get them out for testing, but since so much was in design the boards would have jump wires where they needed a new connection. A lot would come lose or disconnect slightly. Then you'd get a program which said A9E sometimes and A8E sometimes. In the words of Tron you'd "bring out the logic probe" and track it down. Sometimes these connections gave way when it was hotter or more humid. Things we never worry about today. Can you imagine software that worked but not on rainy days? Yep that's the way it was.
We programmed a "star wars" game for our HP41CVs, a RPN programmabe uber calculator the size of a brick.
An MIT student stole mine and I was so pissed about that.
From RSKEY.org - "A stunning calculator even by today's standards, more than 20 years after its introduction, the HP-41C remains one of the best handheld calculating devices ever conceived. The HP-41C line remained in production for over 10 years, practically defining the high-end calculator industry throughout the 1980s. Numerous machines remain in use today, no doubt due in part to versatility offered by the calculator's four expansion ports. The HP-41C has been used in the most exotic places, including the Space Shuttle or the cockpit of the Concorde."
Why was it called the CV? Well because of course it had FIVE TIMES the memory of the 441 byte HP41C. TWO THOUSAND TWO HUNDRED and THIRTY THREE bytes. I mean, can you imagine the luxury of having a thousand bytes to spare? I constantly tell my programmers that I don't want to hear them whine that a 2k memory leak isn't a big deal, that leak was bigger than my whole development platform on the HP 41!
Here is what our first video game looked like: (each line of characters was displayed one at a time, one line per second. The sequence below would take six seconds:
> <
> <
> <
> <
> * <
Now what this was meant to simulate was luke firing his torpedo into the death star. You'd guess the left right position of the tube and enter a number. If you got it right you'd see:
XXXXXXXXXXX
that was a video game from the early eighties geek style. In order to see this game you'd need to load your code up with a strip reader, ten strips about four inches wide and half an inch tall. Having the extra magnetic card reader meant you were cool at a level that granted you immediate VAX access. We were geeks but the word hadn't been invented yet.
A rebuilt HP41cx with a card reader costs over 500 bucks today. That's how much people loved them. Can you imagine writing your senior thesis on a calculator with one line of text at a time?
''The HP-41CX was the same as the HP-41CV but added the Time module (stop watch plus clock with alarms), an Extended Functions / Extended Memory module, a text editor, and some additional functions."
Later the HP 71B came out. While the 41 only had 4k ram cartridges, whopping endlessly huge 32k RAM cartridges could be had for the 71. At the time it seemed like madness, who could possibly write a program that required 32,000 characters?
Subscribe to:
Posts (Atom)