I think a significant hint lies in your use of the word "motivation". In order to be motivated to do something, you have to have conscious thought and a sense that there is something wrong happening. When it comes to accessibility and software development, accessibility issues doesn't even enter the mind of most software developers and if it does, it's overruled by most managers as irrelevant to the majority of the customer base. interesting calculation would be to see if it's cheaper to warehouse disabled people than it would be to make the world accessible to them. from people's actions, I would say that their intuitive sense is that it would be cheaper to warehouse. I'm not saying that's what they think but that's what they demonstrate in their actions.
Maybe I'm just a know-nothing 'tab', but I think your disability has twisted your mind a little bit.
the fact that you raise this question at all shows that you are more aware than 99.9% of all software developers. as for twisted mind, please consider that I'm posting on Slashdot. A little more seriously however try living in a world where applications can degrade the accuracy of your input device. For example, Thunderbird, Firefox, open office, chatzilla, aim, emacs, ulipad, jedit, nvu and others I can't remember right now all cause serious degradation in recognition accuracy. Other applications massively fail with minimal correction techniques (i.e. natural text). For example, tools like pyscripter which uses Smart IDE type technology, fail because if you try to correct a misrecognition, the wrong text is selected, deleted, and overwritten with the correction. then you have applications like VM Ware workstation which doesn't accept any textual input from NaturallySpeaking which means in order to gain access, ssh in, use X11 forwarding to bring up a window to display on your XP X11 server. Then, maybe then you can dictate into an application. But even if you get the application to accept input from speech recognition, you have the problem that the built-in macros for cutting and pasting follow Windows conventions and can't be changed. As you probably can guess, any application without Windows style cut and paste doesn't work very well. A major shortcoming is you can't use the NaturallySpeaking Select-and-Say feature which is an absolute godsend for hands-free editing. Without it, you burn a significant amount of hand time just getting the cursor to the right place and changing the right text. smart completion text boxes are another barrier. For example, in Firefox the search engine bar will drop down when the focus is in that window. If the drop-down is present, you can't dictate into that window. Many JavaScript editors will destroy your text if you try to correct a misrecognition. If focus shifts in the middle of a recognition output, it's effectively like typing random keys on the keyboard and commanding your application to do God knows what. I'm always forgetting recognition is turned on when I'm in Thunderbird. And I'm always losing messages because of many characters shoved into Thunderbird. I could go on with more examples but I hope you get the idea of what you would bump into within the first day of use.
You do not need to take my word on all the failings I've described above. You can verify them for yourself. Pick up a copy of NaturallySpeaking, purchase a real microphone instead of the piece of crap in the box, install, train, use after throwing away your keyboard. Then you'll get a very clear example of just how inaccessible your working set of applications are. Then you get to make a choice about whether you are going to do something to make systems more accessible to speech recognition users or not.
not exactly true. There's a very interesting project called voice coder developed at nrc-it in Canada. It translates limited English expressions into code. The reason that approach was chosen is because most software using the current style of bmpyNms is literally unpronounceable and would require spelling out letter by letter which does incredible damage to your voice as well as your temper. As I pointed out elsewhere, the complete and total lack of a backdoor API also makes it extremely difficult to use speech recognition for programming. The last thing needed is the ability for the IDE to export all symbols and their scope so that the speech recognition environment can create a very high accuracy and grammar using those symbols.
On the other hand, I have been writing Python using speech recognition for a few years and I completely violate all coding standards because I'm not going to burn my vocal tract to comply with some tab's idea of good looking code. I have been using my hands to navigate and I only use voice to create symbols and comments. to make the jump to the next level, we will need editors with the ability to navigate to various features of code such as arguments, methods, nested parentheses, etc. these features are useless for hand users but invaluable for voice users. But, as usual, the problem is that the handicapped need tabs to write the code for them in the initial bootstrap phase.
Reading the comments I'm see a bunch of tabs[1] with no clue about being disabled, the speech recognition market, the history of the product, and how nuance is probably hampered by the management attitude towards money and the history of the code base.
for someone who's been disabled (temporarily or permanently) speech recognition means the difference between making a living and being able to support oneself, a mortgage, family etc. and sitting around on your ass in section 8 housing on Social Security disability. Pain from RSI once made it extremely difficult to feed myself. When you've experienced that level of pain, disability and the associated despair, you get the attitude that anything that gives a disabled person independence and an ability to make a living should be encouraged with all possible resources.
Listening to someone dictating using speech recognition will drive you mad. You would have the same problem with a blind person listening to text-to-speech. But that's not the fault of speech recognition or text-to-speech. That's the fault of management not providing the disabled person with an acoustically isolated environment (i.e. reasonable accommodat.
Desktop speech recognition is a monopoly because it's extremely expensive and difficult to develop speech recognition and there is not a large market. the market consists of lawyers, doctors, and the disabled. There is not enough money to support two companies (or more) to develop desktop speech recognition applications.
NaturallySpeaking is very buggy. There are bugs that cause people problems that were first seen in NaturallySpeaking 5. These are not hidden or hard-to-find bugs. They don't affect nuances ability to sell NaturallySpeaking. There's no reason for them to fix them except for the fact that they interfere with the use of many programs by the disabled. If you are just doing dictation into Microsoft Word or DragonPad, you'll never notice. If you try to dictate into Thunderbird, Firefox, Open office,... you're screwed. For example, I cannot dictate directly into Firefox for this comment, I need to use a workaround for dictation and then paste the result into the text box. The reason why this problem exists is because nuance management has the reputation of not making any change or feature unless you can make a business case and show them they will get revenue from that change. This is not such a bad model because it can keep nuance profitable and product available to people who truly need it (i.e. the disabled). The downside is that it doesn't leave room for changes necessary for the disabled.
I've heard from people working inside dragon that part of the problem also is the code base. It was written by a bunch of Ph.D.'s who are really really good at speech recognition but are not so good at writing code. Also in the last few years, there has the huge turnover and people working on the code as NaturallySpeaking was sold first to L&H and then to nuance. That kind of change alone will wreak havoc on the code base as knowledge is lost and never really acquired by the new people. by the way, I have talked with some people from nuance, and they are basically good people. They understand the needs of the handicapped but they are constrained in what they can do for us because of budget and resources.
When people talk about alternatives with open source speech recognition, only a tab would think they would work for the disabled. Their recognition speed is significantly slower, vocabulary size is smaller, and they are really more projects to keep grad students busy than be anything useful in the real world.
The last problem with speech recognition sits in your lap if you are a manager of a software product or a developer. As far as I can tell, the number of applications that are speech recognition friendly is vanishingly small. It seems to me that software developers go out of their way to make software handicap hostile. It starts with the multiplatform GUI toolkits that do not
crap, crap, crap. I hit every one of those points. Self driven learning, always exploring new forms of technology and then applying it to the current business, etc. etc. Maybe if I hadn't been so good, my hands would still work and I wouldn't be living with constant pain up to my elbows. What's ironic is I left my father's rigging business (machinery moving) because I saw so many people around me losing fingers, damaging their limbs and back etc.. and I wanted to go do something where I wouldn't be injured on the job. Since I had been programming in high school, I thought hey, this is not a bad career. You work indoors, you're not covered in grease, you don't breathe toxic chemicals and you aren't going to get injured by heavy machinery falling on you.
what's missing here is a recognition of "natural monopoly". A natural monopoly exists when increasing competition results in increasing per customer costs.
Another way to think of it is in a town of population X, it costs Y$ to serve the entire town. Every competitor and entering the market using the same type of technology will have to expend Y$ to service that town. If you have three competitors, they need to make their revenue on one-third of the customer. If you have four competitors, they now need to make their revenue on 1/4 of the customers. end result being that the cost for service increases with competition. This model of economics for last mile also applies if you have mechanisms for covering last mile that have unequal costs as long as they are rival services. Non-rival services are like comparing apples and oranges.
Another missing component in the discussion has been structural separation. With structural separation, you have one company/town/whoever maintaining the infrastructure as a regulated monopoly. that infrastructure is guaranteed to have non-discriminatory access for all comers. This way, you have competition where it works (i.e. on services).
Re:Sorta agree with both points of view
on
Singularity Sky
·
· Score: 1
on the never having heard of checks and balances, if you study your history, you will find that the honor Harrington series is derivative of the one of the many times when England and France were at war with each other. That military tactics then as in the book were based on "ships of the wall" and that you could travel faster than you could communicate.
Weber did a good job at translating the politics of the era to fictional future. Hence the pro royalty stance.
obviously, we need to have a longer conversation. Feel free to contact me directly via the link on the contact section of the camram web site.
This is true of all proof of work systems. You could get really lucky and meet the criteria for "done" on the first try. On average however you will take the target amount of time. Which means sometimes it will take longer and sometimes it will take shorter to reach "done".
Now on average, every time you increase the cost of a stamp by a bit, you double the average cost. So if a 22 bit stamp takes 15 seconds on average, a 23 bit stamp will take 30 seconds on average. Now it's also possible to encounter a 26 or 32 bit stamp in the search for a lower value one. There's no magic or exploitation involved, it's just how sha1 and the search for the right completion work. Think dumb-F'n-luck. which is why I choose to use the desired number as a predicate and use a simple go/no go. Other interpretations are possible but less predictable.
Seriously, contact me directly and I highly recommend playing with the hashcash code from hashcash.org and really get a good feel for what it means to generate stamps. There's nothing like hands-on experience at this point.
problem is that the number of bits of collision found is a probabilistic event. You always have at least the number you requested but sometimes you can have as much as 10 or 15 bits more because that is just what you stumbled across in search for the collision. It's always safest to say whether or not it passed the minimum number of bits collision threshold and not that it has a certain number of bits collision.
I suggest you try this using the hashcash executable. Run the process for about a week and log the number of collision bits found versus number of times it was found. Its quite illuminating.
camram project has successfully used hashcash for stamp generation and message acceptance. We find that about 15 to 20 seconds computation is about the right amount to seriously bankrupt spammers. (paper on this coming soon)
zombies are a problem but the nice thing about proof of work puzzles such as hashcash is that they make the zombie machines get hot which is noticeable by normal users. They also run real slow. Again something to draw the users attention to a problem. in any case, the numbers are real close. There's still more spam than the number stamps generated by the number of known zombies. Since the upper bound for spam is set by the number of zombies, this is a serious incentive to kill zombies.
Mailing lists are problematic but if one uses a second type of stamp based on signatures, then the problem goes away. In the meantime, using hybrid system, you do not require anything special of mailing lists and you are no worse off than you are with typical content filters.
The camram project is very close to releasing 0.2 which will make available a hybrid sender pays system which will work for systems handling a single user through a few hundred users. With this release will also come the information of how to convert any content filtering antispam defense into a hybrid sender-pays system like camram.
As of today, 3 systems support sender-pays using hashcash: gnus, spamassassin, and camram. it's important for more systems to support an open standard for sender-pays. So if you are deeply involved in an antispam content filter, please consider adding hashcash as part of the system.
thank you for the correction. My view that the technology career in general (hardware and software engineering) was fading comes not only from reading such unbiased sources as slashdot but also from reading EE Times and IEEE newsletters.
you are indeed in a fortunate place. Make the best of it as you can. It almost makes me want to go back to school but I think it a bit late (almost 50).
I understand that you probably have your heart set on a technology career but I would strongly encourage you to look elsewhere for your life's work. the technology career in the United States is fading. There is significant age discrimination and it is effectively a ten to fifteen year career.
Try some informational interviews at technology companies and just look around and see how the people in the technology staff and first couple levels of management are above the age of 45. If the companies say they have a "dual career ladder", ask how many directors they have on the managerial side. Then ask how many they have on the technical side. if they give you a nonzero number, ask to be introduced to some of them. Another question on the same line is to ask what does it take to become a director for managerial and then ask for the technical. You'll frequently find that the technical rungs have significantly higher hurdles than the managerial side.
Don't be fooled by the typical/.comments of "I'm over X, and I still have a job by being technically hot shit" because they are exceptions that prove the rule. For the most part, your typical your career will be over by the time you are 35-40.
A technology career is also bad for you physically and mentally. Most companies use subtle or not so subtle psychological pressure to encourage staff to work all sorts of hours, usually in the name of teambuilding. It will cost you sleep, health by being increasingly sedentary and obese, and even possibly repetitive motion damages which leaves you with lifelong pain.
The psychological pressure to work long hours will reduce your ability to take time off to take vacations.
The hyper focus mindset it takes to get work done in a cube environment also will impact your ability to form healthy relationships with a partner. Important time off together (see above) will be impaired and nibbled away at by the inability to leave work at work.
So, leave the technology career for others. The smart move into something where you can have a long career and make good money without putting your physical and mental health at risk. take care of yourself. Because not only will nobody else do it, everybody else wants to eat you alive and not in a good way.
went out and bought one today and within a couple of hours, my hands were hurting so bad that I went back to the old cheap Memorex keyboard I was using.
I was really hoping it would work out to be a nice keyboard but unfortunately, I was disappointed yet again. As the data point, I have been disappointed by virtually every keyboard I used for one reason or another be it reliability, key pressure, noise, etc. Maltron, kenisis, ms natural have failed my finger tests. They all suck but they all suck in unique ways.;-)
a thing to remember is that if someone can prevent a spammer from communicating based on identity (or lack thereof), you can be silenced as well.
This is why I have put my efforts into sender-pay systems and specifically the camram project. We invite you to please come and join us in the effort to build a decentralized, user-friendly, freedom-of-speech supporting antispam system and hit spammers in the pocketbook.
Remember when PC chip approached the GHz, there was a bunch of people that were fearing they would interfere with TV and other applicance? None of that happened.
actually, the things that happened were much sooner than that. Back in the bad old days of S. 100 bus systems, there was significant interference to radio and television. If there was a computer turned on, FM radios and over the air television was useless. The FCC stepped in and required certification for emissions levels. As result you'll now find on every piece of gear some form of class A or B certification listing.
As result, there is very little interference no matter what CPU frequency. Assuming of course, you can keep the top on;-)
x10 is very low-frequency control signals. Below the range of most receivers. Not so with the case of BPL. It uses spectrum from roughly 1 MHz all the way up through 80-100 MHz. As others have pointed out, it also has the problem that if something radiates, it can also receive which means any user of the HF spectrum like aircraft and military can easily interfere with BPL
it doesn't work that way. person gets injured, sues company, company buries person with payoff or 3ed party slander campain to defuse the outrage over the event. company continues making shitty product.
A.R. theories are ok if there exists open accuate info on products and a choice of companies. too often, faults are surpressed and companies collude to eliminate choice and you have nothing to say about it.
also A.R. ignored realties of market places esp where companies use their power to crush those that dare to challenge them. the free market tends to a monoply and it must be restrained. the lust for power is not limited to those in government.
Before you flame me about how your favorite information should be free consider that information includes:
- child porn pictures or other snuff
can't justify that.
- virus/worm/hacking tool source code and instructions
exposure of flaws is the only tool shown to get manufacturers to acknowledge the defects in their products and fix them. You expose flaws by creating the tools that exploit them otherwise they can just dismiss the report as unsubstantiated.
- stolen intelectual property (for example: HL2 source)
in a word, Diebold. Diebold has engaged in the coverup campaign to disguise just how vulnerable their electronic voting machines are to tampering.
- [fill in other human rights violation here]
again, any place where publicity or public exposure is necessary to move the rich and powerful from places of comfort.
any good challenge response system will try to identify mailing lists so they won't send out a challenge. This is a relatively safe behavior because at worst, if a spammer tries to mimic a mailing list, they won't get a challenge and their e-mail will sit in the spamtrap.
As for Mailer demons messages, I will argue for making them generate postage as well. That way, you do not need to white list them. Of course, this assumes you are using a proof of work postage system which doesn't need any centralized infrastructure. See hashcash.org/camram.org
You are also quite right that it will take a series of fixes which is analogous to the drug cocktail approach that medical sciences using against drug-resistant diseases. Camram currently incorporates automatic white listing based on traffic and a Bayesian style filter with user settable thresholds in addition to the postage stamp and a handicapped user-friendly postage due notice mechanisms.
I was blessed with a deep look into the making of two desktop speech recognition products. It's an extremely expensive process to make recognition engines which explains the lack of options. The development of sphinx to the same level as NaturallySpeaking will probably suck down about $5 million worth of development effort before you have a usable product. The example you gave of 1000 word vocabulary is a toy recognizer. If you aren't running at least 100,000 words continuous speech recognition at 120 words a minute, it will be more frustrating than useful to anyone accustomed to current standards in speech recognition.
Personally I view command and control speech recognition as worse than useless. The real hand killer is text entry such as this message or program code. Using alternative input device for pushing buttons and menu selection is far less injuring than trying to use your voice in simulate mouse operations.
For a very reasonable sum, code weavers could make NaturallySpeaking work on linux using their wine implementation. this is a win in a lot of ways. It makes speech recognition available on Linux and gives handicapped users an option if they want to leave windows behind. It also gives vendor a sense for the size of the Linux market for their product which would influence the development of a native implementation.
There really isn't much choice in speech recognition. There is only a single vendor for desktop speech recognition today and that is scansoft. IBM has left the market and handed the remnants of the ViaVoice product line to scansoft. The rest of IBM's research is now aimed at IVR systems and not desktop recognition. The only other vendors that I know of is Phillips and, when last I looked, their engine was so far behind the technology curve that I don't consider it usable.
If I had any say in the direction of Gnome accessibility, I push the NaturallySpeaking on code weavers and bridge the NaturallySpeaking API to Gnome. This would solve the immediate problem and provide a gold standard by which one could evaluate other solutions.
unfortunately, the accessibility support for Gnome is only half of the story. There are many more people out there who suffer from "slight" mobility impairments like RSI, arthritis, and other problems of the hands or arms. These impairments keep them from using a keyboard to a significant degree and circus tricks like unicorn stakes or paddles are so inefficient as to be not worth using.
These people need speech recognition and full featured speech recognition like that provided by NaturallySpeaking. The current demonstration recognizers like Sphinx don't come anywhere near the capabilities needed by handicapped users.
today, these people cope by walking away from computers, getting keyboard slaves to type for them, or by using Windows.
The needs of these people have been ignored because they are silent unlike the blind community. They are silent because the chronic pain and day-to-day struggle to make a living leaves no energy for advocacy but this does not make their needs any less important.
Linux will not be completely handicapped accessible until you can operate the system either by discovering your monitor and listening or by discarding your keyboard and speaking.
(This message written using Windows-based speech recognition)
yes it does scale up. One of the principles is that "strangers cost, friends fly free". A mailing list is a "friend" and therefore, there is no charge to receive mail from the mailing list.
Today, mailing lists are not as easily white listed as I would like. In the future when we start using opportunistic signatures, then mailing list white listing will become extremely easy.
problem has already been considered and solved. The camram project uses a recipient bound token as its "payment". There's no need for any central infrastructure, it can't be co-opted by any central organization, it hit spammers where it hurts (throughput of messages, economics) and it can't be forged.
Take a look at the camram project you'll find a practical, working implementation of sender pays email today.
Slashdot Mirror
What would be their motivation for doing this?
I think a significant hint lies in your use of the word "motivation". In order to be motivated to do something, you have to have conscious thought and a sense that there is something wrong happening. When it comes to accessibility and software development, accessibility issues doesn't even enter the mind of most software developers and if it does, it's overruled by most managers as irrelevant to the majority of the customer base. interesting calculation would be to see if it's cheaper to warehouse disabled people than it would be to make the world accessible to them. from people's actions, I would say that their intuitive sense is that it would be cheaper to warehouse. I'm not saying that's what they think but that's what they demonstrate in their actions.
Maybe I'm just a know-nothing 'tab', but I think your disability has twisted your mind a little bit.
the fact that you raise this question at all shows that you are more aware than 99.9% of all software developers. as for twisted mind, please consider that I'm posting on Slashdot. A little more seriously however try living in a world where applications can degrade the accuracy of your input device. For example, Thunderbird, Firefox, open office, chatzilla, aim, emacs, ulipad, jedit, nvu and others I can't remember right now all cause serious degradation in recognition accuracy. Other applications massively fail with minimal correction techniques (i.e. natural text). For example, tools like pyscripter which uses Smart IDE type technology, fail because if you try to correct a misrecognition, the wrong text is selected, deleted, and overwritten with the correction. then you have applications like VM Ware workstation which doesn't accept any textual input from NaturallySpeaking which means in order to gain access, ssh in, use X11 forwarding to bring up a window to display on your XP X11 server. Then, maybe then you can dictate into an application. But even if you get the application to accept input from speech recognition, you have the problem that the built-in macros for cutting and pasting follow Windows conventions and can't be changed. As you probably can guess, any application without Windows style cut and paste doesn't work very well. A major shortcoming is you can't use the NaturallySpeaking Select-and-Say feature which is an absolute godsend for hands-free editing. Without it, you burn a significant amount of hand time just getting the cursor to the right place and changing the right text. smart completion text boxes are another barrier. For example, in Firefox the search engine bar will drop down when the focus is in that window. If the drop-down is present, you can't dictate into that window. Many JavaScript editors will destroy your text if you try to correct a misrecognition. If focus shifts in the middle of a recognition output, it's effectively like typing random keys on the keyboard and commanding your application to do God knows what. I'm always forgetting recognition is turned on when I'm in Thunderbird. And I'm always losing messages because of many characters shoved into Thunderbird. I could go on with more examples but I hope you get the idea of what you would bump into within the first day of use.
You do not need to take my word on all the failings I've described above. You can verify them for yourself. Pick up a copy of NaturallySpeaking, purchase a real microphone instead of the piece of crap in the box, install, train, use after throwing away your keyboard. Then you'll get a very clear example of just how inaccessible your working set of applications are. Then you get to make a choice about whether you are going to do something to make systems more accessible to speech recognition users or not.
not exactly true. There's a very interesting project called voice coder developed at nrc-it in Canada. It translates limited English expressions into code. The reason that approach was chosen is because most software using the current style of bmpyNms is literally unpronounceable and would require spelling out letter by letter which does incredible damage to your voice as well as your temper. As I pointed out elsewhere, the complete and total lack of a backdoor API also makes it extremely difficult to use speech recognition for programming. The last thing needed is the ability for the IDE to export all symbols and their scope so that the speech recognition environment can create a very high accuracy and grammar using those symbols.
On the other hand, I have been writing Python using speech recognition for a few years and I completely violate all coding standards because I'm not going to burn my vocal tract to comply with some tab's idea of good looking code. I have been using my hands to navigate and I only use voice to create symbols and comments. to make the jump to the next level, we will need editors with the ability to navigate to various features of code such as arguments, methods, nested parentheses, etc. these features are useless for hand users but invaluable for voice users. But, as usual, the problem is that the handicapped need tabs to write the code for them in the initial bootstrap phase.
Reading the comments I'm see a bunch of tabs[1] with no clue about being disabled, the speech recognition market, the history of the product, and how nuance is probably hampered by the management attitude towards money and the history of the code base.
for someone who's been disabled (temporarily or permanently) speech recognition means the difference between making a living and being able to support oneself, a mortgage, family etc. and sitting around on your ass in section 8 housing on Social Security disability. Pain from RSI once made it extremely difficult to feed myself. When you've experienced that level of pain, disability and the associated despair, you get the attitude that anything that gives a disabled person independence and an ability to make a living should be encouraged with all possible resources.
Listening to someone dictating using speech recognition will drive you mad. You would have the same problem with a blind person listening to text-to-speech. But that's not the fault of speech recognition or text-to-speech. That's the fault of management not providing the disabled person with an acoustically isolated environment (i.e. reasonable accommodat.
Desktop speech recognition is a monopoly because it's extremely expensive and difficult to develop speech recognition and there is not a large market. the market consists of lawyers, doctors, and the disabled. There is not enough money to support two companies (or more) to develop desktop speech recognition applications.
NaturallySpeaking is very buggy. There are bugs that cause people problems that were first seen in NaturallySpeaking 5. These are not hidden or hard-to-find bugs. They don't affect nuances ability to sell NaturallySpeaking. There's no reason for them to fix them except for the fact that they interfere with the use of many programs by the disabled. If you are just doing dictation into Microsoft Word or DragonPad, you'll never notice. If you try to dictate into Thunderbird, Firefox, Open office,... you're screwed. For example, I cannot dictate directly into Firefox for this comment, I need to use a workaround for dictation and then paste the result into the text box. The reason why this problem exists is because nuance management has the reputation of not making any change or feature unless you can make a business case and show them they will get revenue from that change. This is not such a bad model because it can keep nuance profitable and product available to people who truly need it (i.e. the disabled). The downside is that it doesn't leave room for changes necessary for the disabled.
I've heard from people working inside dragon that part of the problem also is the code base. It was written by a bunch of Ph.D.'s who are really really good at speech recognition but are not so good at writing code. Also in the last few years, there has the huge turnover and people working on the code as NaturallySpeaking was sold first to L&H and then to nuance. That kind of change alone will wreak havoc on the code base as knowledge is lost and never really acquired by the new people. by the way, I have talked with some people from nuance, and they are basically good people. They understand the needs of the handicapped but they are constrained in what they can do for us because of budget and resources.
When people talk about alternatives with open source speech recognition, only a tab would think they would work for the disabled. Their recognition speed is significantly slower, vocabulary size is smaller, and they are really more projects to keep grad students busy than be anything useful in the real world.
The last problem with speech recognition sits in your lap if you are a manager of a software product or a developer. As far as I can tell, the number of applications that are speech recognition friendly is vanishingly small. It seems to me that software developers go out of their way to make software handicap hostile. It starts with the multiplatform GUI toolkits that do not
crap, crap, crap. I hit every one of those points. Self driven learning, always exploring new forms of technology and then applying it to the current business, etc. etc. Maybe if I hadn't been so good, my hands would still work and I wouldn't be living with constant pain up to my elbows. What's ironic is I left my father's rigging business (machinery moving) because I saw so many people around me losing fingers, damaging their limbs and back etc.. and I wanted to go do something where I wouldn't be injured on the job. Since I had been programming in high school, I thought hey, this is not a bad career. You work indoors, you're not covered in grease, you don't breathe toxic chemicals and you aren't going to get injured by heavy machinery falling on you.
Ha.
what's missing here is a recognition of "natural monopoly". A natural monopoly exists when increasing competition results in increasing per customer costs.
Another way to think of it is in a town of population X, it costs Y$ to serve the entire town. Every competitor and entering the market using the same type of technology will have to expend Y$ to service that town. If you have three competitors, they need to make their revenue on one-third of the customer. If you have four competitors, they now need to make their revenue on 1/4 of the customers. end result being that the cost for service increases with competition.
This model of economics for last mile also applies if you have mechanisms for covering last mile that have unequal costs as long as they are rival services. Non-rival services are like comparing apples and oranges.
Another missing component in the discussion has been structural separation. With structural separation, you have one company/town/whoever maintaining the infrastructure as a regulated monopoly. that infrastructure is guaranteed to have non-discriminatory access for all comers. This way, you have competition where it works (i.e. on services).
on the never having heard of checks and balances, if you study your history, you will find that the honor Harrington series is derivative of the one of the many times when England and France were at war with each other. That military tactics then as in the book were based on "ships of the wall" and that you could travel faster than you could communicate.
Weber did a good job at translating the politics of the era to fictional future. Hence the pro royalty stance.
also see Horatio Hornblower for a similar genre
obviously, we need to have a longer conversation. Feel free to contact me directly via the link on the contact section of the camram web site.
This is true of all proof of work systems. You could get really lucky and meet the criteria for "done" on the first try. On average however you will take the target amount of time. Which means sometimes it will take longer and sometimes it will take shorter to reach "done".
Now on average, every time you increase the cost of a stamp by a bit, you double the average cost. So if a 22 bit stamp takes 15 seconds on average, a 23 bit stamp will take 30 seconds on average. Now it's also possible to encounter a 26 or 32 bit stamp in the search for a lower value one. There's no magic or exploitation involved, it's just how sha1 and the search for the right completion work. Think dumb-F'n-luck. which is why I choose to use the desired number as a predicate and use a simple go/no go. Other interpretations are possible but less predictable.
Seriously, contact me directly and I highly recommend playing with the hashcash code from hashcash.org and really get a good feel for what it means to generate stamps. There's nothing like hands-on experience at this point.
---eric
problem is that the number of bits of collision found is a probabilistic event. You always have at least the number you requested but sometimes you can have as much as 10 or 15 bits more because that is just what you stumbled across in search for the collision. It's always safest to say whether or not it passed the minimum number of bits collision threshold and not that it has a certain number of bits collision.
I suggest you try this using the hashcash executable. Run the process for about a week and log the number of collision bits found versus number of times it was found. Its quite illuminating.
camram project has successfully used hashcash for stamp generation and message acceptance. We find that about 15 to 20 seconds computation is about the right amount to seriously bankrupt spammers. (paper on this coming soon)
zombies are a problem but the nice thing about proof of work puzzles such as hashcash is that they make the zombie machines get hot which is noticeable by normal users. They also run real slow. Again something to draw the users attention to a problem. in any case, the numbers are real close. There's still more spam than the number stamps generated by the number of known zombies. Since the upper bound for spam is set by the number of zombies, this is a serious incentive to kill zombies.
Mailing lists are problematic but if one uses a second type of stamp based on signatures, then the problem goes away. In the meantime, using hybrid system, you do not require anything special of mailing lists and you are no worse off than you are with typical content filters.
www.camram.org
The camram project is very close to releasing 0.2 which will make available a hybrid sender pays system which will work for systems handling a single user through a few hundred users. With this release will also come the information of how to convert any content filtering antispam defense into a hybrid sender-pays system like camram.
As of today, 3 systems support sender-pays using hashcash: gnus, spamassassin, and camram. it's important for more systems to support an open standard for sender-pays. So if you are deeply involved in an antispam content filter, please consider adding hashcash as part of the system.
check out http://www.camram.org http://www.hashcash.org
thank you for the correction. My view that the technology career in general (hardware and software engineering) was fading comes not only from reading such unbiased sources as slashdot but also from reading EE Times and IEEE newsletters.
you are indeed in a fortunate place. Make the best of it as you can. It almost makes me want to go back to school but I think it a bit late (almost 50).
I understand that you probably have your heart set on a technology career but I would strongly encourage you to look elsewhere for your life's work. the technology career in the United States is fading. There is significant age discrimination and it is effectively a ten to fifteen year career.
/.comments of "I'm over X, and I still have a job by being technically hot shit" because they are exceptions that prove the rule. For the most part, your typical your career will be over by the time you are 35-40.
Try some informational interviews at technology companies and just look around and see how the people in the technology staff and first couple levels of management are above the age of 45. If the companies say they have a "dual career ladder", ask how many directors they have on the managerial side. Then ask how many they have on the technical side. if they give you a nonzero number, ask to be introduced to some of them. Another question on the same line is to ask what does it take to become a director for managerial and then ask for the technical. You'll frequently find that the technical rungs have significantly higher hurdles than the managerial side.
Don't be fooled by the typical
A technology career is also bad for you physically and mentally. Most companies use subtle or not so subtle psychological pressure to encourage staff to work all sorts of hours, usually in the name of teambuilding. It will cost you sleep, health by being increasingly sedentary and obese, and even possibly repetitive motion damages which leaves you with lifelong pain.
The psychological pressure to work long hours will reduce your ability to take time off to take vacations.
The hyper focus mindset it takes to get work done in a cube environment also will impact your ability to form healthy relationships with a partner. Important time off together (see above) will be impaired and nibbled away at by the inability to leave work at work.
So, leave the technology career for others. The smart move into something where you can have a long career and make good money without putting your physical and mental health at risk. take care of yourself. Because not only will nobody else do it, everybody else wants to eat you alive and not in a good way.
as always, ego before excellence..
Architectural Neutral Distribution Format has been around for years and solves many of the same problems (and more).
I guess it is one more time around the (reinvention) wheel for sun.
went out and bought one today and within a couple of hours, my hands were hurting so bad that I went back to the old cheap Memorex keyboard I was using.
;-)
I was really hoping it would work out to be a nice keyboard but unfortunately, I was disappointed yet again. As the data point, I have been disappointed by virtually every keyboard I used for one reason or another be it reliability, key pressure, noise, etc. Maltron, kenisis, ms natural have failed my finger tests. They all suck but they all suck in unique ways.
a thing to remember is that if someone can prevent a spammer from communicating based on identity (or lack thereof), you can be silenced as well.
This is why I have put my efforts into sender-pay systems and specifically the camram project. We invite you to please come and join us in the effort to build a decentralized, user-friendly, freedom-of-speech supporting antispam system and hit spammers in the pocketbook.
camram antique documentation (too busy writing code to write new documentation)
actually, the things that happened were much sooner than that. Back in the bad old days of S. 100 bus systems, there was significant interference to radio and television. If there was a computer turned on, FM radios and over the air television was useless. The FCC stepped in and required certification for emissions levels. As result you'll now find on every piece of gear some form of class A or B certification listing. As result, there is very little interference no matter what CPU frequency. Assuming of course, you can keep the top on ;-)
x10 is very low-frequency control signals. Below the range of most receivers. Not so with the case of BPL. It uses spectrum from roughly 1 MHz all the way up through 80-100 MHz. As others have pointed out, it also has the problem that if something radiates, it can also receive which means any user of the HF spectrum like aircraft and military can easily interfere with BPL
it doesn't work that way. person gets injured, sues company, company buries person with payoff or 3ed party slander campain to defuse the outrage over the event. company continues making shitty product.
A.R. theories are ok if there exists open accuate info on products and a choice of companies. too often, faults are surpressed and companies collude to eliminate choice and you have nothing to say about it.
also A.R. ignored realties of market places esp where companies use their power to crush those that dare to challenge them. the free market tends to a monoply and it must be restrained. the lust for power is not limited to those in government.
Before you flame me about how your favorite information should be free consider that information includes: - child porn pictures or other snuff
can't justify that.
- virus/worm/hacking tool source code and instructions
exposure of flaws is the only tool shown to get manufacturers to acknowledge the defects in their products and fix them. You expose flaws by creating the tools that exploit them otherwise they can just dismiss the report as unsubstantiated.
- stolen intelectual property (for example: HL2 source)
in a word, Diebold. Diebold has engaged in the coverup campaign to disguise just how vulnerable their electronic voting machines are to tampering.
- [fill in other human rights violation here]
again, any place where publicity or public exposure is necessary to move the rich and powerful from places of comfort.
any good challenge response system will try to identify mailing lists so they won't send out a challenge. This is a relatively safe behavior because at worst, if a spammer tries to mimic a mailing list, they won't get a challenge and their e-mail will sit in the spamtrap.
As for Mailer demons messages, I will argue for making them generate postage as well. That way, you do not need to white list them. Of course, this assumes you are using a proof of work postage system which doesn't need any centralized infrastructure. See hashcash.org/camram.org
You are also quite right that it will take a series of fixes which is analogous to the drug cocktail approach that medical sciences using against drug-resistant diseases. Camram currently incorporates automatic white listing based on traffic and a Bayesian style filter with user settable thresholds in addition to the postage stamp and a handicapped user-friendly postage due notice mechanisms.
I was blessed with a deep look into the making of two desktop speech recognition products. It's an extremely expensive process to make recognition engines which explains the lack of options. The development of sphinx to the same level as NaturallySpeaking will probably suck down about $5 million worth of development effort before you have a usable product. The example you gave of 1000 word vocabulary is a toy recognizer. If you aren't running at least 100,000 words continuous speech recognition at 120 words a minute, it will be more frustrating than useful to anyone accustomed to current standards in speech recognition.
Personally I view command and control speech recognition as worse than useless. The real hand killer is text entry such as this message or program code. Using alternative input device for pushing buttons and menu selection is far less injuring than trying to use your voice in simulate mouse operations.
For a very reasonable sum, code weavers could make NaturallySpeaking work on linux using their wine implementation. this is a win in a lot of ways. It makes speech recognition available on Linux and gives handicapped users an option if they want to leave windows behind. It also gives vendor a sense for the size of the Linux market for their product which would influence the development of a native implementation.
There really isn't much choice in speech recognition. There is only a single vendor for desktop speech recognition today and that is scansoft. IBM has left the market and handed the remnants of the ViaVoice product line to scansoft. The rest of IBM's research is now aimed at IVR systems and not desktop recognition. The only other vendors that I know of is Phillips and, when last I looked, their engine was so far behind the technology curve that I don't consider it usable.
If I had any say in the direction of Gnome accessibility, I push the NaturallySpeaking on code weavers and bridge the NaturallySpeaking API to Gnome. This would solve the immediate problem and provide a gold standard by which one could evaluate other solutions.
unfortunately, the accessibility support for Gnome is only half of the story. There are many more people out there who suffer from "slight" mobility impairments like RSI, arthritis, and other problems of the hands or arms. These impairments keep them from using a keyboard to a significant degree and circus tricks like unicorn stakes or paddles are so inefficient as to be not worth using.
These people need speech recognition and full featured speech recognition like that provided by NaturallySpeaking. The current demonstration recognizers like Sphinx don't come anywhere near the capabilities needed by handicapped users.
today, these people cope by walking away from computers, getting keyboard slaves to type for them, or by using Windows.
The needs of these people have been ignored because they are silent unlike the blind community. They are silent because the chronic pain and day-to-day struggle to make a living leaves no energy for advocacy but this does not make their needs any less important.
Linux will not be completely handicapped accessible until you can operate the system either by discovering your monitor and listening or by discarding your keyboard and speaking.
(This message written using Windows-based speech recognition)
actually, I dealt with this in a rough outline of models in a September 2000 presentation I gave at the Digital Commerce Society of Boston.
Making money in a post Napster world
yes it does scale up. One of the principles is that "strangers cost, friends fly free". A mailing list is a "friend" and therefore, there is no charge to receive mail from the mailing list.
Today, mailing lists are not as easily white listed as I would like. In the future when we start using opportunistic signatures, then mailing list white listing will become extremely easy.
problem has already been considered and solved. The camram project uses a recipient bound token as its "payment". There's no need for any central infrastructure, it can't be co-opted by any central organization, it hit spammers where it hurts (throughput of messages, economics) and it can't be forged.
Take a look at the camram project you'll find a practical, working implementation of sender pays email today.
http://www.camram.org and camram.sourceforge.net