|
|
Welcome to the Frontpage
|
2011 Progress (or lack thereof) |
 |
 |
 |
|
Written by Administrator
|
|
Sunday, 10 July 2011 19:43 |
|
2011 started out very well but has become a disaster. The primary work on the project is performed on the justin.frdcsa.org server which has 4 TB of project data. The project is divided up in many ways, including a git archive of all of the code, and a computer-specific data section that houses the application data. It was customary to store large files there so that they didn't bog down the git repo.
Unfortunately, while making extensive use of the data drive for a program for natural language paraphrase generation, the hard drive developed IRQ errors. This was a serious problem because currently there is no backup infrastructure, which may appear to be a colossal lapse in judgment, but rather is attributable to a lack of project funding. Going forward it is now the aim of the project to obtain a new server with a raid configuration (as well as an offsite backup server) in order to proof the system against possible future data loss. Also, one goal of our fundraising is to have data recovery be performed on the drive that failed.
Why should this matter? The reason this matters is that the FRDCSA is engaged in long term work to improve the state of open source software as it relates to improving the human condition. Whereas many AI projects are military in nature, this project differs in that the goal of the project is to provide tools that benefit individuals (and not military organizations), especially those in poverty and those with disabilities. The proliferation of cheap computers and smart phones, in conjunction with the capabilities of free software, enables essential services to be rendered, including a medical diagnostic system, a meal planning system, and a personal life planning and organization assistant. The aim is to augment the social safety net and to enhance the lives of people with tools for improving organization. Unfortunately in a for profit cultural environment most of these tools have not yet been created as free software. All of these tools benefit from the capabilities of software that is gathered and stored by the FRDCSA. An original objective of the project is to package previously unpackaged yet applicable software for Debian GNU+Linux.
On a more positive note, much technical progress has been made this year. Work has focused on the end-to-end life planning system, with the construction of additional parts of the Free Life Planning system.
http://frdcsa.org/~andrewdo/WebWiki/FreeLifePlanningCoachSoftware.html
We already have a system that helps the user set goals and compute and execute plans to achieve them (SPSE2/IEM). By contrast, action planner is concerned with weighing the importance of these goals in various situations, to enable automated replanning in cases where all the objectives cannot be met or in cases where unexpected and unmodelled failures occur.
Much work had been done on improving the Interactive Execution Monitor, which walks users through plans. Unfortunately most of that seems lost as it was not yet committed to the git repository.
And lately I have taken to working more on the natural language understanding components of the system, and in particular the Emacs environment for asserting knowledge from text. It would be very nice to acquire some of the recent efforts at ontology population from text. Fortunately, most of the methods from the Capability::TextAnalysis module have been made functional again, however, NLU research progresses more along the lines of NL-Soar in that we are looking to create a deliberating process that knows the timing and time complexity of various tasks and uses that information to guide the search for answers. But it is clearly a work in progress.
More information about the failure can be found as quoted in this Facebook post:
The hard drive is definitely hosed, will submit it to recovery operations once I get the funds (perhaps through grant procurement). There was no backup to the secondary drive. The reason the data was not stored in git was it was personal data for applications. Total lost was 600 GB. The main FRDCSA systems are of course still intact but their data sections will have been lost. It should have been rsynced to a remote drive. I am redesigning the data storage systems, would appreciate any knowledge of data storage like NAS or RAID, any advice to get the operation moving in the right direction especially as regards disaster recovery and recovery from hacking. The hard drive was a Seagate 1.5 TB drive that was known to brick frequently, got it on discount without that information. I was writing a natural language paraphrase generation program using the TERp data and was using MLDBM tied hashes with 800MB of data, and the program ran for 4 or 5 hours. Guess that was too much. It was main data storage, contained many virtual machines and did many other data intensive processes. Hoping the platter has not flaked so that there would be a chance of a full or at least partial backup. Drive makes clicking noises and read errors, discontinuing use. The drive itself contained sole backup of my laptop after that filesystem failed. All in all a sad day. However, sometimes I find losses like this refreshing in that you can build a new direction free from the errors of the past. Many design decisions are now undone but I can do it better this time around. Sort of like simulated annealing.
|
|
Last Updated ( Sunday, 10 July 2011 19:44 )
|
|
|
Written by Administrator
|
|
Friday, 08 October 2010 14:32 |
|
It has been so long that I have submitted an update that I kept passing the opportunity by because I forgot how to do it. The good news is the project is coming along really well. Many systems that were only wishes are now facts. Yet life seems to be passing me by, there is still not the unity of structure I would like behind the project. The good news is that in all this time the state of the free software world has moved immeasurably forward. Each new day brings new surprises. I will now proceed to give an overview of the state of the system. I do not yet have the todo list systems working properly. The mechanisms for reporting which tasks have been accomplished are therefore more or less lacking. For this reason, I cannot give a precise list of the accomplishments. Moreover, there is currently no estimation of the value of a particular task, although some code exists which computes it in a suboptimal way. So I cannot readily prioritize the announcements of which tasks have been completed. So, I will simply talk stream of consciousness. Thanks to the Zlatan's influence, much work has progressed on the usability of the project to a wider audience. Therefore, several GUIs have been developed. The advantage here is that the infrastructure for rapidly developing GUIs is moving forward, and that certain processes that are ultimately easier to perform with GUIs can be or have been programmed. The two most significant GUIs thus developed are SPSE2 and Paperless Office. SPSE2 is the shared priority system editor. It is useful for establishing and editing a big picture model of the project. Unfortunately, for it to be truly useful, there still remain serious obstacles. But it already functions and can even export temporal plans based on the facts. It uses FreeKBS2 as the backend and runs reasonably fast and is highly efficacious, just not perfectly so. Paperless Office as well has been implemented, enabling the full cessation of Windows use as Paperport is no longer required. While being a very substantial program, it also requires a lot of improvements to become the system it is meant to be. Releasing the project, which has been its central bane, is coming along. We had one grant for the release rejected, but with the help of some allies we are developing a grant writing campaign manager that should see some funds make it to supporting the release. As well, a video game under development has now been made to use the FRDCSA, tying its release to the project and hopefully accelerating it. Lastly, the Virtual Machine release has been accomplished and we now have build instructions for the project. This means that with a little work we can build the frdcsa-common package. It also means that, for those who apply and are approved to receive a pre-release version of the Panoply GNU+Linux, as it is called, can start using the system and get to know some of its aspects. It is also hoped that those who use it will be able to eventually contribute to its release, by means of a distributed GUI system for "declassifying" or rather deidentifying personal information from the codebases. Currently the system is available in two forms and a third will soon be made available. The virtual machine appliance may be shipped via a storage device like a hard drive or USB stick, it may be downloaded (caution 20GB), and soon a multi-DVD release will be made. As the git repos are continuously filled and new software of use to others becomes available, the VM images will be upgraded and released. But to pretend that this was all the news would be to reduce the tremendous progress all across the project to a few special areas that have taken only a fraction of the development time. There are just too many fascinating subjects to explore and it does not seem appropriate to mention one in favor of another, so I will simply conclude with an appeal to apply for the pre-release version and to become active in what may someday become the FRDCSA community. Oh, POSI is being woken from the ashes. It's website had to be moved because the original domain name lapsed, so it is temporarily being placed at http://posi.frdcsa.org . As POSI was conceived outside the range of the FRDCSA and is not a project (although it uses the FRDCSA tools), it is not an appropriate location and will be fixed when the resources are available. To all who have helped, your contributions have meant more than you probably know. |
|
Written by Administrator
|
|
Thursday, 12 February 2009 00:33 |
|
Wow, so I’m really cruising along on the development of various software systems. I wrote a web front end for the software that will soon run POSI (http://posithon.org). It is a goal tracking and knowledge representation system used to model people’s goals, interests, and abilities and use that information to help form teams to solve specific problems. Related to this is the Job-Search resume matching program, which had been written earlier. I’m half-way through writing a web front end so anyone can submit their resume and have the results processed. This includes creating a daily update script for the scraper. The resume matcher takes your resume, parses all the different skills and experience, and uses this to query Craigslist job postings to find the best matches. Lastly, I’ve written code for Reasonbase that populates a wiki with information that has been extracted from text. It runs OpenCalais along with other text analysis procedures on the submitted URI, and these provide information for all the named entities, which are then fed into a Wiki. The first instance of this, the Tara Knowledge Archives, is (partially) online. http://posithon.org/reasonbase/tarawatch Tell your friends, join #posi @ freenode to work with others on interesting projects of your/their choosing. The FRDCSA release is coming along, although I haven’t spent too much time on it recently. Of course, all the software I mentioned above with be copied via git to the VM before being released. There are so many projects I want to work on. But reality is calling. Disability will soon expire and then I’ll be looking for a job. So if you know anyone that wants to hire, please let them and me know. I am
also open to doing consulting once the disability has ended. If you are looking for consulting work, I am writing code to consolidate consulting tasks using the POSI infrastructure and information. |
|
|
Good Things Come to Those Who Wait |
|
|
|
|
Written by Administrator
|
|
Monday, 19 January 2009 06:23 |
|
Hello friends and bots, the FRDCSA will be available to download soon - in all its glory. A working, configured virtual machine will be the easiest way to distribute the system for now given the myriad dependencies and other configuration items that have not been ironed out yet. Just load it into the Virtual Machine software of your choice and you have the nascent FRDCSA at your fingertips. I would like it if those who are considering downloading it email me, but of course that's not necessary. As there is no coherent documentation, please feel free to contact me for explanation of how the software might suit your needs. Also, all who discover the software systems of the FRDCSA could help by documenting them. Okay, thanks. I'm sure that you won't be disappointed, but chances are you won't know which tools are useful to you. |
|
Relative Consistency Proofs |
|
|
|
|
Written by Administrator
|
|
Saturday, 27 December 2008 21:27 |
|
In the course of a proof, at each step, when you make an inference, you have to protect the inference from leading to a false conclusion, because the axioms and or rules of inference could very well be inconsistent. What assurance do you have for the safety of the conclusion? How has it been secured? And where would you put a proof of the consistency of the axioms and the rules of inference? Ultimately, perhaps this must reside in the mind of the prover. It is here that the prover may be tricked. If she wishes to verify some set of axioms and rules of inference, she must either do it in another proof system, or in her head (or perhaps I am mistaken). Anyways, this is what a fresh pair of eyes yields upon rediscovering mathematical logic. |
|
|
|
|
<< Start < Prev 1 2 3 Next > End >>
|
|
Page 1 of 3 |
|
|
