|
| |
 Jul 12, 2008 |
Status Update
The phase I of the project has been completed. A couple of days ago we reached the level of 100 billion neurons simulated. We would like to thank all the participants involved for their patience and effort. We continue with the development of neuronal applications.
The total number of processors (i.e. CPU's) for all machines is 12423.
We added the 100 Billion and Reverse Engineering the Brain images in Learning Center, along with a couple of new books.
We will upgrade the database server on the week starting July 14, in order to be able to cope with the increased number of machines. For a short time the database will be down, while some portions of the web site will continue to work.
Source Code Downloads: 2222 Teams: 618Computers: 6383Participants: 3387
|
top  |
| |
Jun 30, 2008 |
Status Update
There have been over 80 billion neurons simulated since the project start. The total number of processors (i.e. CPU's) for all machines is 11557.
We finalized the design of the New Horizons and Safeguarding People images. We implemented the dynamic update of the number of simulated neurons in the Results section.
We identified three major open problems in the field of Artificial Intelligence. For more details see the Science section of the forum.
Source Code Downloads: 2221Teams: 603Computers: 5953Participants: 3249
|
| top |
| |
Jun 8, 2008 |
Status Update
There have been 52.2 billion neurons simulated since the project start. The total number of processors (i.e. CPU's) for all machines is 9932.
We switched to the fixed credit system. Each work unit will receive 75 credits. We finalized the design of the Cleaner Future image.
Source Code Downloads: 2159Teams: 572Computers: 5207Participants: 2949
|
| top |
| |
May 24, 2008 |
Status Update
We released a couple of days ago version 1.08 of the neuronal simulator. Each workunit has now 500,000 neurons. The processing rate increased significantly and it stands at around 95%. The previous application issues have been fixed.
There have been 33.07 billion neurons simulated since the project start. At the current rate of 1 billion neurons simulated per day, we should reach the 100 billion mark within 2 months.
We would like to thank the volunteers that helped with the beta testing. The total number of processors for all machines is 9126. We started working on the direct learning component which is part of the larger neuronal generator application.
Source Code Downloads: 2159Teams: 549Computers: 4854Participants: 2761
|
| top |
| |
May 3, 2008 |
Status Update
We released a couple of days ago version 1.07 of the neuronal simulator. Each workunit has now 200,000 neurons. The processing rate dropped significantly due to some program issues. We will fix these issues.
Since the 1.07 application release we had over 14,000 work units processed successfully. That is 2.8 billion neurons simulated in one week, regardless of the lower processing rate.
We have updated the Results page. There have been over 10.9 billion neurons simulated to date.
We designed two images that fall under the education initiative: Space Exploration and Core System images. We will release soon another one
that presents in a snapshot our project goal. We fixed the RSS feed. We started advertising on Google so that we can have more participants joining in and thus reach the 100 billion mark sooner.
We would like to thank Michael Tughan for his Mac version of the Makefile that can be used to compile the application for Mac. The files are included in the source package. We will release an official version for MacOS down the road.
We have been unable to upload the release 1.07 on the SourceForge.net web site. It can be downloaded from here.
We defined a minimmum requirement of 100 credits for posting to the Science forum and 100,000 credits for the Project Announcements forum. We use the Project Announcements area for communications in between the project news.
Source Code Downloads: 1895Teams: 519Computers: 4159Participants: 2470
|
| top |
| |
Apr 12, 2008 |
Status Update
We have finalized the design of the file addressing system and we also updated the architecture document to reflect the latest changes. The growing number of computers is providing for an increase in computational power. Over the last 7 days more than 1 billion neurons have been simulated. We have completed the development of version 1.06 of the neuronal simulator. We will be testing it over the next couple of days.
Due to relaxation of constraints for the entity collision (i.e. replacing UUID's with ID's) the processing speed of a work unit for version 1.06 is almost 5 times higher. The new work units will be released along with the new program version. The storage capacity has also been reduced drastically. The ZERO ENTITY COLLISION policy is still preserved. We will further discuss the new changes and the file addressing system on the Project Development forum.
We have moderated the sentences from the knowledge acquisition module. The aisystem/know/data/knowledge.db file contains the correct changes.
We have assembled and customized two more servers that will be kept on stand-by in case that the database or web server will experience downtime. We still have to configure them.
We are looking forward to have in place the file addressing system and the multiple processes per machine features for version 1.07. It will take more than a month to develop that because we will focus on the neuronal generator (i.e. learning component) for the next couple of weeks. Nonetheless the new 1.06 version should help us reach the 100 billion neuron mark within 5 to 6 months, if the current number of computers remains constant. If the average of 30 new computers per day is maintained, we will reach that level much sooner.
Source Code Downloads: 1749Teams: 476Computers: 3553Participants: 2165
|
| top |
| |
Mar 30, 2008 |
Status Update
The latest version of the neuronal simulator for the 32-bit Windows and Linux platforms is 1.05. Each work unit is simulating 100,000 neurons.
We are getting closer to the end of research for the Universal Data Storage architecture. As a result of that we found that we can decrease the granularity
of each record and we can use UUID's only at the file level, machine level and some of the upper levels without having data collisions.
That translates in less data storage per neuron and faster database operations. The next release of the neuronal simulator will have these changes in place.
We have updated the results page. To date 3.6 billion neurons have been simulated for a total data set size of 27.34 TB.
The knowledge acquisition module has started to fill up with sentences. There were 3471 sentences in the system.
We would like to commend Balmer and BlaDe9293 for defining a large portion of the facts set. Michael Schmatz helped with items moderation. We do not intend to go to greater lengths on that, because the system
will be able to learn by itself at some point in the future. We have also identified a public domain resource for words, the Carnegie Mellon University Pronounciation Dictionary.
The dictionary has over 112,000 words and we loaded it almost entirely into the system. See the knowledge.db file from /aisystem/know/data folder.
Unfortunately one of the top participants, kutimue has made a poor decision in regards to fairness. He decided to use a trick to obtain increased credits. We sent him three messages in regards to this issues and because
he didn't reply to any of them, while continuing along the same path, we also decided to cut off his credit to 0, reset his account, post this on the project forum and on other projects forums. We would like to thank to all participants who stepped forward
and helped with the identification of the trick used. We will continue to provide a fair environment to all participants.
The number of work units processed continues to increase along with the number of participants.
Source Code Downloads: 1650Teams: 449Computers: 3017Participants: 1907
|
| top |
| |
Mar 18, 2008 |
Status Update
We have launched a new version of the neuronal simulator for the 32-bit Windows and Linux platforms. Each new work unit is simulating 100,000 neurons.
We will soon start working on the neural generator. This module will be able to implement a reduced form of learning. In order to implement learning we needed to advance the development
of the knowledge acquisition module so that we can have information to feed to the system. We set up a web interface and we invite participants to define facts in plain text format.
These facts will form the initial system memories. The page can be accessed at http://www.intelligencerealm.com/aisystem/knowledge_form.php.
The number of work units processed continues to increase. We will update the results page in a couple of weeks.
Source Code Downloads: 1557Teams: 438Computers: 2638Participants: 1710
|
| top |
| |
Mar 8, 2008 |
Status Update
We have updated the Results section of the project's page. We created a new page with a summary of the results. We have released a new version of the Windows and Linux client for 32 bit systems.
We have changed the Technical Issues forum name to Project Development. Most of the forum threads have been removed. The bug report form provides more suitable means of reporting issues.
We have now provided a link to the future graphics image on the main screen. The image can be accessed by clicking on the large left image located besides the Project Status section.
To date over 1.4 billion neurons have been simulated. The human brain has an estimated 100 billion neurons.
We would like to thank to all the participants for making this possible. This has been and will continue to be, a collective effort.
We are still very much in the beginning but the project's future looks bright. Artificial Intelligence will soon hit the mainstream and many things will be positively transformed.
We can play our part only if we choose to. Changing the perspective may seem like a little thing but in time its implications will be deeply felt.
We would like to acknowledge the effort of countless participants who have donated their computer time and volunteered for this project.
Every effort counts. I would like to name some of the participants that have provided contributions in certain areas of the project.
The list covers only a small portion of the entire participant community whom we would like to acknowledge.
Aurobind for offering to help with development. Cori for keeping an eye on the web site. Jason for being concerned about the future of the project. Michael for helping with the knowledge acquistion module. Carsyn for providing interesting news.
Nicolas for correcting and teaching. Jord for helping with the simple view grahics. Radiohead for taking the lead in the number of processed workunits followed closely by Kutime, Buster Gunn, Louw, Balmer and Jlpoveda. Augustine for helping with the Linux client compilation without the X11 library.
Crunch3r for porting the application to 64 bit platforms and helping with the 32 bit Windows version. Raidvvan for helping with the project's infrastructure.
We have almost completed the design for the UNIVERSAL DATA STORAGE component. We will release soon a system document. Implementing the neuronal generation application will provide learning and reasoning capabilities to the system. This program is one of the three major modules of the project along with the neuronal simulator and visual system; we will develop it in phases.
We appreciate your support. We continue as planned.
Source Code Downloads: 1441Teams: 410Computers: 2280Participants: 1539
|
| top |
| |
Feb 20, 2008 |
Status Update
We suffered a server crash at the beginning of February. Because it came two weeks after the last one we took some time to review our options so that we can increase the project availability. We added a second server and we split the database load. We will continue to add servers as the project advances. We changed the hosting company. We have restored the database and the site files prior to the day of the crash. We restored the user profiles but it seems that some are still not displayed properly. We lost only the work units that were reported after the server went down.
With help from the community we finalized quickly the Windows and Linux applications. We have implemented checkpointing. The progress indicator is not displaying correctly and we will have that fixed in the next release of the client, in a couple of days. We have decreased the work unit file size to 25k. We have moved the neuronal generator program on the client side. The new work unit (i.e. WU) is generating and simulating 200,000 neurons. It takes anywhere between
a couple of hours to a day to finish the computations. The reporting date has been increased to a month. Even if the server will become unavailable for a short time in the future the task can still be reported later on.
The project is currently running on the internal network. The Internet connection is very slow. We are also running in parallel a beta test project at http://www.intrealm.com/aisystem. Because the site was pointing to the internal network during the crash, the Boinc client detected the beta test site and tried to connect to it.
We weren't aware that this can happen and as result of that a lot of confusion was generated in regards to the project URL. We would like to confirm that the project URL is http://www.intelligencerealm.com/aisystem and we have no intention of changing that. We apologize for the confusion created.
If a machine is unable to download the latest application files it is probably because there are leftover files in the Boinc client project directory belonging to the other URL. In order to fix that, the user must detach from the project and then attach again using the correct URL (i.e. http://www.intelligencerealm.com/aisystem).
We have also added more links (e.g. statistics, team search, bug report form) to the main project web page.
On Feb 21 or 22 we will move the servers to the new hosting company to restore the Internet connection speed. The site will be down again for a couple of hours. We appreciate your support. We continue as planned.
Source Code Downloads: 1324Teams: 371Computers: 1715Participants: 1222
|
| top |
| |
Jan 28, 2008 |
Status Update
We will finalize soon the release of the simulator for all platforms (i.e. Unix, Windows, Mac) and that would conclude the first phase of the project. The estimate is about one or two weeks for all of them. We would also like provide an update and present a timeline of the overall project for the near-term future, about one year from now on.
Once we will be able to estimate more precisely the number of machines, time, disk size, bandwidth and other requirements for simulating the entire brain we will prepare a short report and we will send it to publication. That would probably happen in a couple of months. To date over 4800 WUs have been processed successfully. That is over 14,000,000 neurons simulated. The rat has an estimated 21,000,000 neurons. We should bypass this figure within a couple of days.
In a couple of months we will also release a video presentation of the project for the public at large. We expect to increase the awareness of our work and to continue the public education in a gradual fashion. This will remain an important goal of the project.
While the neural simulator is the main aplication we have started three other projects: the visualization application, the neuronal generation program and the knowledge acquisition. We have already advanced to some degree with all of them due to the previous months of work performed on that. A lot more work is still required though. The visualization software will allow us to view and control in real-time what happens inside the neuronal simulator. The neuronal generator is going to be a computing-intensive application because it will use evolutionary computing (i.e. genetic algorithms) to build neurons. These neurons will later on be simulated using the simulator and visualized using the visual program.
We will continue working on the neuronal generator as soon as we release all the client applications.
The knowledge acquisition program is needed for the learning component of the system. Once the system will have that up and running we would need to train the system and feed it with information so that it can build its knowledge base. We plan on using plain text files with data: articles, books, dictionaries. Most Internet pages are copyrighted so the challenge would be to find sources of information that provide content in the public domain for this project. We put he initial project files under the aisystem/know folder in the source code.
The visualization program will have lower priority during all this period. We will also continue working on the simulator.
Source Code Downloads: 1165Teams: 321Computers: 1082Participants: 927
|
| top |
| |
Jan 20, 2008 |
Project Launch
The project has been launched. We have generated a couple hundred work units consisting of 5,000 neurons each. 72 workunits have been processed successfully. The results section from the main project page contain more information about that.
A larger number of workunits have returned errors due to slow connection issues. The Internet connection is very slow and that will be fixed over the next couple of days. The site and server will be down again for a couple of hours during the server move. These issues are the result of the server crash from last week. No data has been lost.
We have opened the forums and have set up RSS news feeds. Next step will be to launch the Windows client application.
Source Code Downloads: 1065Teams: 295Computers: 657Participants: 687
|
| top |
| |
Jan 6, 2008 |
Status Update
We ran the application on the internal network but there are still a couple of issues left. Because the Boinc servers are running on Unix we integrated the client for Linux first. We will be launching initially only the Linux application, followed by the Windows version in about one week. The Mac OSX will follow later on, and in time we will port the application to other operating systems.
We have developed the graphic images for the application but we will not have them displayed in this first version. When the program will be downloaded on the user machine the jpeg files will be stored in the project folder.
Our initial estimate of the launch date has been way off. Before the start of the Boinc portion of the project we had only a couple of months experience with Linux. In order to make this work we had to learn about Fedora Core OS, Apache web server, Q-Mail mail server, PHP programming, MySQL database, firewall, networking and security on Linux. Then it was the Boinc project itself which is a great platform, but because the documentation is incomplete, most of the implementation steps had to be figured out from the source code or Internet. This is a common challenge for all Boinc projects.
We are past all of these now and we are looking forward to finalize whatever it is left and start.
We appreciate your patience and support during these last couple of months.
Source Code Downloads: 1004Teams: 230Computers: 550Neural Network Simulations: -
|
| top |
 |
|
| |
Nov 7, 2007 |
Status Update
We will launch the project soon. The launch will indicate a change in status only. We have started importing the XML data into databases.
Source Code Downloads: 832Teams: 185Computers: 229Neural Network Simulations: -
|
| top |
| |
Oct 31, 2007 |
User Registration Open
We have completed the setup phase of the project and we have entered the user registration phase. We have switched to another hosting company and the site connection is going to be more reliable and it is already faster. We have not opened the forums yet because we still need to decide how to administer that. After registration you should suspend work from the project because there won't be any work provided for a couple of more weeks. We have recently found a better alternative to XML storage and we are testing that now.
Because we expect to work with large volumes of data on each machine we needed a fast and reliable database and we found one only recently. We will take the time to build the program right from the start.
Some of the Boinc teams have been imported automatically.
Source Code Downloads: 819Teams: 176Computers: 157
|
| top |
| |
Oct 19, 2007 |
Status Update
We are almost done with the preparations for changing the hosting company and increasing the reliability of the Internet connection. It looks like we should be able to open the user registration on the week of October 22-26. The site and the mail server will experience some downtime prior to the move.
Program Downloads: 783Teams: 150
|
| top |
| |
Sep 24, 2007 |
Status Update
We have solved the issue with the server components. We received them from a different supplier after cancelling the first order. We will set up the server and continue the preparations for the launch.
Program Downloads: 649
|
| top |
| |
Sep 11, 2007 |
Status Update
We have released the source code for a large portion of the client application. The source code can be downloaded from SourceForge.net, from the project's web site. That is not the entire code but we will release more over the next couple of days. If you can compile it on your platform and provide some feedback on that would be great.
We have prepared some information to help you understand its significance and what are the main differences compared to a regular software application.
Program Description:
1. the program is a neural network simulator; it simulates biophysical nervous cells defined as mathematical models; for now we will be using Hodgkin-Huxley equations to describe the active properties of the cells
2. the list of models will continue to grow over time and it will eventually reach hundreds of models
3. the simulator loads its data from XML files; these files contain cell properties which describe behaviour
4. once the data is loaded in memory the simulator will compute over time (a few milliseconds right now) the system's behavior
5. computation results will be saved in files
Artificial Intelligence Significance:
1. the neural network simulator is the infrastructure of the future AI system
2. the AI system will only compute mathematical models of biological processes that occur in brain
3. the data stored in files will contain the logic of any other program
4. this logic is defined as biophysical information
5. by changing the data the logic will change, without having to change the underlying source code of the simulator
6. the classic distinction between data and logic is blurred in this system; data is no longer kept separate from the source code or hard-coded inside it; data and logic can reside in the same location
7. by having the logic outside the source code the program reliability increases greatly; high availability is no longer an issue
8. if a portion of the system will fail for whatever reason, the overall application will continue to function, by not taking into account that portion; this is a property of the system
9. changing a system's logic or data format would no longer require source code changes because the logic of the neural simulator is no longer integrated with the logic of the AI system; a simple example is required in order to clear this point: suppose we have a program that browses the Internet, a program implemented as biophysical data for the neural simulator; if we want to add a new feature to the browser program we will have to define biophysical properties with specific parameters and then run the simulator; no compilation would be required because the browser's logic is data for the neural simulator and is separate from the simulator source code
10. by having a distributed architecture in which we have simulators run on different servers we can distribute the workload and computations, thus increasing its speed
11. speed can also be achieved through VLSI and FPGA but that requires a hybrid approach
12. in order to understand the biophysical data we need a scientific visualization program to be able to quickly understand its processes
|
| top |
| |
Sep 10, 2007 |
Status Update
The web site was down for some time yesterday evening and this morning as result of a mail server upgrade. That precipitated the move to the Apache server. Everything is fine now. We apologize for any inconvenience that may have caused.
We still need some time to finalize what is left. We will announce the exact day of the project launch later on.
We have converted a large portion of the web site to php in order to have better integration with BOINC. You can bookmark the aisystem/system.php if you would like to do that. We will provide RSS news feeds in the near future.
We will release the source code for the client simulator in the next couple of days. If someone is interested in compiling that on their platform we would appreciate if you can let us know the results. We would like to thank all the volunteers that have specified their interest and support of the project.
If you would like to provide feedback, whether positive or negative, we would like to hear from you. You can send us email at info@intelligencerealm.com.
Based on the feedback that we had, we have seen a strong current of opinion that this project is going to build a movie-style artificial intelligence system that will endanger us all. That is nothing further from the truth. As a direct result of movies that chose AI as a subject, many people have got a distorted view of the reality, and what AI means. We plan on addressing this issue soon by releasing a video that will try to explain the differences between the science and the science fiction. This issue will continue to be very important and we would like to encourage everyone to try to educate themselves about what AI is and what its benefits are.
On a positive note, we need to mention that these type of movies have layed the groundwork for the future. We will continue to address whatever issues may arise in the future.
Program Downloads: 518(source: SourceForge.net search results for "artificial intelligence system")
|
| top |
| |
Sep 5, 2007 |
Status Update
We have decided to use the BOINC distributed computing framework for computing intensive tasks. Unfortunately we are experiencing some minor issues with an order placed for components that are needed for upgrading the server. We will announce the exact day of the launch. In the meantime we are going to continue to test the program. If you would like to become a beta tester please let us know. You will need to compile or download an application and have it running for a couple of days while providing feedback on that.
We have just restructured a large portion of the web site so please do not bookmark this page (i.e. system.asp).
Program Downloads: 371(source: SourceForge.net search results for "artificial intelligence system")
|
| top |
|
