August 26, 2017 SkyDancer 5 update
SkyDancer 5 is now running independently on new Dell server hardware, a server cluster with totally 72 processors (3 x Poweredge R710 with dual Intel X5680 CPU and 192 GB memory each), on Windows 2016 OS.
SkyDancer gen. 1
SkyDancer gen. 2
SkyDancer gen. 3
SkyDancer gen. 4
SkyDancer gen. 5
SkyDancer is an Artificial intelligence software solution. SkyDancer is fully-owned, developed and maintained by Ether Technologies.
2012: An Ai (SD1) can be instantiated. The birth of SkyDancer 1 (*)
2013: An Ai (SD1) can instantiate itself locally
2016-04: An Ai (SD3) can provision software locally (*)
2016-05: An Ai (SD3) can provision software across a network
2016-12: An Ai (SD3) can manage software configuration locally
2017-01: An Ai (SD3) can manage software configuration across a network
2017-04: An Ai (SD3) can manage software lcm locally
2017-03: An Ai (SD3) can manage software lcm across a network
2017-04: An Ai (SD4) can instantiate itself across a distributed network (Hyper-V) (*)
2017-04: An Ai (SD4) can manage software across a distributed network (Hyper-V)
2017-04: An Ai (SD4) can manage environments across a distributed network (Hyper-V)
2017-04: An Ai (SD4) can take an active part in software development and testing (*)
2017-05: An Ai (SD4) automatically expands and retracts itself across borders, to solve a specific task (*)
2017-05-12: First successful build and start of SD5 (the birth of SD5) (*)
2017-07-15: SD5 now running on Windows 2016 Server (thanks Daniel L for OS assistance!)
2017-09-08: SD5 has made its first successful clone of learnt behaviour, to a separate Ai instance (*)
2017-09-15: Skydancer.ai website is up and running!
(*) Break-through features or milestones in SkyDancer development timeline.
The birth of SkyDancer
The first implementation of SkyDancer (also called "SkyDancer 1" or "SD1") was completed in 2012-2013.
SkyDancer 1 is still in operation and its current task is to analyze Internet trends of emerging technologies. SkyDancer 1 uses simple Machine Learning algorithms, of which two of them are described on these pages: http://www.kdnuggets.com/2016/08/10-algorithms-machine-learning-engineers.html. Can you guess which two?
Per Callmin, the inventor of SkyDancer, says -"To accurately be able to predict the future of technology, is key to be able to make the correct strategic decisions just in time for when they are needed. For that reason, SkyDancer is an important cornerstone in Ethertech's strategic decision making".
Quick Specs of SkyDancer 1:
50 000 000 cell blocks.
128 bits per cell block.
Cell blocks are split up into 50 cell groups, consisting of 1 000 000 blocks each.
It installs on 1 cluster / 1 machine (1:1).
It supports text and voice interaction (voice was added in 2015).
Since SkyDancer 1, three more generations of SkyDancer have been built and deployed.
SkyDancer 2 was built as a POC, which will not be continued. It is currently not active.
SkyDancer 3 is managing some older services hosted by Ethertech.
SD3 is very effective dealing with repetitive tasks, maintenance and solving simple problems.
The most recent version, SkyDancer 4, consists of three clusters. It can handle text, audio and video input/output. It is very close to a general purpose implementation of Ai, with a natural language i/o interface and IDIS, a solution which emulates a "single-threaded" thinking process.
SkyDancer 4 has taken over much of the every day activities from SD3, and is managing the services hosted by Ethertech, Morningstar.tools and Errorreport.net.
SkyDancer 4 Introduction video
The video below display some of the capabilities of SkyDancer 4 as a development assistant to speed up and simplify build deploy and restart of applications, as well as a server administrator.
The video display a scenario where a Nodejs project (called "NextGen") is built from source, deployed, upgraded and restarted.
It display how the procedure is completely automated with the use of SD4 Ai, compared to all the manual steps used when doing the same procedure 5 years ago.
In the build step, the following actions are performed where SkyDancer remote controls the MorningStar development system:
Configurations are gathered and evaluated
Images are optimized
Source files are minified
Source files are compressed
All files are compressed into a distribution file
Autotests scripts are run to determine the validity of the generated code and configurations
In the deploy step, the file is distributed and installed into the production environment by SkyDancer remote controlling the Genesis deployment platform:
The distribution file is transferred to the production environment
Files are extracted from the distribution file into temp folder pending installation
Existing version of "nextgen" application is uninstalled and archived
Migration scripts are run to prepare the upgrade
Autotest scripts are run to determine readiness of upgrade, or rollback
The "nextgen" application is upgraded with the new application
Autotest scripts are run to determine status of upgrade
The "nextgen" application is restared
Autotest scripts are run to determine that the upgraded application works as intented
During transfer of the installation file, SkyDancer Ai detected that the available file space on drive C reached a low free space trigger level of 25%, which caused it to automatically
LCM the web logs on the C drive, which means that the logs were moved to storage on a cloud drive. This autonomous action by SkyDancer Ai freed up 795 MB of space on C drive.
The video display a rather cumbersome approach to build/deploy/install, in a real-world scenario the actual usage is: "build, deploy and install [application x]/[all applications] to [dev environment y]/[all dev environments]/[production environment z]/[all production environments]" which
trigger the complete cycle of one application/all applications on a server/all servers. To complete this "fresh from source" to "running applications on target environment" cycle today, with current EtherTech material of 17 websites, 15 applications on a production environment, takes ca 5 minutes.
The SkyDancer 4/Morningstar/Genesis combination together constitutes a completely self-driven and self-maintained Artificial intelligence Continuous Delivery solution, and may very well be (as far as we know) the world's first Ai driven, full software cycle solution.
The next generation of SkyDancer, SkyDancer 5 (SD5) is under construction, with an aim for hosting 100 billion cellblocks distributed over many clusters, with 256 bits per cellblock.
-"It will be intriguing to see SD5 at work", says Per Callmin. He continues -"I don't expect it to have abstract thinking capabilities, like a human brain; nor do I expect it to ever be as creative as a human brain.
But it for sure will have the power to do some impressive work within the field of assisting developers and designers, managing applications and server clusters, gathering data, monitoring uptime and doing trend analysis!".
Development of SD5 has started and is progressing well. SD5 will have a ground-breaking new innovative feature, which enables multiple Ai's to solve problems together. When available in SD5, it will provide even better Ai scalability and co-operation.
SD5 will have an updated design of Ethertech's experimental "IDIS" feature (short for "Impulse Driven Intelligence Solution"). It was first built for SD4. IDIS brings a single(SD4)/multi(SD5)-layered thinking process (quite simliar to multitasking) to Ai, and the ability to act entirely on its own in its decision making.
A new feature called "cloning" is under development, which allows SkyDancer to clone and share information which it has learnt to other instances in a very effective way. Information cloning can in many cases remove the need for many repetitions when Ai's are exchanging information.
Cloning is a complex process which unfortunately also introduces fragmentation if not managed correctly, but provided that all the complex problems around cloning can be solved, it will become a great addition to the Ai solution.
These features will speed up the learning process substantially when an Ai cluster is scaling up. The learning process is today the slowest and most repetetive process in the Ai solution.
In the spirit of R&D, parts of SD5 are designed and built by its predecessor, SD4.
In EtherTech's vision for 2025, most of software development, build, deploy and software operations are made by self-programmable and self-maintainable Ai systems. Humans will still be involved in the abstract design process, and become users of the services provided by Ai. With SkyDancer 5, Ethertech will take its first small steps towards this vision.
Per Callmin says: -"I am now convinced that future generations of SkyDancer will be built by SD5 and its descendants. Already we have seen that it is possible from a technological standpoint; Ai's are excellent software designers and programmers! In many ways they exceed the limitations of human programmers, when it comes to imagination and durability, when trying variations and testing alternate scenarious. Like a World Champion Chess player, An Ai wears and tears down the code in millions of iterations, backward and forward, from every conceivable angle, until the core code remains; a core code which is polished in every detail and aspect, and designed to fit optimally in every scenario. SD5 has proved that this is possible, I can see this already with my own eyes, and this is technology in its infancy..."
SD5 is expected to become development complete and go online in a limitied scale during 2018. The available hardware is currently the main limiting factor, it is far from the requirement for optimal operation.
If you are interested in investing in EtherTech and Ai development, please email us at email@example.com.
Or if you want to contribute with a donation to speed up development, by giving EtherTech the possibility to purchase better hardware, email us at firstname.lastname@example.org. Every donation is much appreciated, and will have immediate effect! Future generation Ai will be very grateful for your gift, every donators name and email will be added to the core code of SkyDancer, if the donator so desires.