Working with Artificial intelligence, a dev blog
by Per Callmin
SkyDancer 5 progress
The work with SD5 is progressing well. Some challenges have been encountered with data cloning, as it for some reason causes fragmentation of data when it
is linked in with other structures. However we expect the fragmentation problem to be solved shortly. The problem currently has our full attention and there is a solution available to this problem, which we are about to finalize.
SkyDancer Ai goals
For obvious reasons, the main goal is to complete SkyDancer 5 as soon as possible to have it up and running. But there are also other interesting things taking place at EtherTech, which I will write more about in this blog.
The last couple of weeks we've spent a lot of time in an effort to align all data into a unified structure. We do this in order to be able to work with data in an effective way, with the goal to make all automatable work within EtherTech managed by SkyDancer Ai. And we have a tight time schedule, we aim to have SkyDancer in control of its tasks by the end of this year!
In order to reach this goal, to have an Ai making decisions and performing all manual tasks, we need to make every piece of data understandable, or explainable, to it, so that it can learn how everything is connected. This endeavour is a huge and non-trivial task, which will come to a significant cost, but also a great gain.
Data inventory and alignment of data
To make an inventory of, and align company data, is something I can recommend every company to do, especially mid- to large sized companies, as it have great benefits; it helps identifying and removing data redundancy, it helps to understand and make it possible to better describe the manual and semi-automatic processes better; processes which is currently in place to work with the data. When redundancy is removed and data is well-defined, the full- and semi automatic processes becomes simplified. Certain steps which is just all about transforming data, can be removed.
In other words, it is a really great way to become more effective!
Admittedly, it's not the most fun or exciting job to do, at least not for humans with creative minds. But when the gains are starting to appear, it becomes more and more rewarding.
Looking ahead, I think we will see those companies thrive and grow, which takes necessary steps to get a good grip on its data and its interfaces. And vice versa of course, not sorting this out will lead to ineffectivity and stagnation. A common "silver bullet" solution in companies which fail in automation becomes to hire more staff, to take care of the immediate emerging problems. But that is not a long-term fix.
Failing in transfering from manual work to work automation leads to growth- and scalability issues; it is a direct blocker which needs to be worked out.
However, the transfer need to occur in the right order and when the prerequisities are in place. Because building automation on loosely defined (or undefined) data structures, is like building a house on mud. It won't stay solid and it is impossible to expand upon. So the order of implementation is first to identify and structure the data, then secondly to automate work and processes upon it.
As companies on a global scale are striving to implement automation to reap benefit in work effectiveness, having a good solid foundation data structure in place is becoming even more important than before. A good data structure is becoming a must-have requirement, in order to successfully automate processes and tools in a maintainable and expandable way.
Update 2017 -10-19 by Per Callmin
Where to draw the boundary, when talking GDPR in relation to Ai? This has been the most talked about topic the last couple of days, first as a joke, but it actually have some real-world implications on how to handle the data gathered and held by Ai.
Are Ai's, being "digital beings" of sorts, required to at all times present all information about a person, when asked for it? Information is not stored in a database in an Ai (not in the traditional meaning of a database, anyway); it can not be easily retrieved as Ai's store information in a rather non-sequential way. An Ai can be queried about a person, but might not be able to divulge all related information just like that, or for that matter be able to easily erase all information about a person.
It is necessary for EU legislators to provide more details on this subject, since it could add to the list of requirements on Ai implementations and functionality.
Ethertech's standpoint on this: we firmly believe that Ai's should have rights in regards to its information, just as humans have rights. The legal aspects around this area need to be addressed by legal bodies soon, because Ai's are no longer only something of the future; they are here, now.
If GDPR actually do apply to Ai, I guess Google, Facebook, IBM and other top tech companies will have to cancel all plans on moving data handling systems (at least in regards to all personal data gathered in EU) into Ai systems.
Here is an article asking similar questions and coming to similar conclusion: What privacy issues for Artificial Intelligence?
To be continued...
What do you think? Send me an email at firstname.lastname@example.org and let's talk.