Quote:
Originally Posted by roby_dk Machine learning and AI is currently a hot topic in my company, Ericsson. |
Good to see another colleague on the forum! Hope you are enjoying all the AI training!
Back on topic:
In practice automation historically, in general, has been used to perform human tasks, more effectively, efficiently, more safely and in certain cases also at a lower cost. Think robots producing (parts) of car, assembly and painting. Think an autopilot on a plane. Think a submersible welder robot. Or administrative tasks.
Whereas the cost associated with performing a certain task can be a major factor, truth be told, the other factors are often much more important long term.
Humans are not very good at repetitive, simple tasks. Nor are they particularly good at focussing at a given task for very long. The more complex or more repetitive the quicker errors slip in. That is just human nature A paint robot will paint a car in a certain manner, with a near 100% accuracy and consistency 24/7. No human or group of humans can match that.
So in many cases where automation has been introduced you tend to see quality, consistency of output etc. improving.
There are some theories that we as humans should concentrate on creative thinking and tasks only as that is where probably humans will always differ from machines.
A lot of these discussion around AI always end up comparing it to human intelligence. Whereas that is interesting debate and it always catches the news, it is not where AI is being used for to date. It’s main application are actually in line with how and where automation has always been used. To (help) perform task humans are not particularly good at.
In our Mobile Networks we are using AI/ML for instance to move away from traditional reactive/preventive maintenance to predictive maintenance. That requires ingestion and processing and (trend) analysing of vast amounts of data. And it needs to be done in relative short terms. Enter AI/ML, it is simply impossible for humans.
Mobile networks become increasingly more complex, more (complex) technology (e.g. 5G), more devices and the need to provide much more customer specific type of services. Traditional methods of operating networks simply can not scale up.
What is equally important and just as complex: How do you actually use AI in your organisation. In more cases than not, you can not just “bolt” on some AI.
In our case we have developed, ground up, a completely new way of delivering our services for telecommunication networks to our customers. New organisation, new (digitised) processes, new tools, new job roles and competencies, new commercial and pricing models etc.
AI in itself does not provide anything. You need to use AI in a complete eco system to deliver the capabillilies you require to deliver the required output. So, together with our customers we define that output and put the necessary capabillilies in place to deliver on them.
https://www.ericsson.com/en/managed-services
Whereas AI probably still has a long way to go, it is being used extensively already in many industries and applications. What it is not used for is to replicate a human, or human intelligence. That is more for the lab R&D environment.
You will find it in places/applications where humans simply won’t be able to provide the necessary output. Usually a combination of complexity and vast data sets that need analysing. In some case in (near) real time.
It is interesting to see that in the car industry (and some other industries) the label “hand made” tends to used in relation to high quality and exclusivity. Whether that is actually true might be in the eye of the beholder. Most likely if we look at it from a factual and measurable point of view it is most likely not!
Morgans are still, largely, hand made. What it means is if you have need to replace say a wind panel, you order a new panel. That new panel won’t fit exactly, because Morgans are hand made. And each panel needs to made fit specifically for a specific car.
Jeroen