If you have a wealth of information but do nothing with it, then it is worth precisely nothing. Changes are coming. Now.
We have become extremely efficient in gathering information. Big data is a new frontier, burgeoning and expanding exponentially. It was estimated in 2012 that all the data in the World took up about three zettabytes (that’s three trillion gigabytes, folks), and it is predicted that by 2025, this will top 160 zettabytes, heading towards a yottabyte territory. Yeah, me neither.
We are not terribly efficient at processing it, though.
Estimates suggest that around a third of this big data could be useful, although I suppose that its usefulness depends on who is processing it and what their motivations are, but regardless, that’s an awful lot of information that is crying out to be interpreted and understood.
And herein lies the problem. We are terrific at gathering data, but not very good at converting it into knowledge; we are currently processing around 0.5% of this big data, and there is clearly something wrong with that picture.
The slightly peculiar thing is that whilst theoretically efficient, smart processing of this data will give rise to useful results, we are squarely in Donald Rumsfeld territory – the landscape is full of unknown unknowns. Famously, (and before the era of big data, apparently) Walmart know that there is a link between purchasing diapers and beer on a Friday night, but is that particularly useful to know?
And don’t even get me started on correlation and causation.
So, we find ourselves in a peculiar position. We are standing at the foot of Mount Bigdata, we don’t have much rope, and whilst we are pretty confident about the direction we should be heading in, we don’t actually know if we really want to go there anyway. We probably do, but it is inconceivable that we will reach the zenith by taking the quickest path. I think I have milked this metaphor to death. I was going to stick crampons in there at some point; you should count yourself lucky.
This brings us to on to the topic in hand : AI robotics.
I might need to explain this robotics thing. I’m sure we have all seen the remarkable (and slightly foreboding) progress that cutting-edge metal monsters are making, but they are a different type of animal (okay, machine) altogether.
AI robots do not exist tangibly; they are complex, state of the art software applications. Broadly speaking, they come in three distinct categories: script-based reproduction, supervised machine learning and unsupervised machine learning (which is where the really interesting stuff is going on).
A simple way to define the concept of robotics is a process that eliminates the need for manual repetitive, tedious tasks. That could mean a house bot doing the vacuuming and dishes (and perhaps it will, one day), but right now it is all about processing information.
I’ll give you a good example. Suppose that an employee spends one day a week working on spreadsheets. Suppose that they have to import data and then extrapolate salient results to be exported into another spreadsheet. It doesn’t really matter what the details are; this – or something like it – takes place in offices all across the globe every day.
A simple AI robot could be trained to undertake this task, and it would do so independently, in a fraction of the time it would take a person. And as well as the obvious advantage of saving that time, robots don’t get bored or distracted and make mistakes. They don’t get sick. They don’t have bad days. There is little doubt that machines are superior to us when it comes to many mundane, repetitive tasks.
The peculiar thing about this example is that spreadsheets are already a much quicker way of processing data. Before they came along, compiling figures was a clunky, time heavy task, subject to human error. The eight hours that this fictitious employee takes to arrive at a set of useful figures may have taken a week or so before spreadsheets were so ubiquitous.
Technology never stands still though, and undertaking necessary tasks as quickly and efficiently as possible is the raison d’etre of much technology. There are complex philosophical and moral considerations here, but it is clear that the application of AI robotics technology will cut down on running costs (e.g. wages), and that in order to compete and be profitable, a modern business needs to be at the cutting edge, all the time.
It doesn’t matter what anyone might think – we are on a path, and the pursuit of profits that dominates business and commerce means that we are stuck on it, for better or for worse.
And this is why AI robotics will become more prevalent and influential in the foreseeable future.
The spreadsheet case above was an example of script-based reproduction. It works a little like the macro function on a spreadsheet. This is not particularly clever tech. It isn’t self-regulating, and errors in the script stage will mean that it will fail, and will not know it has failed.
Supervised machine learning is a step up from this. We are getting into Siri and Alexa territory. Here, the robotic software can make decisions within a framework, using what is referred to as a trained set of data. If you were party to the coding that gives rise to the robot’s response, you could probably predict it. Probably. However, the rest of us mere mortals could not, which is why this sort of AI is becoming so prevalent. Have a look at this, for example. This is supervised machine learning incarnate. And yes, the prospect that I might have a conversation with a machine complete unknowingly is one that I am a little bit perturbed by.
The third level of AI robotics is very interesting. Unsupervised machine learning is at the very cutting edge of this technology, and pertains to the processing of big data and the mining of this data. The coding gives the robot the potential to explore and interpret the data as it sees fit. It will make connections and extrapolations, without being directed to do so.
These results will then be examined by humans, who will identify useful information and feed back to the robot about what it has produced, which means the process will constantly improve. An example of this is Google Translate, which is processing pertinent information all the time in its quest for perfection. It'll get there soon.
You must understand that we are talking about huge amounts of data. Huge. It is inconceivable that even the most exceptional of savants could begin to process it. What these unsupervised machines are doing is way, way beyond the scope of what we could achieve on our own. This technology is burgeoning, and we cannot predict where it is going to take us.
Corporations and large firms are switching on to how important data mining is. Budgets that didn’t exist two years ago now have at least six noughts in them, and this is just the beginning.
We’re i-Recruit. We’re new, and we understand the importance and growth of AI robotics.