MACHINE Learning is essentially high-speed evolutionary learning.
Computers perform well when a task requires repetition. Many things can be learned by using that technique. We all probably remember how multiplying numbers was difficult until you found out your way of understanding it – whether it is from the multiplication table or a specific logic attached to that.
Machine learning can replace most of the tasks that require repetition if we know well enough how to define them. Computers will then follow the instructions and learn.
However, the learning that happens is not about being innovative or creating new ideas.
It is rather learning all the details by trial and error.
When we think machine learning is fast it is actually only fast when it comes to time; it might not take many days or weeks but when we compare how many times a computer needs to go through a task using this trial and error method it is extremely slow.
So, in the end, the computer performs many iterations of the task given and finds out which steps actually matter or provide the best outcome, and it then can optimise the process based on that.
The exact same thing happens when you visit a blacksmith and watch the difference between an experienced master at work, and the young trainee – it is all in the details.
Machines Instead Of People?
Limits of machine learning is a hot topic and the debate often revolves around the question of whether machine learning will increase or decrease the number of jobs.
As mentioned, computers outperform humans when tasks require repetition. When a new employee joins a factory, someone needs to train that person. After the person learns and is qualified for the job there still needs to be a quality assurance department that notifies the management if things are not running smoothly. Computers will replace this and training is conducted by deploying a piece of software code.
This means, in the future, there will be more job design and definition work in all companies and once that has been done well computers will be deployed, and the people can shift their focus to the next task instead of supervising the process.
Machine learning only works if humans are educating themselves enough to discover what is the root cause of their challenge at hand and how to communicate the solution to a computer who doesn’t understand context.
Computers can innovate but to a very limited extent – basically to allow variance in the repetition pattern – and requiring boundary conditions given by the staff.
It will be a cheap way to clone the future workforce!
Mikko Niemela is the CEO of Singapore-based Cyber Intelligence House.
He is a panelist at the upcoming Keep It Going: the Nth revolutioN, which will focus on Machine Learning & Design.
If you would like to attend the session on 22 November 2018 from 2pm – 5pm email firstname.lastname@example.org. More details if you get a seat.
You May Also Want To Read:
At the recent the Nth revolutioN, the discussions on The Good, The Bad And The Ugly Of The Future covered things technological but also focused on the quality of life in light of the changes that are thrust upon society.
Read more articles from the Nth revolutioN at http://www.storm.sg/keep-it-going/