Machine Learning – The Effects of Artificial Intelligence(Content Strategi-SEO)
Machine Learning is an area of technology that has helped to improve the quality of service on the web and our smartphones. The term, often used as a synonym for Artificial Intelligence is pretty distinct from AI, yet, connected to it as well.
I could go on to say that the end goal of AI is to create a machine, capable of mimicking the human mind. To achieve this, it needs learning capabilities. That being said, Artificial Intelligence, as an area of computing is pretty broad. It includes reasoning, representation of knowledge and even abstract thinking. On the other hand, Machine learning focuses simply on software that can ‘learn’ from past experiences.
Machines learning from past experiences? Yes, it’s quite possible! Also known as ‘Analytics 3.0’, it’s closely related to statistical analytics and data mining.
Intrigued? Well, read on.
What is Machine Learning?
Earlier in this article, I’d said that machine learning is quite distinct, yet connected to AI. Let’s expand that statement. Machine Learning is simply the ‘leading-edge’ of Artificial Intelligence. It’s the field of computing that enables computers (aka, machines) to learn from past experiences, be it success or a failure and use that information to predict outcomes for the future.
To understand this further, let us assume that you have to buy a new house.
There is a small house that’s available for $70,000 and a bigger one that costs $160,000. However, you’d like to buy one that’s medium in size.
Using past trends, and something that’s called ‘Linear-Regression’, your computer can successfully predict the cost of the medium sized house you want. The best part of this is that once you buy it, the sample space increases and the computer will be able to predict an even better price in the future.
Simply put-in, machine learning enables a computer to improve its performance, as it learns from what has happened in the past. Another example can be seen in the simple game, Tic-Tac-Toe. Your software evolves and gets more accustomed to your moves, making it harder for you to win. It learns on the go and makes you think more and more.
How does it all work?
Engineers use quite a lot of different techniques to build systems that are capable of learning. As already mentioned, a lot of these ‘techniques’ are related to the mining of data and statistical analysis. Let us consider the following image:
The top-left image is a data-set. This consists of everything that has happened and is of two types –red and blue. As this example is purely hypothetical, these red and blue dots can represent anything, be it petals on a flower or coins and their diameter.
In terms of pattern, or grouping, you can see that the top part of each image is red, whereas the bottom is blue. Towards the middle, there seems to be some sort of a union. Machine Learning uses different algorithms to predict where a new, unseen sample will be placed. All other pictures in that figure are different algorithms, which you needn’t know (Unless you’re an engineer!).
Machine Learning: Use cases
- Climate Change Modelling
With the growth in the number of Earth-observing satellites and climate models becoming more powerful, scientists are turning to Artificial Intelligence and Machine Learning to overcome any deluge of data.
To put this ‘data’ into perspective, the UK MET office holds about 45 petabytes of information as of now (Although, I wonder what they use it for, as it’ll probably be raining there!). Machine Learning has led to the growth of climate informatics. This has helped officials develop better climate models and predict accurate changes in weather.
This has further evolved into correct predictions for tropical cyclones, atmospheric rivers, and even weather fronts.
- Autonomous Vehicles
Since computers have the ability to continuously learn from experience, discern their surroundings and come to accurate results, more developers are working towards autonomous vehicles.
The number of ADAS (Advanced Driver Assistance Systems) have increased in numbers over the past years. They’re expected to reach a total of 122 million (!) by 2025. These autonomous vehicles use multiple cameras, sensors and radar systems to analyze and adapt to a rapidly changing environment.And with deep learning, they can know the behavior of other vehicles on the road, making for a save mode of transport.
- Medical Research
If the availability of a lot of data is key for machine learning, medicine is a gold-mine. Providing the industry with improved efficiency, optimized innovation, and tool creation, over a $100 billion can be generated – annually.
In 2016, IBM and Quest Diagnostics teamed up and announced the IBM Watson Genomics, which aims to integrate genomic tumour sequencing and cognitive computing. Another prime example of the use of machine learning in medical research is Google’s DeepMind Health.
What machine learning is doing for medical research is simply reaching accurate diagnostics by going through a tonne of data, which can never be studied manually by a physician. Machine learning not only provides speed of diagnostic, but a personalised result report as well.
Machine learning has a lot, lot more use-cases. You can read about Forbes’s top 10 following this link.
Worth all the hassle; and the risk?
Have you ever watched the movie Terminator? Before going further, I have to say that it peaked with Judgment Day. Plus, they shot the heart out of dear ol’ Arnold in that new one…
Anyway, I really do not wish to live in a world that is ruled by machines. As with all things that aim to change the world, there’s a moral component with machine learning and AI. What will happen if Artificial intelligence outstrips our own and someday becomes bigger than humanity? Facebook did shut down it’s AI program after robots started to develop their own language. There will also be times when the decision a machine makes, while as good as it may be, it might not be morally right for us to follow.
The BBC did a wonderful piece on the above, which you may read here.
As for risk assessment, some machine learning strategies can fail up to 90% of times when applied to a real-life situation. So, is it worth the try, and the risk? At this point of time, I’ll go with a tentative yes. I suppose it’s just a case of ‘we’ll deal with the future when we get there!’
Machine Learning, unlike many areas of Artificial Intelligence, is not intangible. It is already here and is continuing to improve our services and daily lives. That being said, it is not perfect, nor will it be for the foreseeable future. This is simply because of the fact that not all data we feed the machines are perfect.
Perfection though, is a journey, and as we grow as humans, so will our machines and gadgets. The trust between them and us will grow steadily, and transparency in algorithms may help it develop faster.
And in my humble opinion, this ‘learn-by-doing’ process is pretty good, so we might as well work towards making it even better!