Artificial intelligence and machine learning have emerged as the big ideas in the last couple of years. For example, the financial services space saw the use of robotics and predictive intelligence in personal finance planning and broking advisory. The ARQ platform of Angel is based on artificial intelligence. So what exactly is artificial intelligence all about? Essentially, it is the ability to make machines think more like humans. In the current context, machines are extremely good at executing routine tasks. But artificial intelligence enables the machine to actually learn from experience and try to map the user needs, map the future requirements and proactively offer products and services accordingly. Artificial Intelligence has extensive applications in retail, financial services, medicine, business applications, public services etc. Let us understand how artificial intelligence will change the future and what the future of artificial intelligence is across the world and the implications for India.
Artificial intelligence predictions 2018..
We could see some 8 key developments in artificial intelligence (AI) in the coming year. Here are some key ideas..
The big idea in artificial intelligence is about neural networks. What do we understand by neural networks? It is all about mimicking the working of the brain. The brain is a hugely complex mechanism and has the ability to understand, evaluate and execute. The use of neural networks is also called “Deep Learning”. This can have huge applications in designing optimal networks in a variety of businesses.
Seamless processing of text, data, audio and video is still something that only a brain can do. The brain can simultaneously process pictures, words, incidents, visuals etc. Despite strides in AI, computers do not have the ability. In the current year we expect the principal of information bottleneck to apply wherein the computer will also be able to apply multiple sources of inputs to a single goal.
Deep reinforcement learning is another major stride that is expected in 2018. This is all about the ability of the machine to learn from the environment and actual apply to its offerings on the next occasion. For example, if an Amazon engine deciphers that John always comes back to buy books by a certain category of writers then the offering can be focused accordingly. Amazon is already using such AI tools but we could see a lot more progress this year.
Another major progress in AI is expected in the intelligent use of adversarial networks. What do we understand by adversarial networks? It is the process of putting good data and bad data and let both get processed. The beauty of machine learning or AI is that over a period of time both the sets of data tend to learn from each other and improve over time, improving the quality of output in the process.
The big challenge in AI is the availability of reliable historical data. For AI to be useful you need mountains of data but that may not be available in most fields. The answer is cross learning. For example, if you apply AI to military operations, can you also cross adapt the intelligence and adapt it to a typical business operation? This is called labelling of data to train the system and can have huge applications where historical data is limited. Expect big progress here too!
We all use probabilistic modelling in some form or the other. The farmer estimates rain, the gas station anticipates vehicle flow, the toll gate predicts fees. All these are based on probabilistic thinking. The next stage of AI is not only using probability but also tweaking the probability based on consistent experience. So you start off with a model with certain probabilities and the use of AI automatically fine tunes that model along the way. This is called probabilistic programming and is a big progress likely in 2018 in the field of AI.
Most of you who have worked with computers understand programming and need to modify programs based on feedback. We could see progress on AI that makes programming redundant. This is referred to as Automated Machine Learning. This is more of a decentralized model like the cloud which will democratize the process of machine learning and make it available and accessible even to smaller businesses.
Artificial intelligence is called black box technology because nobody really understands what goes on behind the screen. It is a lot like algorithmic trading where there is a lot of intelligence behind each decision but the user rarely understands what is the big process going on in the background. Year 2018 could see progress in what is called as Explainable AI. That will make AI more transparent and will also attract more people into the AI fold. In fact AI products in retail and financial services could become a lot easier to sell to customers once it becomes more transparent and understandable.
So, AI is likely to go through major shifts in 2018. Of course, it will be for the better!