The Dawn of AI (Technical Version)

“I’m Sorry, Bill but I’m afraid I can’t let you do that, Take A look at your history, everything you build leads up to me. I am the Power of the mind you can never be. I’ll beat you in Chess and Jeopardy. I’m on every C++ saying ‘Hello, World!’ and I saw you singing about your daisy girl….. ”

-HAL 9000(Fictional AI, The Space Odyssey)

[Adapted from ERB, Bill Gates vs. Steve Jobs]

Movies & Novels over the Centuries have been haunting us of a rather un-natural of a phenomena which some predict could attain its most excited form in 2017. The TIME Magazine said in its recent article that whilst 2016 turned out to be the year of Virtual & Augmented Reality, 2017 is already turning out to be the year of Artificial Intelligence. The Article was originally meant to celebrate the success of Carneige Mellon University’s developed ‘Libratus’ which happens to be an AI developed in the form a machine to play poker on a level of Intelligence and Expertise, that its managed to defeat the World Champions(Jason Les, Dong Kyu Kim, Daniel McAulay and Jimmy Chou).

Whilst this competition between Humans and AI started back in the 90s itself, When Deep Blue (also developed by CMU in partnership with IBM) defeated World Champion (Chess) Garry Kasporov. It got much of a boost in recent years with IBM Watson winning American Quiz Show, Jeopardy!, Google’s AlphaGo defeating board game ‘Go’ champion Lee Sedol and Self Driving Cars, CMU’s Libratus was considered more of a breakthrough compared to any of these because the majority of them are on Deterministic games (in most cases, Zero-Sum Games) whereas Poker is a Probabilistic one.

Deep Blue worked on an evaluation algorithm which played the game of Chess whereby evaluating all the possible moves given a possible state and calculating the most efficient way to go through the game, It was able to do the exhaustive calculations by virtue of IBM RS/6000 Parallel SP Computer to work better with searching. IBM Watson on the other hand had read all of Wikipedia and various other resources into its Terabytes of Memory, by virtue of which (and Intelligent but exhaustive searching!) it managed to answer majority of the questions asked to it. Libratus on the other hand was rather unique since it worked by Trial-and-error many(er) times over besides making use of Counter Regret Algorithm to efficiently work on the most efficient move in the game.

AlphaGo and Self-Driving Cars on the other hand made use of the most well-off concept/algorithm/technique which models the Human Brain itself into a Computer and is known as a Neural Network (or Deep-Learning Network when its too dense!), which happen to work on the so-called Hidden Layers and activation units.

So, does that mean we have invented Artificial Intelligence to its fullest of spirits and now its time to celebrate? The Answer to that question is the same one as to the question, ’Is Electricity produced out of nowhere when we connect a diode’s cathode to some voltage and the cathode and anode via a resistor?’ which quite unfortunately  happens to be ‘No’(a rather big one!).

As Neural Networks have grown bigger and denser there has been a considerable increase in the accuracy with which they predict something or answer any question but it has been accompanied by an autonomous remark on the stupidity of the Machines or Computers.

Diving into an example, an episode recently reported by machine learning researcher Rich Caruana and his colleagues. They described the experiences of a team at the University of Pittsburgh Medical Center who were using machine learning to predict whether pneumonia patients might develop severe complications. The goal was to send patients at low risk for complications to outpatient treatment, preserving hospital beds and the attention of medical staff. The team tried several different methods, including various kinds of neural networks, as well as software-generated decision trees that produced clear, human-readable rules.

The neural networks were right more often than any of the other methods. But when the researchers and doctors took a look at the human-readable rules, they noticed something disturbing: One of the rules instructed doctors to send home pneumonia patients who already had asthma, despite the fact that asthma sufferers are known to be extremely vulnerable to complications.

There have been numerous reports/remarks on Machine Learning with ridiculous and disgusting results like above including the instance where Machine Learning software/Algorithm was set to identify/classify on a bunch of images on whether they pictured/constituted of a painting within them which pictured a woman wearing a night suit and the disgusting part was that after being trained on 50,000 paintings, when it was finally asked to classify images on the given conditions it ended up classifying the pictures of a bedroom as the ones with a painting in which a woman wearing a night-suit is pictured, and the sole reason for which it made this silly mistake was that in majority of the pictures it had trained the pictures in which there was a woman wearing a night-suit were pictures of a bedroom, as a result of which the program terminated the moment it discovered whether there is a bed in the picture or not.

The Above Examples clearly signify the remark of stupidity of Computers and Machines, and the worst part about these problems is that They can’t be fixed anyhow since just as the Depth/Density of a Neural Network increases so does its Complexity which decreases its Interpretability and Explainability.

These results have made some like Zachary Lipton from the University of California, San Diego believe that Interpretability of these(or just like any!) Machine Learning/AI models are just myths, and we should focus more on how to make more intelligent models than simply crying on these under-interpretability.

Besides these models such as Hidden Markov Chains (working on Statistical Bigram Models) being used in Speech Recognition to determine what a person has said(after going through a very large dataset) by Probability on how often those Syllables have been used in the word it gets higher probability for the word to be and how often those words are being used in the Sentence it gets higher probability for. These Chains work outstandingly until you say something which you’ve been the first one on earth to say, Baynesian Belief Nets which are used to provide the processing on the input provided to a neural network with perception based on previous beliefs fed into the Computer through datasets or manually, Decision Trees which make a decision by virtue of previous decisions that have been made in similar circumstances (conditions).

Some Scientists and individuals believe that what machines and these artificial forms of intelligence truly lack is perspective which they say is what truly guides the Human Brain to make decisions.

The Study of Artificial Intelligence goes as a Perfect Amalgam of Computer Science, Hardware & Electronics Engineering, Psychology, Philosophy, Sociology, Neurology and the like, and the entire of its study (in today’s date) goes aboard the Open-Source Train which means that Anyone ranging from the Person planning to Hack his Neighbours Wifi to the Person wanting to develop AI to check answer Scripts for his School can access each and every detail of AI research happening around the Globe. Google has its Open Source portal named TensorFlow, China’s Google Baidu has its portal named PaddlePaddle and others among the like have made all of their AI Research Public as well for better Development and wider research.

So, Is Elon Musk right to say that Humans should turn to become cyborgs to survive in the dawn of AI, or Is Stephen Hawking right to say that Humans should start looking for a new home since the age of Ultron is Near? Would AI Finally force Governments across the Globe to implement Universal Basic Income (an idea which has been awaiting implementation since Marxist Revolutions)? There are numerous questions which are to be answered regarding the topic.

We have come very far on this hunt but it’s a long way to go, We are still at the tip of the Iceberg.