The Dominance of Artificial Intelligence (AI)
Artificial Intelligence (AI) is the mantra of the current era. The phrase “AI” was first coined in the late 1950’s to refer to the heady aspiration of realizing in software and hardware an entity possessing human-level intelligence. Intoned by technologists, academicians, journalists and venture capitalists alike, there is a significant misinterpretation of the phrase that led to the err of humans. But this is not the classical case of the public not understanding the scientists — here the scientists are often as befuddled as the public. The idea that our era is somehow seeing the emergence of an intelligence in silicon that rivals our own entertains all of us — enthralling us and frightening us in equal measure.
According to AI experts, everything from our jobs, to the food we eat, to the sports we play, to the software we write will be affected. Artificial Intelligence is a broad term that encompasses machine learning and deep learning. The success of modern AI is less due to a breakthrough in new techniques and more due to the vast amount of data and computational power available. Importantly, even an infinite amount of data won’t give AI human-like intelligence — we need to make significant progress on developing artificial “general intelligence” techniques first. Some approaches to doing this involve building a computer model of the human brain — which we’re not even close to achieving.
Consider self-driving cars. For such technology to be realized, a range of engineering problems have to be addressed that may have little relationship to human competencies (or human lack-of-competencies). The state of encompassing cars that are driven by its own engine and control will be vastly more complex than the current air-traffic control system, specifically in its use of massive amounts of data and adaptive statistical modeling to inform fine-grained decisions. It is such an effort on the focus on human-imitative where AI may be seen as more of a distraction than traction.
This expectation that intelligent artifacts should by necessity be human-like artifacts blinded most of us to the important fact that we have been achieving AI for some time. As AI evolves, more of our personal data and information might prove to be at the mercy of the almighty technology that had engulfed the world in half of its sphere. A skilled adversary could fool the ML into giving the wrong output by carefully crafting a perturbation to what otherwise appears to be a legitimate input.
Papernot explained that “this type of attack can be generated using a process similar to the one used to train the ML system: that is, instead of computing the derivatives of the error of the system with respect to the training parameters (as one would to optimize a model), one can instead compute the derivatives of the error of the system with respect to the input itself. In this way, an adversary or researcher can systematically find perturbations for any input in order to trick any model they want to target. The general technique can work against ML in any type of application — for example, image recognition, audio transcribing, or malware detection.”
Also known as membership inference attack, attackers may also seek to compromise the confidentiality or privacy of training data. Done through careful observation of how different inputs affect a model’s predictions, if a particular model was overfitting to its training data set, it will be very sensitive to inputs similar to the training set’s outliers — these points can be inferred through careful testing.
Picture this. A robot that can win the best chess player in the world every time. In fact, a program has already been built for this way back in 2015.
AlphaGo Zero, later evolving into a generalized version called AlphaZero, is Deepmind’s AI program which learned the game of chess from scratch playing against itself and subsequently defeated the strongest chess engine stockfish. The closest a program has won AlphaZero is Stockfish, which was a result of +155 -6 =839 in favor of AlphaZero, just a few games lost to Stockfish out of the multiple games. It would be sheer luck if the strongest chess player can win 2 games against strong engines like Strongfish.
To err is human — is that why we fear machines that can be made to err less? — John Naughton
The way to beat AI?
Many of us fear AI. The illusion of AI dominance in the world is misleading to a certain extent that sparks fear in all of us, causing it to be tangible. We should never think of AI as a substance that will take over our lives. We should, in fact, learn to work with it. There may be certain things AI can do better than us, but they will never beat the way humans think and act.
I once attended an Artificial Intelligence Conference by Google. The talk was held by Toby Ruckert and Yiliang Zhao where they discussed the significance of AI and what the outlook of AI is. The most intriguing speech of the conference that left a deep impression on me is this —
Artificial Intelligence can beat humans in a rat race easily, no doubt. In the years to come, it will dominate the world, but we don’t know when that might be yet. It will be more intelligent than us, it will be more versatile than us, it will be more capable than us. But there is one thing that humans can beat Artificial Intelligence in, that is — Creativity.
Humans can find multiple unique solutions to a problem. AI can’t.
Will AI be taking over the whole world tomorrow? Definitely not.
Unless you choose to live remotely and never plan to interact with the modern world, your life will be significantly impacted by artificial intelligence. Instead of having a negative impact on society, AI will improve the way we live and creates a more diversified social well-being of ‘humans’.
So, humans. Instead of worrying about the negative impacts that AI can have on us, or forging a plan to take down AI, it is time we understand more about AI and how we can better work with it.
One piece of advice to you?
Hustle.
For a list of references that explains more about AI, check out Sam DeBrule writing on machine learning. Succinct references by him that will make you drool over AI. Or refer to the list of tools that have been provided by Liam Hänel ♛ because you definitely need a pair of utensils before you can drool over something ‘AI-licious’.
