Has Artificial Intelligence become Alchemy, and is this a good thing? Some experts argue that AI research may be the next step in the evolution of human intelligence. However, this debate has been raging for decades, with many people coming to the discussion with preconceived notions and prejudices. So let’s consider some of the evidence that points to AI as a new form of alchemy.
If we can create a machine with human intelligence, and then give it free rein, would that be a great thing? If so, what are the consequences? What kind of social systems would be impacted? For example, would a gold-mining company dominate the world? Or would it destabilize health care systems? And what about politics and religion systems? Could the ultimate worker or soldier be created? Such unknowable outcomes are scary, especially for those without a thorough understanding of how such systems function.
European leaders have long sought to keep their secrets, but alchemists knew that opening up would be dangerous. When gold became plentiful, manufacturers feared for their survival. Similarly, the technology giants use user data to train algorithms and develop money-making black boxes. As these companies gain wealth, regulatory responses are weak, and legal responses are missing. The emergence of new technologies means that we must protect our privacy.
Machine learning algorithms are the leading AI implementations in use today, including neural networks and deep learning. However, some researchers have criticized machine learning as alchemy. According to Google AI researcher Ali Rahimi, it’s all about trial and error and has little scientific foundation. This is due in part to publication bias and unsystematic use of trial and error. The problem, Rahimi says, lies in the fact that researchers don’t know why some algorithms work while others don’t.
While the debate is largely centered around how AI will affect the future of humanity, it’s also important to remember that this is a complex debate and not just a technological one. The debates over artificial intelligence (AI) have taken a bizarre turn recently, with articles in the Guardian and Science magazine comparing it to alchemy and magical thinking. This debate must be addressed before it becomes a hot topic for debate.
A major problem is that researchers don’t understand AI algorithms, and often fail to explain their methods. A recent study in Science showed that an algorithm with its complexity stripped away was more effective at translating, even though the researchers didn’t know what they were doing. As a result, this article summarizes recommendations from data scientists at Alphabet and the authors of the paper. It also highlights some steps for improved understanding of algorithm behavior.
Until now, the goal of artificial intelligence has been to create an algorithm that can apply its intelligence across domains. However, this goal isn’t even close to being achieved. Instead, there is no single theory that can explain the way we behave in any domain, and we are not even close to reaching that goal. In other words, there is no general theory of artificial intelligence that can explain the behaviour of human beings.
Fortunately, research at all levels is still possible. Today, machine-learning algorithms are widely available and can be used by anyone with basic computer science skills. However, despite the availability of a wide variety of machine learning algorithms and their configuration options, most researchers cannot fully explain the reasons behind the success or failure of one algorithm over another. As a result, AI remains a highly experimental discipline, and this makes it difficult to develop a definitive theory.
In a similar way, a similar phenomenon has happened in chemical engineering. In the 1990s, there was an explosion of AI publications. Instead of the rise and fall of the emergence of artificial intelligence, clustering algorithms were chosen over genetic algorithms. The difference between clustering algorithms and artificial neural networks is due in part to the difficulty of creating new algorithms and the lack of powerful computing. So, despite the rapid growth of AI, the chemical engineering community did not see much progress until the mid-2000s.