_edited.jpg)
AI, otherwise known as artificial intelligence, is a set of technologies a computer uses that enable it to perform various tasks, including the ability to see, understand, and translate spoken and written language, analyze data, make recommendations, etc.
As technology has continued to advance rapidly throughout the decade, the topic of Artificial Intelligence has become of utmost importance. Most people have likely heard it during class, with teachers telling them not to use it on assignments, but AI is so much more than a tool students use to cheat.
While AI can be useful and beautifully complex, the possibility that AI can grow to outsmart its creators, humanity, is a threat that this technology poses and doesn’t help with humanity's already existing fear of a world taken over by machines.
This year, the Nobel Peace Prize was awarded to two scientists, John Hopfield, a professor emeritus at Princeton University, and Geoffery Hinton, the so-called “Godfather” of AI and machine learning. It was awarded to them for creating a structure that can store and reconstruct information and a method that can independently discover data properties. This is necessary for the extensive artificial neural network that AI uses nowadays.
Even though both of these scientists received praise and a prize, including 1 million dollars, both scientists have expressed their unease regarding the use of AI. Hinton even cautioned that even though this technology could revolutionize subjects like healthcare, he is “worried that the overall consequence of this might be systems more intelligent than us that eventually take control.”
Even though it’s good to approach AI cautiously, it’s essential to recognize the significance of these developments within the scientific world. Scientists have been able to spot patterns that, without this advancement, they wouldn’t have been able to find. AI is changing the course of finding exoplanets for the better, and there are many more examples.
The following important question is how these scientists even achieved such a grandiose achievement. The technology developed by Hinton and Hopfield was modeled after the human brain, building artificial neural networks as computer simulations. In 1982, Hopfield translated the process known as associative memory, which is used inside the brain to find things stored in our memory, into a machine. This led to the Hopfield network, a procedure within a computer that can store patterns and recreate them. Suppose the computer is given incomplete information or a slightly distorted pattern. In that case, it can find the stored pattern most similar to the incomplete or distorted one and use it to reconstruct it. This structure was thought out thanks to Hopfield’s previous knowledge of molecular biology and neuroscience.
Furthermore, Hopfield’s system has nodes joined by connections, which resemble the brain’s neurons and the synapses between them. Hopfield discovered that you could train these artificial neural networks by making the connections between nodes weaker or stronger, just like you would train neural connections in the brain. He described the state of the network as equivalent to the energy spin system found in physics: the energy is calculated by a formula that uses all the values of the nodes and all the strengths of the connections between them. The Hopfield system is programmed by an image fed to the nodes, which are given the value of black (0) or white (1). The network connections are later adjusted using the formula so that the image is low energy. Later, when the second, more distorted image is fed, it uses the pattern previously saved by the computer to see whether the node has lower energy if its value changes. If the computer finds that the energy has been reduced, a black node is now white, or vice versa, the node changes color. Using this method, the computer keeps going until it can’t find more improvements and often produces an exact replica of the original image.
This advancement in itself is impressive, but there is also the topic of interpreting this data. In 1985, Hinton further developed the Hopfields system to create the Boltzmann Machine, which used his knowledge of statistical physics. Hilton named his system after Ludwig Boltzmann, a physicist who created an equation that describes how the probability of the states of individual components within a system depends on the amount of energy available. Hinton applied this equation to the system by creating visible nodes that receive information and hidden nodes that form a hidden layer. Both nodes and corresponding connections contribute to the system as a whole. The machine updates node values one at a time, leading to pattern changes while maintaining the network's overall properties. Each pattern's probability is based on the network's energy, as defined by Boltzmann's equation. When the machine halts, it generates a new pattern, illustrating its role as an early generative model. Continuing, this version of AI learned from examples rather than instructions. The more a specific pattern showed up, the higher the probability that it was correct.
Of course, this early system of the Boltzmann Machine is ineffective, but Hinton has continued to work on it and tune it to be more efficient. It is clear that thanks to the scientific work from the 1980s and onward, Hinton and Hopfield have helped lay the foundation for machine learning. While this technology is still under development and will likely continue progressing as humanity advances, these scientists lay the necessary groundwork to make this a possibility. Thanks to them, AI is where it is currently. It is unsure where AI will end up, scientifically and socially. Still, it's essential to recognize where it comes from, something that the Nobel Peace Prize effectively does with this year's nomination.
Comments