INVITATION TO CYBERSECURITY 40 represents the forefront of capabilities and the height of complexity. As of this writing, AI is synonymous with machine learning and large language models that have given rise to generative AI technologies. Generative AI is computer technology capable of generating coherent media such as text, audio, images, and video. For example, intelligent chatbots speak English fluently and can provide expert answers to questions in seemingly every domain of human knowledge. They can also write computer programs, fiction, poetry, and much more. While not perfect, many believe it is inevitable that they will improve over time and will someday be practically flawless. Regardless of how good computers get at generative AI (and at any other type of AI), foundationally, they are still using Boolean logic to process binary input signals to produce binary output signals. Due to layers upon layers of abstraction and encapsulation and extreme amounts of processing, the relationship between inputs and outputs is impossible to trace. This gives rise to the appearance of true intelligence, but it is an illusion. In 1948 Claude Shannon anticipated generative AI when in his study of information theory he proved that languages have statistical properties. He demonstrated this by manually selecting random letters from a book. As he increased the complexity of his calculations he was able to randomly produce semi-coherent sentences. Due to the escalating amount of work involved in each step, he was not able to proceed further. What Shannon could not do at that time, computers today can due to the enormous amount of data available to them and their incredible processing power. The AI technologies that have produced generative AI are based on the statistical nature of information. While it is incomprehensible that coherent language can be produced based sheerly on statistical probabilities, that is their underlying program. This is not how humans process information, and this is easily verified by comparing the minimal amount of energy human brains consume to the vast amount of energy generative AI models consume. In short, AI in all of its forms, is still just a computer program. Human intelligence is needed to write it, feed it inputs, and evaluate its outputs. Its outputs are deemed good or bad based on human evaluation—that is the ultimate criteria proving that computers serve humans, not the other way around. We need to embrace AI as a tool that augments human intelligence. We are able to do more with AI technologies than we can without them—they multiply productivity. As for cybersecurity, AI is another technology that needs to be secured, the goal of secure AI, and it is also a technology that can help enhance cybersecurity, the goal of AI for cyber. Because AI will assist both cyber attackers and cyber defenders, it is not clear if it will change the overall balance of power in cybersecurity. Chapter 9 covers some ramifications of AI for cybersecurity. 2.5.2 Quantum Computing “I think I can safely say that nobody understands quantum mechanics.” - Richard Feynman
RkJQdWJsaXNoZXIy MTM4ODY=