Introduction
It is important to state that artificial intelligence (AI) has seen massive gains in the achievement of new developmental milestones as well as improvements in the past several years. The given persuasive essay seeks to showcase the real threat of AI to human civilization and why the general public is underestimating its dangerous elements. The problem is that AI is a serious and existential security risk, but people are unable to garner appropriate emotional responses toward it. If an AI is an information processing tool and it continues to improve regardless of the pace, it is destined to surpass human intelligence, which will lead to loss of control; thus, it is dangerous. The development of AI requires a pause for a year, during which the key AI developers and governments must create principles and frameworks to ensure that general AI does not emerge as an existential threat.
Overview
A proper discussion on AI requires a solid definition of the technology first. AI can be described as human intelligence that is artificially simulated; hence, its core characteristics include carrying out functions that humans are capable of, self-learning, and self-improvement (Shchitova, 2020). As of today, there are already narrow AIs at work, which are commonly used in self-driving cars, internet searches, and facial recognition (Conn par. 2).
Therefore, “AI outperforms humans, but only in low-level tasks,” but general AI or AGI will outperform humans in all domain (Wang par. 3). As a result, one can observe that AI is already deeply integrated into current platforms and instruments; hence, the competency overtaking is a question of when. It is also important to remember that AI risk does not come from it becoming actively malevolent and hostile towards humans. Its immense competence in its task can pose an indirect threat.
The Parallels Between AI and Human Intelligence Processing
The first argument is that intelligence essentially revolves around processing information in a manner akin to the human brain. Such an understanding positions narrow AI systems, already in widespread use, as fundamental exemplars of the concept. For instance, “AI systems are being used in the service of disinformation on the internet … from deepfake videos to online bots manipulating public discourse by feigning consensus and spreading fake news” (“SQ10. What” par. 9). These systems demonstrate an advanced level of intelligence by processing vast amounts of data to adjusting content that matches with individual preferences and behaviors.
Hence, it becomes evident that AI’s information-processing capabilities mirror human cognitive functions to a certain extent. The AI systems analyze, interpret, and respond to data in ways that showcase an evolving level of ‘intelligence,’ meaning they continuously learn from new data inputs, enhancing their efficiency and accuracy. The learning mechanism is currently limited to specific tasks or domains; however, it signals the potential trajectory toward more advanced AGI systems.
Furthermore, the current deployment of narrow AI in a multitude of sectors shows its integral role in modern society. From healthcare diagnostics to financial market predictions, AI’s applications demonstrate its versatility and potential to surpass human abilities in specific tasks (Rita par. 8). This reality anchors the premise that AI is on a path to eventually emulate and even exceed human intelligence through its information-processing prowess.
As a result, there is a need to open critical discussions about the ethical and safety considerations of such advancements, given the progression from narrow AI to AGI. The latter entails a transition from task-specific intelligence to a broader, more autonomous form. Such a transition raises questions about the control and regulation of AI systems, as there is a possibility that AI could make decisions independently of human oversight or ethical considerations. Thus, AGI needs to emerge with complete alignment with human interests and control mechanisms – there must be a pause to ensure the alignment.
Inevitable Advancements in AI Technology
The second argument is that AI will continue to develop because of its immense value to humans. It focuses on the relentless advancement of AI technology, encompassing both software and hardware. Humanity has a deep-seated reliance on and values technology, which acts as a primary catalyst for this continual improvement (Müller 53).
Without a deliberate pause, AI development will only stop due to cataclysmic, civilization-destroying events, such as global famines, pandemics, nuclear wars, or climate catastrophes. A significant aspect of this argument involves the analysis of AI’s role in current and future societal structures. AI technology does not exist in a vacuum; instead, it intertwines with economic, social, and political dimensions. The pursuit of technological enhancement, especially in AI, is deeply embedded in the fabric of modern civilization. In other words, there is much to be gained from powerful AI – a significant incentive that explains why the pursuit transcends other considerations, including potential risks and ethical dilemmas.
AI development is propelled by its inherent value to human society; hence, there is resilience and momentum for AI advancement against most conceivable challenges. It reflects a societal commitment to technological progress as a paramount objective by overlooking the potential for unintended consequences. Given that AI improvement will continue, there is a certain inevitability regarding AI’s evolution (Müller 74). As AI continues to advance, it becomes increasingly integrated into different facets of daily life and critical infrastructure. Such an integration enhances the efficiency and effectiveness of many processes but also increases reliance on AI systems.
In simpler terms, the current economic and societal frameworks exert a massive push towards AI development, and they lack a self-preservation mechanism for humanity’s safety and well-being. A company developing AI only incurs losses from halting work on the technology for safety reasons; due to competition, no developer will halt operations. Hence, there needs to be a global intervention that halts all development efforts until a safety framework and alignment are properly established.
Human Intelligence – Not the Pinnacle of the Spectrum
The third argument concerns intelligence itself and seeks to explain that human intelligence does not represent a limit or an upper ceiling. The statement aims to show that as soon as AGI surpasses the human intelligence mark, the competency gap will widen dramatically, leaving no time to implement safety measures (Müller 91). In other words, it will be “too late” by then to do any action in a meaningful manner; hence, at this moment, humanity is nearing a deadline by which the alignment can be achieved. The most unmistakable evidence that shows why human intelligence is not at the peak of the intelligence spectrum is already demonstrated by narrow AI that surpasses humans in specific tasks.
It is important to emphasize that there is a term for the moment when AGI surpasses human intelligence – the intelligence explosion. Evidence suggests that “at this point, which is estimated to occur between 2040 and 2070, it is hypothesized that an AGI will have the capability to recursively self-improve by creating more intelligent versions of itself” (McLean et al. 649). It marks the event in which AGI will supersede human competency and self-improve at an exponential rate, which will be incomprehensible and inconceivable for human beings.
In addition, one should be aware that human cognitive abilities have significant limitations, such as susceptibility to optical illusions and poor memory retrieval. Brain biology also puts human intelligence at a disadvantage because neurons transmit signals much more slowly than electrical wires (Müller 68). Therefore, an equal AGI will learn in a couple of days what a human will require for years. This shows why the explosion will be massive and swift; hence, any safety concerns must be raised and addressed before this moment. Given there is no incentive to implement proper safety protocols, they must be implemented through a universal pause.
The Challenge of Controlling AI Pre-Intelligence Explosion
The final argument concerns the current systems and structures that not only fail to incentivize safety over fast-tracked development but also actively sabotage it. Capitalism is about competition; hence, the existing framework is a “winner takes all” game, which even fails to ensure that the winner will not suffer from their own creation. It is stated that “in the classic arms race, a party could always theoretically get ahead and win.
But with AI, the winner may be advanced AI itself. This can make rushing the losing move” (Grace par. 6). The latter statement shows that AI cannot be regarded as nuclear weapons or other form of technology – there is no winner at the end. Experts are already calling for a six-month pause on the training of AI systems more powerful than GPT-4″ (Romo par. 5). In other words, there is an acknowledgment among leading developers that AI development will not self-correct for safety purposes.
Even more troubling is the fact that the existing leadership in the AI domain are already at odds with each other – there is no cohesion and unity among them. Sam Altman is the CEO of OpenAI, which is a leading AI development company (O’Brien par. 3). Recently, he was fired from the company by the board but was reinstated later due to either a power struggle or safety concern (Vallance et al. par. 1). Regardless of the actual cause of the drama, it showcases two possibilities, both of which are deeply problematic.
One explanation is that if there was a power struggle over the management and leadership of the company, it reveals that those who are supposed to develop AI with carefulness are engaged in pursuing selfish interests. Hence, they cannot be relied upon for developing AI responsibly. The second possibility is that some members of the board were concerned about safety and were not pleased by Altman’s commercially inclined leadership, which also made the whole situation troubling. The latter means that commercial interests have won since Altman was reinstated as CEO (Vallance et al. par. 13).
Both possibilities indicate that even AI leaders themselves are incapable of developing AI safely; thus, pause is necessary. The lack of adequate control mechanisms is not merely a technological issue but also reflects broader societal and ethical challenges. It involves questions about who is responsible for AI development and how to balance innovation with safety. The rapid advancement of AI necessitates a proactive approach to risk management, yet the evidence suggests a reactive and fragmented response from both the industry and regulatory bodies.
Conclusion
In sum, a year-long pause on AI development is mandatory to ensure that humanity’s future is safe from AI, during which the alignment of AI and human interests must be established. Firstly, intelligence is a matter of processing information, and the artificial version of it is a mere simulation of the human brain. Secondly, AI is highly valuable for humans, which is why it will continue to be developed and improved with certainty. The only way it will halt is if there is a global catastrophe or active intervention.
Thirdly, human intellect is not an upper limit of how intelligent an entity can be; hence, the ‘explosion’ will be dramatic and swift. Lastly, the existing framework promotes competition, selfish power struggle, and disregard for safety. The latter means that there is no cohesion and unity among them. If all these premises are true, which they are, then it means that the only possible solution to the problem is global intervention – a pause. One year should be enough to make sufficient progress to ensure security.
Works Cited
Conn, Ariel. “Benefits & Risks of Artificial Intelligence.” Future of Life Institute, 2015.
Grace, Katja. “AI Is Not an Arms Race.” Time, 2023.
McLean, Scott, et al. “The Risks Associated with Artificial General Intelligence: A Systematic Review.” Journal of Experimental & Theoretical Artificial Intelligence, vol. 35, no. 5, 2021, pp. 649-663.
Müller, Vincent C. Risks of Artificial Intelligence. Chapman & Hall, 2020.
O’Brien, Matt. “Artificial Intelligence Raises Risk of Extinction, Experts Warn.” PBS News Hour, 2023.
Rita. “Why AI Could Be a Threat to Humanity.” Surfactants, 2023.
Romo, Vanessa. “Leading Experts Warn of a Risk of Extinction from AI.” NPR, 2023.
Shchitova, Alikhadzhieva. “Definition of Artificial Intelligence for Legal Regulation.” Proceedings of the 2nd International Scientific and Practical Conference on Digital Economy (ISCDE 2020), vol. 156, 2020, pp. 616-620.
“SQ10. What Are the Most Pressing Dangers of AI?” Stanford University, n.d..
Vallance, Chris, et al. “OpenAI Staff Demand Board Resign Over Sam Altman Sacking.” BBC News, 2023.
Wang, Xiantao. “Artificial Intelligence and the Loss of Humanity.” Berkeley Political Review, 2020.