In April of 2021, the European Commission proposed a general framework for regulatory laws governing the development and use of Artificial Intelligence (AI). Subsequently, in June of 2023, the Members of the European Parliament (MEP) adopted a preliminary version of the AI Act. This act places restrictions and, in certain cases, outright bans on specific AI systems, depending on the level of risk associated with their use. Among the practices prohibited are social scoring, real-time biometric identification, predictive policing, and indiscriminate scraping of facial images from the Internet. Notably, high-risk applications, such as the recommender systems used by social media platforms, will face stringent transparency requirements, requiring the disclosure of AI-generated content. The AI Act encourages testing AI systems within regulatory “sandboxes” (isolated, controlled, real-life environments) before their deployment within the EU community. Furthermore, it allows room for exceptions, and it safeguards the right of citizens to file complaints. At the time of composing this case, the European Council is deliberating the final form of the Act.
It is widely acknowledged that the evolving capabilities of AI will probably be disruptive. Advances in automation will profoundly affect certain sectors of the job market. We must anticipate many unintended consequences from flawed programming, hidden biases, or interactions with complex environments. We must anticipate potential misuse of AI by bad actors and rogue states. However, from the beginning of AI development, a handful of people have warned about a possible existential threat to humanity due to the very nature of AI.
It seems inevitable that a “superintelligent” system will eventually emerge—an AI system that surpasses human intelligence in virtually every aspect. While many computer programs excel in specific tasks, such as playing chess, they do not qualify as superintelligent because their abilities do not go beyond that task. A superintelligent system, often referred to as an Artificial General Intelligence (AGI), would have a generalized intelligence superior to humans in virtually all areas. The greatest existential threat to humanity is often associated with the possibility of AGI, and it is linked to the notion of a “technology singularity,” a point at which technological development becomes uncontrolled and irreversible. A superintelligent system, unlike the fictional HAL 9000, would presumably outsmart any human attempts to control it. In the words of I. J. Good, “the first ultraintelligent machine is the last invention that man need ever make, provided that the machine is docile enough to tell us how to keep it under control.”
The competitive pressure within the AI development community are likely to push developers to innovate faster than safety measures can keep pace. Historically, governments are slow to respond, making it likely that laws will be enacted, modified, or repealed too late to achieve their intended results. Out of apparent concern over the monster they are creating, many AI developers are urging governments to impose controls on their own work to delay the arrival of the singularity.
One notable AI, GPT (“Generative Pretrained Transformer”), was developed by OpenAI, and uses deep learning to generate high-quality text. ChatGPT, a variant of GPT, primarily produces conversational text in a natural language but can also write programming code. As of July, 2023, ChatGPT is all the rage, but other AI models are piling up fast. Auto-GPT, for instance, can decompose complex tasks into subtasks. As expressed in Wikipedia, “Auto-GPT was used to create ChaosGPT , which, given the goal of destroying humanity, was not immediately successful in doing so.”
From the 2024 National Ethics Bowl