ChatGPT burst onto the scene, quickly amassing more than 100 million users in mere months. This achievement was awe-inspiring given that ChatGPT was the first general-purpose AI application ever to be used for the same tasks that only humans could before – things like research, providing legal advice, giving parenting tips, and even digital companionship. Realizing the massive potential of these advanced AI technologies, and wanting to ensure their responsible implementation, it became imperative to act quickly. Without full US control of the AI field, it is likely that other entities might move in to steer the development of AI in ways that do not align with our ethical standards or our best wishes.
The technology behind ChatGPT is called a Large Language Model (LLM). LLMs make use of deep learning algorithms and neural networks, allowing them to make good use of language to reach conclusions. They can be quite powerful, sometimes even demonstrating SAT test scores in the 90th percentile range, but they are not foolproof. Sometimes, LLMs are tricked into providing inaccurate or biased information and can even fabricate spurious results. Research has shown that a version of ChatGPT can be tricked into giving bomb-making advice, highlighting the flaws inherent in these technologies.
To keep up with the tremendous pace of AI growth, tech giants like Google and Microsoft, as well as Chinese companies like Baidu, Tencent, and Alibaba, all have their own AI products in some stage of development. These firms and countries compete for global AI supremacy, but their efforts have left governments and institutions feeling unprepared to cope with the ever-evolving landscape that is AI. The fear of revolutionary technology being implemented too quickly led a number of technologists, academics, and organizations to call for a pause on the rollout of advanced AI, giving everyone time to consider the implications and build a framework for safe, responsible AI use.
Rather than drop the brakes on US AI development, many suggest the allocation of resources towards the safer use of AI. For example, technologists need data, algorithms, and response validation methods that are up to date and high quality. They could also put more energy into distinguishing inappropriate queries that were not meant to be a part of the system’s function. Forms of regulation have also been proposed, though considering their potential for misapplication, their introduction must be done with great care.
Recently, some of the brightest minds in AI signed an open letter that called for a six-month pause in the implementation of new AI technology. This debate over the speed of AI deployment should not dominate the conversation, but rather, how to replace existing AI applications with safer, more responsible versions. To that end, collaboration between government, industry, and academia is essential to producing an AI ecosystem that is safe and facilitates further AI advancement. The Government has the ability to curate experts to advise on AI safety, use the National Academies to inform policy, and back research and development projects that explore safe AI use. Private industry can create consortiums that foster safety practices and develop AI ratings agencies. Academics can create AI research centers and appropriate K-12 curriculums, while universities can increase advanced AI degrees.
It’s time for everyone to come together and make safe AI a reality. If the US is to lead the way in ethical AI implementation and use, it needs to start making moves today. The world is counting on us.
