Andrew Liu ’27
One-Line Summary: Considering AI’s many novel problems, how should the Federal Government control such a rapidly changing and innovative new field?
On September 13, 2023, Elon Musk and other big tech CEOs met in Congress to discuss the potential dangers of AI. Amid the technological craze of chatbots like Chat GPT and thousands of new AI business applications, these leaders were instead considering the existential impacts of AI on humanity. When prompted about AI’s effect on mankind, Musk responded, in a tone notably more serious than usual, “There is some chance that is above zero that AI will kill us all. I think it’s low. But if there’s some chance.” Although not every CEO in the room agreed on the extent of the action, by the end of that day, they had all agreed that the government must play a role in regulating AI.
As the only development of its nature, AI creates many unique societal problems, even if concerns regarding rogue AI and human apocalypse may be blown out of proportion. For instance, when AI gets mistakenly trained on false data, biases and inaccuracies will leak into results. In doing so, AI may contribute to the spread of misinformation. Another issue with AI has been its ability to create deep fake replications of people’s faces, a major risk to privacy rights worldwide. Copyright infringement between AI developers and journalism outlets has also been a problem with the technology that has recently appeared. By training on actual news articles, AI imprints upon real people’s work; when prompted, it will spit out that same information. Is the machine infringing on copyright, or is it simply “learning” from published material and reproducing what it comprehends? The choice is handed to the federal government, which has no precedent to work from. This is the most significant issue with regulating AI: the concept is so novel—a human endeavor not controlled by humans—that the moral line between right and wrong is blurred.
In its ability to mimic humans, AI can exhibit traits of reasoning, awareness of surroundings, and problem-solving to an almost eerie degree. Unlike search engines, AI creates answers off the cuff, building them through learned patterns. The differences between humans and AI are even more distorted when considering augmented AI, the combination of AI with a human user. Should AI be prosecuted like a human, considering its ability to “think?” From a different perspective, some argue that AI still isn’t human enough to be treated as its own entity. After all, the pattern-fitting nature of AI doesn’t count as a mind of its own. Chat GPT is only inputting one letter after the next using probability, not writing essays with emotion and reason. Its intelligence is a direct function of the data it is trained upon. If this is the case, is the data set instead to blame, or even the coder who assigns that set? Government regulators must toe the line with questions as moral as they are logical. Another difficulty with regulating AI is the pace at which the technology is growing. Many of AI’s new developments have only come to fruition in the past few years. Moreover, new advances are being made each day. In this fluid economy, regulation must keep pace or be left in the dust. Considering the recent pace of policymaking, with frequent deadlocks between parties, the political system may make this type of regulation infeasible.
Despite the challenges of regulating AI, the EU has been very proactive in passing new standards to limit the new technology. The EU has some of the most developed AI legislation, including far-reaching enforcement throughout one of the largest extended economies in the world. With a risk classification system, the EU AI Act will automatically ban AI that violates guidelines against biometric privacy, deep fakes, and other issues. With this umbrella protection, the EU hopes to eliminate AI’s challenges. However, critics have said that this strong approach might negatively affect the technology’s development. The compliance burden is said to be too high for small businesses to compete, stifling competition and ultimately hurting the industry. In fact, because harsh regulation introduces uncertainty around the future viability of new developments, it significantly lowers the levels of risk-taking in a market.
Currently, the US Federal Government has almost the exact opposite issue. With most regulation coming from existing laws that apply to AI, Congress has yet to pass AI-specific legislation on a comparable scale to the EU. Acts like the AI Disclosure Act of 2023 seek to reduce ambiguity between AI and human work, but these laws won’t directly change how AI collects or interprets data. By allowing companies more freedom to experiment, the US relies on a trust-based system where corporations, not Congress, call the shots. As a result, the AI market gains viability while consumers will feel the brunt of negative effects; loss of data privacy and informational bias being two examples.
Ultimately, the issue of federal AI regulation must balance innovation and safety. By cracking down on AI, the technology may never reach its true potential to help humanity. However, at the same time, without careful and regulated growth, AI may do more harm than it does good. As with every new technology, there comes a time when apparent value must be weighed with possible risks. For AI, that conversation may be more complicated than any other before.
Sources:
https://time.com/6554118/congress-ai-journalism-hearing/
https://www.americanbar.org/groups/journal/podcast/what-could-ai-regulation-in-the-us-look-like/
https://www.reuters.com/technology/musk-zuckerberg-gates-join-us-senators-ai-forum-2023-09-13/
https://mitsloan.mit.edu/ideas-made-to-matter/machine-learning-explained
https://www.imf.org/external/datamapper/PPPSH@WEO/EU/CHN/USA
https://hbr.org/2021/03/ai-should-augment-human-intelligence-not-replace-it
https://www.mckinsey.com/featured-insights/mckinsey-explainers/what-is-ai




Leave a Reply