An image generated by ChatGPT

Introduction

With the rapid growth of AI, it has made significant contributions to various sectors such as healthcare, law, and education in modern society. However, this rapid development has also introduced a range of challenges that cannot easily be solved. Generative AI, in particular, which has recently made remarkable advancements, introduces complex issues that are challenging to address. Some of the major concerns posed by generative AI include the following:

Body

1. Privacy issues and data mining

Users may not be aware that generative AI usage can be utilized for data mining purposes, leading to potential privacy breaches and the unauthorized collection of personal information. For example, a 2023 paper by Milad Nasr^[1] showed that LLMs can be prompted into dumping private information such as email addresses and phone numbers.

2. Copyright concerns

Since generative AI collects a vast range of data, the chance of plagiarism increases. For example, the lawsuit^[2] The New York Times filed against OpenAI last year showed many examples in which GPT-4 re-created New York Times stories nearly identically. Also, the image generating AI, Midjourney has the plagiarism issues^[3] of which generating the copyrighted animation characters or scenes from movies identically. These issues are being addressed by AI developers, but remain unresolved.

3. Misinformation

Generative AI also has hallucination issues, which is a response generated by AI that contains false information presented as fact. This hallucination had a case^[4] of misleading lawyers in America to submit fake court citations which resulted in paying $5,000 fine.

4. Identity verification

The articles or images generated by AI are often indistinguishable with human-made works, which may lead to depreciation of human artwork. There actually have been an AI generated artwork which won a prize without notifying the source to the judges^[5].

5. Crime/Racism prevention

Biased data during the AI training might result in racist responses^[6]. Also, generative AI can provide information on illegal or harmful activities, such as ways to break into a house or build a bomb^[7]. Recently, crimes related to deepfake images generated by AI also have been a big issue in Korea.

Conclusion

If these issues are not addressed in a timely manner and AI development continues to accelerate unhandled, these challenges could escalate into uncontrollable problems in the future. To ensure that AI advances ethically and in a positive direction, it is essential to implement regulations aimed at guiding its development.

In fact, discussions on addressing these potential risks have already begun, as seen in events such as the AI Safety Summit^[8]. Furthermore, EU AI Act^[9] which is a European regulation on artificial intelligence, actively working to regulate AI applications which can cause unacceptable, or high risk to the society. By fostering a framework of ethical standards and regulations, we can steer the future of AI toward innovation that serves humanity while minimizing risks.