The Double-Edged Sword: Understanding AI's Impact and Challenges

This article is inspired by my conversation with Choak Visavayodhin on FaceThis article is inspired by my conversation with Choak Visavayodhin on Facebook (the post is in Thai, but using AI, you should be able to roughly understand what it's about). We were discussing our opinions and concerns regarding the development of Artificial Intelligence, particularly Large Language Models (LLMs).

Will AI replace humans?

When discussing the potential of AI to replace human jobs, it is crucial to differentiate between the types of AI in question. Narrow AI, designed for specialized tasks, has a real potential to supplant human roles, particularly in repetitive task environments. The risk of job displacement in these sectors is significantly high due to the efficiency and consistency of Narrow AI in performing such tasks.

However, the situation is different when it comes to Generative AI. With this technology, the workforce is being transformed rather than having human jobs directly replaced. Employers will always favour the most productive and creative workers, so those who are skilled at using Generative AI in the workplace are likely to outperform those who are not. This change emphasizes how crucial it is to become proficient with AI tools rather than letting them take center stage.

Challenges in commanding machines through LLMs

The phenomenon known as AI hallucination significantly affects the trust and dependability of controlling machines with Large Language Models. This concern is paramount as it impacts the accuracy and safety of automated responses.

However, based on my personal encounters (and fights) with digital assistants such as Google Assistant (now Gemini) and Siri, the use of LLMs to command machines should focus on improving their comprehension of human directives. We can address challenges caused by AI hallucinations and improve communication between humans and machine-directed tasks by not requiring AI to generate content from scratch and making sure it asks for the user's permission before executing instructions.

Generative AI's Capabilities and Misuse by Criminals:

The discussion on Generative AI becomes particularly intriguing when considering its capabilities and the associated risks, especially regarding criminal misuse. Generative AI's prowess lies in its ability to produce content, make decisions, and learn from vast amounts of data, which can be leveraged for innovation, creative outputs, and problem-solving across various sectors. However, this power also opens avenues for exploitation.

Criminals can harness Generative AI's capabilities to create sophisticated phishing schemes, generate fake yet convincing content, or automate fraudulent activities at an unprecedented scale. The underlying issue does not stem from the AI itself but from a lack of comprehensive AI literacy among the general populace. Without a clear understanding of Generative AI's workings, its strengths, and its vulnerabilities, individuals and organizations remain susceptible to manipulation and deceit.

This scenario, like the previous push for Digital Literacy, emphasizes the critical need for widespread AI literacy. Our comprehension of and safeguards against the possible abuse of Generative AI must advance along with its capabilities. Mitigating these risks and responsibly utilizing AI's potential requires educating the public about how to identify AI-generated content, comprehend AI's limitations, and promote ethical AI use.

Note: I wrote the Thai version of this post as a comment to Choak Visavayodhin 's Facebook post. After that, I used ChatGPT to assist me in translating, rewriting, and rephrasing. If needed, I then use QuillBot to further refine my work. This is a process called 'Draft & Craft', to borrow Choak's phrase. Instead of spending a couple of hours writing this article, I was able to finish it in about 20 minutes thanks to these AI tools.