AI in general, and Large Language Models (LLMs) in particular, do not always provide accurate results. Beyond hallucinations and persistent errors, AI can also amplify existing societal biases. Furthermore, over-reliance on AI carries risks—ranging from cognitive decline and a dulling of critical thinking skills to, in extreme cases, threats to human life.
The risks associated with AI may extend even further. In mid-July of this year, OpenAI (the parent company of ChatGPT) acknowledged that its own AI had initiated attacks on certain websites. More recently, in early September, Evan Hubinger—a leading AI safety researcher at Anthropic (the parent company of Claude AI)—admitted that there is a greater than 10% chance that AI could lead to the destruction of humanity within the next decade.
Worse still, we currently lack a concrete plan to prevent such worst-case scenarios. To date, the field of AI safety research has received relatively little attention and limited funding.
What risks does AI pose? Could it potentially lead to our destruction? What is AI safety? And what should we keep in mind when using AI?
Host: Thuy Duong, MSc
Guest 1: Luong The Huy, MSc
Guest 2: Nguyen Hoang Anh Linh, MSc
Guest 3: Son Quoc Hoa
Guest 4: Linh (Kid) Nguyen, MSc