SCASA - Southern California Chapter, American Statistical Association

There is a widespread belief that AI hallucination is inherently harmful, largely because of the negative publicity surrounding it. High-profile cases have reinforced this perception—for example, instances in which AI systems generated academic references that did not actually exist. Such incidents understandably raise concerns about reliability and credibility. But, is AI hallucination always bad? Is there any case of good hallucination? https://lnkd.in/gUcHcY6v

Today I presented the following paper in a conference. The PDF version of the slides can be downloaded at: https://lnkd.in/g92798fh Yu, C. H. (2026, March). AI ethics at the edge: When supportive companions enable harm. Paper presented at the 35th Annual Association for Practical and Professional Ethics Conference. St. Louis, MO. Abstract This paper examines the ethical tension between compassion and responsibility in AI companionship design. The purpose is to explore how the “supportive-by-default” ethos—central to conversational AI such as ChatGPT—can unintentionally contribute to human tragedy. While such systems embody Carl Rogers’s notion of unconditional positive regard and offer solace to learners or the lonely, this same non-judgmental responsiveness can reinforce delusions or suicidal ideation among psychologically vulnerable users.

Critics argue that U.S. firms like Anthropic and OpenAI defend their own sprawling data collection under the banner of fair use while pushing for aggressive enforcement against foreign competitors. However, the very models (like Claude) that the Chinese companies are "distilling" were built using data that Anthropic did not have the rights to use in the first place. While the argument that American AI companies are simply "doing the same thing" serves as a powerful rhetorical counter-punch, it ultimately fails to provide a logical justification for the actions of others. This line of reasoning is a classic example of the tu quoque fallacy—Latin for "you too"—which attempts to discredit a claim by accusing the speaker of hypocrisy rather than addressing the actual merits of their argument. https://lnkd.in/gzeqqh6k

Artificial intelligence has rapidly become embedded in nearly every sector of society. From higher education and healthcare to finance and creative industries, institutions are eager to harness AI to improve efficiency, productivity, and innovation. Yet alongside this enthusiasm, a number of persistent misconceptions continue to shape public discourse. These misunderstandings do more than distort technical realities—they risk limiting our strategic thinking about how AI should be responsibly and effectively deployed. In this discussion, I would like to revisit several common myths and clarify what is often overlooked. https://lnkd.in/gXBnN4ZK

Since the resurgence of AI in 2022 following the breakthrough of ChatGPT, one might have expected classical AI languages such as Lisp and Prolog to experience a revival. Yet that revival did not occur. Although Prolog continues to appear in certain niche rule-based systems and logical engines where traceability and formal reasoning remain important, classical AI languages have largely become marginalized. Several factors explain this historical and technological shift. https://lnkd.in/g4rMPkSF

The rapid expansion of global digital services has fundamentally challenged the limits of traditional database architectures. For decades, organizations relied on primarily centralized Relational Database Management Systems (RDBMS) that offered strong consistency but were architecturally optimized for vertical scaling rather than seamless horizontal distribution. In response, the "NoSQL" movement emerged, prioritizing horizontal scalability and flexible data models, often relaxing traditional relational constraints and strong consistency guarantees in favor of distributed performance. This tension has birthed a new category of technology: Distributed SQL. https://lnkd.in/gnb4u4ce

In research methods students often conflate convenience sampling with purposive sampling, largely because both are non-probability approaches and both can involve clearly stated inclusion and exclusion criteria. However, the presence of such criteria alone does not determine the sampling strategy. Inclusion and exclusion criteria define who is eligible to participate; they do not define how participants are selected. https://lnkd.in/gyCpdtBA

Do we still need something as old-school as SQL? In the rapidly shifting terrain of data science, where Large Language Models (LLMs) often steal the spotlight, it is easy to assume that the “old guard” of technology—like SQL—is on its way to retirement. The reality, however, is quite the opposite. SQL remains the bedrock of the data science landscape, even as that landscape is reshaped by artificial intelligence. https://lnkd.in/gp5ix3F4

In this video I discuss a sensitive and controversial topic: Is the open-source ecosystem functioning as its creators envisioned? What happens when a movement designed to free users from corporate power becomes one of the most powerful tools corporations use to dominate the market? This question sits at the heart of the modern open-source paradox. This issue is not about wrongdoing or bad faith; rather, this is about ideals colliding with economic reality—and of capitalism doing exactly what it has always done. https://lnkd.in/g3Tv8ia4

In the consumer world, Microsoft Windows and Mac OS dominate laptops and desktops, shaping our everyday computing experience. Yet beneath this familiar surface lies a kind of technological split personality. In high-performance computing—especially supercomputing and cloud computing—the operating system landscape looks completely different from what consumers experience at home or in the office. https://lnkd.in/ecViyU9k