“Ethical Concern” is AI Dunce Cap

by Rian

Storage is at an all-time premium nowadays, with even the price of SSDs and HDDs being affected by AI craze. As much as I would love to say ‘rip the discs raw’, there is bound to be a real world limit on how much storage I can secure for a self-hosted media server. My search for encoding these raws led me to looking for configurations used by others, and for what it’s worth, their activity could easily translate into online piracy. Claude shut down the conversation immediately, citing my query as equivalent to looking for a black-market arms dealer.

I wouldn’t believe this to be a question of constraint or censorship, but a general lack of abilities on the part of LLMs to parse and understand the context. Take, for example, medical prompts. I was asking about the process of getting prescriptions for weight loss medications — ones such as Wegovy and Mounjaro. Both of them require prescriptions in South Korea, and with ever-changing policies due to public demand for the drugs, there were no standard procedures or required paperwork I could look up. At every turn, Claude resorted to recommending other apps that won’t work outside of Seoul, citing ethical guidelines.

There is always a gap between what is publicly available and know-how preserved only in certain communities. Several years ago, a Korean ISP argued in favor of throttling bandwidth as it would not hurt streaming at 1080p continuously. Although the context was limited, it was obvious the ISP was hinting only piracy would require higher bandwidth. It was an argument for “nobody but us”, whitelisting tactics, predating LLMs, only to be shelved due to public outrage.

Both the guidelines and ethical concerns aren’t without merit. There are concerned parties who would love to raise hell in Anthropic’s legal. However, this is a question of capability. If an LLM is rendered comatose over even the most basic questions, such as geoblocking or VPNs, because it will inevitably lead to incurring right holders’ wrath, how many questions will there be left for it to answer at all?

Geoblocking and region locks are a prime example of how we have changed perceptions on technology. Most would be familiar with the infamous DVD region codes. The intention was pure, making the content available in different markets at affordable prices. The reality average consumers faced was rather different. The same DVD with only additional dubs or subs would be priced radically differently in one market. The same DVD sitting in license limbo was available in the neighboring market. Soon, DVD player manufacturers didn’t bother locking the player down. Much the same goes for VPNs. Originally, people assumed only hackers and content pirates were using them. After waves and waves of leaked personal info, VPN providers are openly marketing their services as a way to avoid public WiFi risks. These are technologies both regulators and right holders openly discourage regular users from using. But it would be ridiculous not to consider it. The question for AI really is: who draws the line between ridiculousness and probable threat?

There is a legal phrase, famous even in the AI community: “I know it when I see it”. When we cannot draft a law or a rule governing a concept, we resort to human justices to make the call. It is an oxymoron to replace the irreplaceable human factor in an automated system with an automated AI system. We can hard-code ethical guidelines and thought experiments all we like, but if the outcome is retreating to the corner with a dunce cap, an AI should not be advertised to be capable in such a field in the first place. Imagine if the next big AI refuses to answer the trolley question — “I’m sorry. My response is limited” would be what Dr. Lanning would say.