> > Yes, here is the link to Alexander Hanff's obituary on The Guardian's website: https://www.theguardian.com/technology/2019/apr/22/alexander...
> Now ChatGPT was outputting fake links to articles that have never existed - why?
> [...] then it doubled down on the lie, and next fabricated fake URLs to supposed obituaries to support its previous response, but why? Self preservation
To quote the FAQ page: "ChatGPT has no external capabilities and cannot complete lookups. This means that it cannot access the internet, search engines, databases, or any other sources of information outside of its own model. It cannot verify facts, provide references, or perform calculations or translations." If you ask it for a link or some specific information that didn't appear often enough to get memorized, it'll give a plausible guess as to what that would be.
There's a problem if people are expecting it to be an oracle of absolute truth. You do get a pop-up warning on first use and a "Limitations" list at the top of each new chat, but it's easy enough for people to miss/ignore.
> Imagine had one of my children or other family members gone to Microsoft’s Bing implementation of ChatGPT and asked it about me and had received the same response?
Bing's chat mode can look up sources and is instructed to only use information from those sources. Not that it's perfect, but it is at least designed for this type of problem and lists the real sources for you to check.
> For example, under such a definition it would be perfectly fine for an AI to kill a person infected with a serious infectious virus as that would be considered as minimizing the harm. But would we, as a civilized society, accept that killing one person in this situation would be a simple case of the ends justifies the means? One would hope not as civilized societies take the position that all lives are equal and we all have the right to life - in fact it is enshrined in our international and national laws as one of our human rights.
If the implied situation is that killing one person is the only way to stop the deaths of many people (however contrived the scenario has to be to allow only these two choices), I think many people would accept the killing of one person. "All lives are equal" doesn't mean one life is equal to 1000 lives.