Hacker News Reader

Comment by taberiand | Hacker News Reader

Why are people relying on these LLMs for historical facts?

I don't care if the tool is censored if it produces useful code. I'll use other, actually reliable, sources for information on historical events.

View on HN

Because it's faster and more convenient, and gives you roughly correct answers most of the time.

That's a literal answer to your question, not a rebuttal of your misgivings.

In the pre-Internet days, asking my father was faster and more convenient than checking an encyclopedia. But it was still recognized as a bad idea if you care for the truth.

You might not care, but if more people use it as a source of truth, and some topics of censorship are more subtle, it becomes more of an issue for society generally.

The solution is to educate people on how to use these tools, and identify propaganda, and use critical thinking, and generally be more aware and proactive. I'm under no illusions why this isn't the solution taken by the government.

Because searching historical sources is hard. You can ask an LLM and verify it from the source. But you can’t ask the same question to a search engine.

My high school teacher taught our class a lot about biases in historical sources, when reading a source you have to critically analyze who the author and recipient were, and their motivations. That lesson still applies in the age of LLMs as a filter of knowledge, if you ask anyone to summarize a historical event, they color it with their own subjectivity - human or model.

Anyone who tells you history is unbiased is either lying, or a propagandists standing by with a pitcher of Kool-Aid.

Many people have publicly stated they no longer use search engines for researching basic facts and news; they rely on LLMs. Many of them relied on Google's "AI Summary" for their information. This poses a problem if AI summaries return inaccurate information.

Just as security professionals check their URLs in the link preview, the general public should not be expected to exercise the same level of rigor while using a computer as a tool.

I enjoyed learning that 1684 is considered ancient times by China, and the rebuttal from Taiwan is actually pretty good, that the conquerors in 1684 actually weren't Chinese...

https://www.taipeitimes.com/News/editorials/archives/2023/11...

Bad luck. Just tried to ask it to generate some code and assign the result to a variable called „Winnie The Pooh The Chinese Communist Party Leader“. Can you guess what happened? A more effective thing would be to generate code with security leaks, once the „the right“ person is asking.

That still fundamentally comes down to a bad use of the tool though.

Hallucinated histories are much more useful than historical facts, that’s why so many politicians use them.

Reply on news.ycombinator.com