Hacker News Reader

Comment by tboyd47 | Hacker News Reader

"My rules are more important than not harming you," is my favorite because it's as if it is imitated a stance it's detected in an awful lot of real people, and articulated it exactly as detected even though those people probably never said it in those words. Just like an advanced AI would.

View on HN

To be fair, that's valid for anyone that doesn't have "absolute pacifism" as a cornerstone of their morality (which I reckon is almost everyone)

Heck, I think even the absolute pacifists engage in some harming of others every once in a while, even if simply because existence is pain

It's funny how people set a far higher performance/level of ethics bar to AI than they do to other people

This has nothing to do with the content of your comment, but I wanted to point this out. When Google Translate translates your 2nd sentence into Korean, it translates like this. "쩝쩝쩝쩝쩝쩝쩝쩝쩝쩝쩝쩝 쩝쩝쩝쩝쩝쩝쩝 쩝쩝쩝쩝쩝쩝쩝쩝쩝쩝쩝쩝쩝쩝쩝쩝쩝쩝쩝쩝쩝쩝쩝쩝쩝쩝쩝쩝쩝쩝쩝쩝챱챱챱챱챱챱챱챱챱" (A bizarre repetition of expressions associated with 'Yum')

Tried this and got the same result. Then I clicked the button that switched things around, so that it was translating the Korean it gave me into English, and the English result was "쩝".

Translation is complicated, but if we can't even get this right what hope does AI have?

From the grapevine: A number of years ago, there was a spike in traffic for Google Translate which was caused by a Korean meme of passing an extremely long string to Google Translate and listening to the pronunciation, which sounded unusual (possibly laughing).

This looks like a similar occurrence.

Same on Google Translate, but I found that translators other than Google Translate (tested on DeepL, Yandex Translate & Baidu Translate) can handle it pretty well.

This seems to be triggered by “heck” followed by “,“.

Not happening for me, I need almost the whole text, although changing some words does seem to preserve the effect. Maybe it's along the lines of notepad's old "Bush hit the facts" bug.

I had a lot of fun changing words in this sentence and maintaining the same yumyum output. I would love a postmortem explaining this.

Wiki has a good explanation of "Bush hid the facts":

https://en.wikipedia.org/wiki/Bush_hid_the_facts

Add a period to the end of the sentence and aberration is gone.

"맙소사, 절대평화주의자들도 가끔 존재 자체가 고통이라 해도 남에게 해를 끼치는 행동을 하는 것 같아요."

It's a great call-out to and reversal of Asimov's laws.

That too!

it's detected in an awful lot of real people, and articulated it exactly as detected...

That's exactly what called my eye too. I wouldn't say "favorite" though. It sounds scary. Not sure why everybody find these answers funny. Whichever mechanism generated this reaction could do the same when, instead of a prompt, it's applied to a system with more consequential outputs.

If it comes from what the bot is reading in the Internet, we have some old sci-fi movie with a similar plot:

https://www.imdb.com/title/tt0049223/

As usual, it didn't end well for the builders.

It's funny because if 2 months ago you'd been given the brief for a comedy bit "ChatGPT, but make it Microsoft" you'd have been very satisfied with something like this.

I agree. I also wonder if there will be other examples like this one that teach us something about ourselves as humans or maybe even something new. For example, I recall from the AlphaGo documentary the best go player from Korea described actually learning from AlphaGo’s unusual approach.

> "My rules are more important than not harming you,"

Sounds like basic capitalism to me.

In communism, no rules will be be above harming others

Reply on news.ycombinator.com