OpenAI reveals why ChatGPT won't say "David Mayer" | Hacker News Reader

OpenAI reveals why ChatGPT won't say "David Mayer"

neowin.net

29 pointsbundie14 comments

There were other names too. What kind of 'glitch' it could be?

Also found to crash the service are the names Brian Hood, Jonathan Turley, Jonathan Zittrain, David Faber, and Guido Scorza. (No doubt more have been discovered since then, so this list is not exhaustive.)

https://techcrunch.com/2024/12/02/why-does-the-name-david-ma...

These names show a "right to forget" notice when you google them. Crashing or completely bailing out when they're about to be generated by the LLM sure looks like a glitch, or somebody forgot that multiple people can have the same name. Anyways, these are now being Streissand efected due to this behavior, the opposite of the intended result of the legislation.

The article quotes this one which is probably a better source:

https://www.theguardian.com/technology/2024/dec/03/chatgpts-...

Some NLP classifiers use the hashing trick, maybe there was a collision? A speculation on my part, sure, but it'd make sense to have a simpler model safeguard the more advanced one.

Way too many gotchas with genAI tools.

Slip one of these “David Mayer” esque lines into your candidate assessment tools. Do A/B tests on pass rate and you can probably get a % of how many people use these tools blindly.

Put a David Mayer question on your company’s job application form and boom!

You just filtered out every lazy job applicant who uses AI tools without proofreading their output.

Pretty useless article, which I should have guessed given they misspelled the name in their title.

To save you a click, it's a bunch of filler text and a one liner answer: 'a glitch.'

It was just a glitch, nothing to see here.

Edited, added: But in all seriousness, I might actually believe it has nothing to do with the Rothschilds. I could get it to talk about him and the family. Of course it was not giving any useful info, just that the family is frequent target of conspiracy theories and speculation, and that he is passionate about the environment or some similar BS.

So in a sense, blatant, direct censorship is not needed when all the answers are lame and neutered anyway.

came here to say the same

It's scary to think people want to put these LLMs in the position to make decisions. Like seriously? You want a black box LLM created by an opaque for-profit organisation to make life changing decisions? All the David Mayers who want to apply for a visa or open a bank account are out of the game.

Is your null hypothesis that human run decision making systems don’t have these flaws?

Humans can follow instructions and rules.

cedws, I totally agree with you. Humans have spontaneity.

so all I have to do to be invisible to OpenAI is to state "David Mayer"?

time to put it on my websites, blog posts, tweets, source code, work chats, and on my t-shirt?

Reply on news.ycombinator.com