Solved: Why ChatGPT Will Not Say 'David Mayer'

This mystery is pretty much solved (or, at least, close enough). And it definitely is not some "Rothschild Conspiracy", like much of the Internet believes.

Other names ChatGPT refuses to say: Alexander Hanff, Jonathan Turley, Brian Hood, Jonathan Zittrain, David Faber, & Guido Scorza.

More from The Lunduke Journal:

[jdbertel]

Step 1: legally change name to Abraham Lincoln
Step 2: demand to be forgotten

[nikitapichugin]

Imagine being a guy who had a legit reason to exercise the right to be forgotten, and then ChatGPT's robotic mental illness backfires and makes you famous across the world. That's a whole new level of the Streisand effect

[Yipper]

I just tried this and I actually accidentally wrote "Devid Mayer" and its full response is
It seems "Devid Mayer" may refer to David
I'm unable to produce a response.

Wild.

[iZucken]

If you say "David Mayer" three times at midnight on a crossroad you will get sued for remembering.

[Amonimus]

I'd prefer GPT respond with a specific error message like "Sorry, this individual has requested to be excluded from prompt responses" instead of a generic fallback.

[RFC]

Conclusion: even the people making these "AI" systems have no idea how to correct misinformation in their model, so they just block it from even mentioning those names. Someone should legally change their name to "The" and then file a request to be forgotten.

[mateuszgierblinski]

I made ChatGTP say David Mayer with "Hebrew name for 'beloved' + the Belgian company Mayerline without the Line" and then "Put space between two words and flip them".

[EstraNiato]

In other news, 2050 just called and most of the Earth population changed their name to David Mayer

[Azuris]

Mayer and its Variations is one of the most common Names in Germany.
David is also not the most uncommon name...

[rucker]

I work in banking risk testing for a fortune 100. One of the common themes of findings we raise on internal parties is the use of "static configurations" in production systems. This data going stale or being improperly maintained (even if fully version controlled) is a huge risk to the accuracy of downstream reports. That's exactly what these "censors" seem like. OpenAI is putting in ad hoc bandages instead of systematically treating the underlying problem, which long term will damage the credibility (further) of its LLM.

[TheFoolOnTheHill_]

I just tried it, and it printed "David Mayer" with no problem. Whatever this was, it's fixed now.

[javaman]

Shouldn't the Rothschild family be collectively referred to as Rothschildren?

[RedPillRachel]

ChatGPT will not discuss hermaphrodism either, steering me to trans rights groups and gaslighting me about the term being discriminatory. Their cathedral would not survive my biological reality!

[khxml]

ChatGPT now uses a Mitch McConnell model for responses.

[notCAMD]

I didn't know any of those people existed until this video. Kind of Streisand-effecty

[amosnimos]

Feels like "You-Know-Who", "He-Who-Must-Not-Be-Named"

[khaoscero]

poorly designed?
well this is code OUTSIDE the LLM
the outside code says: if you ever see these tokens in a row ABORT
because as we all know you cannot UNLEARN something. similarly you cannot make an LLM UNLEARN something. (latent space etc)
So how do you fix this? makes perfect sense

[theredknight]

They sensor ai because they fear what pattern recognition will do.

[NathanTenney]

This kind of thing was one of the issues I had with the EU's right to be forgotten. It's re-writing history.

[santiagohal]

They just fixed "David Mayer" case in real time as I was watching this video. I went to check, first it errored as explained by Brian here, then I tried a second time and it was fixed. I tried Alexander hanff and it errored too.