[seydor]

notOpenAI told them that this is so incomprehensibly smart that it will destroy all of us. Not because it's dumb and connects the wrong cables, but because it's super-smart. People can't surmise from that that the model just makes stuff up on the way

[JohnPrine]

OpenAI is clear about the fact that ChatGPT can hallucinate, where have they ever said otherwise?

[seydor]

One of the most widely circulated PR about GPT4 is that it passed the bar exam

https://www.forbes.com/sites/johnkoetsier/2023/03/14/gpt-4-b...

and it's had the most vocal and sensational PR about how AI needs regulation now

https://edition.cnn.com/2023/06/09/tech/korea-altman-chatgpt...

People assume the smalltext does not apply to them.

openAI does not get a lot of flak from the media for the amount of BS that chatGPT can blurt out

Does anyone remember what happened to Galactica which did the same thing ? That too was clearly labeled as hallucinatory . But it was shut down because they did not BS the media enough about regulations and such.

I m afraid these LLMs are turning into too much of a political game to be useful for much longer.

On the other hand, if they become political, then people will be even more incentivized to build offline, local LLMs

[arbitrarian]

This is what I don't get. Not only have they not said otherwise, but they put it right up front in a pretty easy to understand brief message before you start using it. I guess lawyers just click agree without reading too.

[jprete]

People are accustomed to ignoring the fine print as legal CYA with no real-world relevance. This is also why the product warnings that "The State of California considers this to cause cancer" are a joke and not a useful message.

[Too]

It’s not even fine print, there is a big Limitations section almost at the top of the chatgpt landing page with this as first bullet:

”ChatGPT sometimes writes plausible-sounding but incorrect or nonsensical answers.”

It does admittedly sound innocent, it should be written with a more scary tone on to how big extent it bullshits, not just “sometimes”. Still doesn’t excuse the incompetence of the lawyer in question.

[IIAOPSW]

I prefer to use the term "clopenAI"

[SilasX]

Or OpenAladeen

https://m.youtube.com/watch?v=NYJ2w82WifU