I'm surprised they would publish this honestly. Especially considering that they're in a lawsuit.
Probably didn't talk to a lawyer.
GPT-4 can pass the bar exam, so they can just ask it.
OpenAI and Elon Musk
331–340 of 1001 posts
Re: OpenAI and Elon Musk
#332Earlier quoted context omitted.
>they may be right that open sourcing AGI would lead to too much danger. I think this part is proving itself to be an understandable but false perspective. The hazard we are experiencing with LLM right now is not how freely accessible and powerfully truthy it's content is, but it is precisely the controls upon it which are trying to be injected by the large model operators which are generating mistrust and a poor und…
I'm not sure I buy that users are lowering their guard down just because these companies have enforced certain restricts on LLMS. This is only anecdata, but not a single person I've talked to, from highly technical to the layperson, has ever spoken about LLMs as arbiters of morals or truth. They all seem aware to some extent that these tools can occasionally generate nonsense. I'm also skeptical that making LLMs a fr…
> They all seem aware to some extent that these tools can occasionally generate nonsense.
You have too many smart people in your circle, many people are somewhat aware that "chatgpt can be wrong" but fail to internalize this.
Consider machine translation: we have a lot of evidence of people trusting machines for the job (think: "translate server error" signs) , even tho everybody "knows" the translation is unreliable.
But tbh moral and truth seem somewhat orthogonal issues here.
Re: OpenAI and Elon Musk
#333Earlier quoted context omitted.
I was talking about his public image - 'founder' of Tesla, doing everything for the love of humanity etc. It's all bullshit, and these emails show it again. He is an extraordinary businessman of course - but everything he does, he does for money and power - mostly coming from government subsides btw. Until recently though, he managed to make most believe that it was to save the planet and the humanity.
At this point most people didn't think Elon was a good guy even before these emails leaked
Yes, he has shown his true self when he called the British diver, who was rescuing kids out of cave in Thailand, a pedophile. That was more than 5 years ago.
Re: OpenAI and Elon Musk
#334Re: OpenAI and Elon Musk
#335Did people really think leading the development of the most important invention in human history wouldn’t involve a little bit of drama? I know everyone wants OpenAI to be a magical place that open sources their models the moment they’re done training, but it’s clear that they’ve chosen a reasonable path for their business based on both practicality and risk reduction. If they had gone another way, today they’d be an…
> ... chosen a specific path for their business based on both practicality and risk reduction
The risk reduction didn't go so well now that Elon is putting up lawyers to force them to become more "open".
This organization thrives in the dark and they know their secret to success depends on it. Would save every one a lot of time if they came out as a proper non-profit and dropped "open" in the name/branding.
> If they had gone another way, today they’d be an unceremonious branch of Elon’s empire
They substituted this dream with Microsoft.
Re: OpenAI and Elon Musk
#336IMHO Musk just wants to slow down OpenAI, so his company can catch up.
That being said, I am pretty sure there will be other conversations supporting Musks claims
Re: OpenAI and Elon Musk
#337> by opensorucing [sic] everything, we make it easy for someone unscrupulous with access to overwhelming amount of hardware to build an unsafe AI
by open sourcing everything, we make it easy for someone unscrupulous with access to overwhelming amount of hardware to build an unsafe BROWSER
by open sourcing everything, we make it easy for someone unscrupulous with access to overwhelming amount of hardware to build an unsafe OPERATING SYSTEM
Re: OpenAI and Elon Musk
#338Earlier quoted context omitted.
They claimed that GPT-2 was probably not dangerous but they wanted to establish a culture of delaying possibly-dangerous releases early. Which, good on them!
Do you really think it is a coincidence that they started closing down around the time they went for-profit?
No, I think they started closing down and going for profit at the time they realized that GPT was going to be useful. Which sounds bad, but at the limit, useful and dangerous are the same continuum. As the kids say, OpenAI got "scale-pilled;" they realized that as they dumped more compute and more data onto those things, the network would just pick up more and discontinuous capabilities "on its own."
That is the one thing we didn't want to happen.
It's one thing to mess around with Starcraft or DotA and wow the gaming world, it's quite another to be riding the escalator to the eschaton.
Re: OpenAI and Elon Musk
#339> Elon understood the mission did not imply open-sourcing AGI. As Ilya told Elon: “As we get closer to building AI, it will make sense to start being less open. The Open in openAI means that everyone should benefit from the fruits of AI after its built, but it's totally OK to not share the science...”, to which Elon replied: “Yup”. Wow the most open OpenAI has ever been is when someone sue them. On the other hand, th…
> this shows Elon doesn't care jackshit about the lack of openness from OpenAI. He's just mad only that he walked away from a monumental success. Yup he's pretty transparently lying here. He committed to fund for $1B dollars. Then when they wouldn't make him CEO and/or roll the company into Tesla he refused his commitment after paying out only 4% of it. Claimed the company would fail and only he could save them (agai…
[dead]
Re: OpenAI and Elon Musk
#340I guess Sam Altman finally somehow found Ilya Sutskever after saying he didn't know where he was and whether he was still working at OpenAI [1] [1] https://officechai.com/startups/sam-altman-is-unsure-if-ilya... .
can you please put a space in front of "https"?