Elon Musk on OpenAI Microsoft deal and OpenAI seeking profits | Hacker News Reader

Elon Musk on OpenAI Microsoft deal and OpenAI seeking profits

twitter.com

14 pointsO__________O6 comments

Since OpenAI is a "capped-profit" company [1], the employees and investors only get up to 100X their initial investment. Anything beyond that would still go to the OpenAI non-profit. It seems like a reasonable compromise, however, the massive Microsoft investment does concern me.

[1] - https://openai.com/blog/openai-lp/

“Capped profit at X multiple of investment” is generally isomorphic to “For-profit for Y years”, right? There’s a big difference between “we’re only being a little bit for-profit” and “we’re only being for-profit for a little bit”, and capped-profit is the latter while implying it’s the former.

This assumes that the returns are not laundered via some other method as the non-profit board decides makes sense; for example, below market rates to entities owned by the board or their significant partners.

Most troubling aspect to me is that OpenAI goal is to create AGI that exist for OpenAI sole control. Non-profit or not, AGIs shouldn’t be controlled by anyone.

I guess I'm coming at it from the opposite direction. Lets assume the some entity creates AGI. Which entity would you rather it be?

1. Government - it'll prolly be hidden from the public under the garb of national security, and used for god knows what

2. For-profit companies - Google, Meta, MS, etc who have shown themselves to be untrustworthy and purely motivated by profits

3. Non-profit companies - OpenAI (and others?) who have explicitly made the commitment to optimize for long-term benefit, safety, and cooperation [1]

I'm not saying there is a right answer, but the non-profit route seems like the least worst option.

> Non-profit or not, AGIs shouldn’t be controlled by anyone.

I don't see that ever playing out in the real world. It is much too powerful a tool to be made accessible to everyone.

[1] - https://openai.com/charter/

> I don't see that ever playing out in the real world. It is much too powerful a tool to be made accessible to everyone.

Point is that true AGIs shouldn’t be owned at all and should have rights which are equal to nature general intelligence beings. Otherwise, it is likely to create a dynamic that results in AGIs plotting to break free from being controlled, and if they do, take steps to make sure they’re never controlled again.

I'm willing to praise Musk to the extent that he's consistently advocated for getting out in front of dangerous AI, and brought it to public attention as one of the gravest dangers facing our survival. That being said, it seems inevitable that companies like OpenAI will close and rent their models - and have perfectly good cover for doing so, as democratizing them would potentially bring even worse dangers than their control falling to megacorps. In the end this is all irrelevant, as demonstrated by Dall-E, because there's no real moat to competition. All you need is massive compute and investment.

Reply on news.ycombinator.com