A new partnership to promote responsible AI | Hacker News Reader

A new partnership to promote responsible AI

blog.google

17 pointssgift9 comments

"Advancing AI safety research to promote responsible development of frontier models, minimize risks, and enable independent, standardized evaluations of capabilities and safety."

So what's the plan for those who don't want to build "safe, aligned" AI? It is already evident from those who attempted to exploit the primitive LLMs we already have that someone will attempt to do this.

I see no solution in concept to the premise that alignment theory has proposed. It is simply a paradox as it stands. I've expanded on this quite a bit here, FYI.

https://www.mindprison.cc/p/ai-singularity-the-hubris-trap

Are Palantir or equivalent defense-associated companies in China going to be part of this?

The CEO of Palantir just published an open letter calling for a Manhattan Project for superintelligent AI weapons. https://archive.is/KSOv4

https://news.ycombinator.com/item?id=36873875

It would probably be a good idea to have different groups/regulations for civil and military tech. Trying to get people with very divergent goals to operate under a common set of guidelines seems like a disaster waiting to happen.

Just my opinion...

The divergence being one group attempts to build hyperspeed superintelligent AI command, control, and swarm deployment as fast as possible for purposes of global dominance, whereas another group uses it responsibly to help society.

You're right, of course the US and China should keep strict separation between moral and strategic concerns. That has always led to a safer world in the past.

"Trying to get people with very divergent goals to operate under a common set of guidelines seems like a disaster waiting to happen." -- you think it's safer for the US and China to NOT work under common guidelines and instead race for world domination by deploying the most powerful superintelligent AI-enhanced military force??

I should have been more clear - civil entities should be one group and have one set of guidelines, and military entities should be another group and have another set of guidelines (US, China, World)

You were perfectly clear and 100% wrong. Treating these things as separate concerns will result in hundreds of millions of deaths.

It’s hard to say until more details are forthcoming but this reeks of a cartel.

What do you mean? Google is so responsible with all of our user data. What could possibly go wrong entrusting them to be responsible with AI? /s

Is this Google's way of ensuring that all future content from OpenAI is watermarked? I am specifically saying OpenAI because their model is far ahead of the competition, and that will only amplify with GPT-5 and so on.

That's how I interpret this anyway... They promote safety and all that, but underneath a lot of it has to do with bottom lines.

Reply on news.ycombinator.com