esyir

Spoken like the safetyists that impede beneficial progress due to their impossible demands for 0 risk

There are internal controls as to the use of personal data within Google. Contrary to the popular opinion, personal data is not actually a free for all within Google

I think the biggest thing that people refuse to accept is that the optimal number of accidental deaths is non zero

And this carries on for all the other sources of injury and other "bad" things. Provided, of course, the cause it's explicit and direct.

I'll add an expansion here. It's more useful to you locally, as you have excess compute that's generally wasted. If you're serving multiple user and trying to max output, you might cost some in this case

I do think that this is natural. When you use LLM coding tools, you're becoming a lot more like an architect/staff/manager, rather than the direct coder. You're setting out the spec, coming up with the design, and coming up with the high level structure of the project.

However, this comes at the cost of losing track of the minute details of the implementation because you didn't write it yourself. I find it a bit analogous to code I've reviewed vs code I've written.

However, I've found using AI for code structure summary and questioning tends to be a good way to get around it. I might forget faster, but I also pick it up faster.

My impression is that, seeing the rising AI use rates everywhere, that you are in a bubble.

Could be me too, but seeing China's general societal infatuation with AI outpace the US by orders of magnitude, I think that's a bit less likely.

Sure, then they can go demand said standards for social media platforms including expected amount per N post, just as car companies are not expected to have car fatality rates be 0.

The fact is that simple scale means that there will always be something, no matter how abhorrent. Small scale doesn't change this, it just concentrates it.

We're talking about CSAM right? Which all platforms remove proactively, build models to remove and essentially always respond to when informed.

Demanding some perfect immediate magic response there is the equivalent of asking car manufacturers to prevent all deaths.

If car manufacturers cannot bring car related deaths to zero, they too should no longer be legitimate companies.

For a community full of engineers, I'm always surprised that people always take absolutionist views on minor technical decisions, rather than thinking of the tradeoffs made that got there.

Not the OP, but my interpretation here is that if you model the replies as some point in a vector space, assuming points from a given domain cluster close to each other, replies that span two domains need to "tunnel" between these two spaces.

I think what my ideal is basically DnD, but with an AI DM.

This is something that I'm hoping the current LLM and future AI work eventually get us to. If we can get persistent context and memory, or at least a simulacrum of that, we could get to truly dynamic reactive worlds

I'd say that the internet has also strongly lowered the barriers to external propaganda and influence, which is another major factor here. When you've got a huge swarm of "people" with no stake, or even a negative stake in your country, that's a naturally destabilizing factor

You mean like the countless western "safety", copyright and "PC" changes that've come through?

I'm no fan of the CCP, but it's not as though the US isn't hamstringing it's own AI tech in a different direction. That area is something that china can exploit by simply ignoring the burden of US media copyright

Just a note, chatGPT does retain a persistent memory of conversations. In the settings menu, there's a section that allows you to tweak/clear this persistent memory