Hacker News Reader

Comment by jere | Hacker News Reader

My experience online is if you don't show full throated rejection of all things related to AI, you're labeled an AI bro. None of the antis care how you feel.

I had my decade old game completely cloned on steam (clearly done using AI as it was almost fully reimplemented in another engine). And I had several people gleefully tell me I deserve this because I've used AI.

I have complex emotions here (overall dread and especially hate for slop, since it's not just bad quality but also endless lies), but no one cares about that nuance.

View on HN

> I have complex emotions here (overall dread and especially hate for slop, since it's not just bad quality but also endless lies), but no one cares about that nuance.

Many do, but I sympathise for your general point. I think the loudest voices… well, the boosters mostly own the platforms, and half of them already had terrible reputations; the normal people who hate it garner a lot of sympathy for entirely understandable reasons; and then there's people like me (abnormal nerd that I am) who are on lesswrong and therefore have had many discussions about doom and what to do to reduce it and if we can use the AI to help reduce it or if this is just asking for trouble.

Did y'all reach any consensus on that?

On if using the AI to help reduce risk, being a good or bad idea?

Not really. Yudkowsky has just recently posted a blog-sized comment which compares most attempts to know if the AI is misleading you while it does so, unfavourably, with people who think they've found a way to violate conservation of momentum.

Specifically, if you think you've found a way, you've probably bodged the maths and not noticed; and that science doesn't work by debate it works by experiment.

Some of the replies (me included) are like "hang on, this isn't shaped like physics, it's shaped like maths; debate does work for maths".

Then again, I also noted I'm skeptical of the AI companies putting in the effort to do this kind of testing, or listening when the answer is "no": Race dynamics, and their money depends on the answer not being an emergency stop button.

Reply on news.ycombinator.com