Hacker News Reader

Comment by ctoth | Hacker News Reader

I thought this was interesting, Erik is a great thinker and writer, but I felt as though the bad stuff he doesn't like here, specifically the weird videos for kids, might equally be blamed on photoshop, YouTube, or any other technology in the chain. Alternatively, it can be blamed on specific people and their specific actions.

Like, blame the person that makes the weird videos, not the technology that can both make weird/alien scripts for kids videos and also write my React components and describe what's outside my doorbell camera. It's some serious baby with the bathwater action.

General-purpose technologies are general-purpose.

The behavior he's upset about is currently human behavior. I don't know of anyone who has wired up GPT and asked it to make money and it has decided to start making YouTube videos.

View on HN

The old "guns don't kill people, people kill people," argument... completely ignoring people use guns to kill people.

We have a school shooting something like once every two days in the US, but can't put children before gun ownership.

I don't know how we survive this.

I'm reminded of Balk's Law (Everything you hate about The Internet is actually everything you hate about people.)

At a certain point you have to acknowledge that if a technology enables harmful action on a scale previously unimaginable you have to start being very, very careful.

Machinists and gun makers are well aware how easy it is to make a machine gun or other illegal firearm, but they don't because that knowledge is locked behind experience, norms and expensive machines. This is why it was potentially a big deal when people started printing guns, not because it wasn't possible before to make dangerous or illegal firearms, but now it was an order of magnitude easier.

We have always had conmen, but running a good con was hard and you were limited to one at a time. Now with voice cloning and LLMs it is an order of magnitude easier, and infinitely more replicable.

I'm not sure the world is ready for the golden era of grifting, but I'm also not sure how you could ever stop it.

Just to add one thing: LLM’s as a force multiplier are much more than an order of magnitude (which is 10x). How long does it take a human to write a 1000-word article? Let’s say an hour, which is being generous. It won’t be a high-quality, well-researched article, but a human could probably get to that word count in an hour.

How long does it take an LLM to write a 1000-word article? Not 6 minutes —- that would be an order of magnitude faster. It would take ChatGPT maybe 10-20 seconds to spit out a 1000-word article. That’s 2 orders of magnitude faster.

Reply on news.ycombinator.com