What do you mean specifically by safety? It seems everyone, young and old, is running AI in YOLO-mode with full system access.
It's why I built the yoloAI tool, in fact!
That looks useful. I currently run OpenCode inside a containerd/Kata container and I see you support that backend. One thing I've been lazy with because it hasn't bit me yet is assuming the agent won't touch ".git", and diffing afterward. Being able to diff/apply modifications from the sandbox copy would be useful.
AI safety is still an issue with bad AI. Not only do you have to protect against what it can do, but also protect against what it might accidentally do, the lies it tells that the masses believe, or people seeding the internet with misinformation specifically for AI to pick it up and regurgitate it to people without a proper source. It’s like money laundering, but for misinformation.
All these things are safety concerns of AI, but just rogue AGI.
If you asking that do experienced programmers think AI safety experts (i.e. ai will doom the world) are overstating theie position the answer is that things are not the same and you cannot predict the future based on the past.
A much better approach would be to just see for yourself what you think. Use it and see how good/bad it is because anything you read or see on social media is likely just marketing.
I would use the word naïveté, but arrogance too, sure.
Define “older”, please.
Someone is older in late 30s. :)
Or did you mean “those who experienced mainframes”?
Thanks.