Hacker News Reader

Comment by dyauspitr | Hacker News Reader

I can’t wait till world models are good enough that you can ask robots to work autonomously to geo transform spaces.

You more or less can if you can afford the robots. Technically you need a dozen people or so to run the machines, but if you can afford the machine and costs to run it the costs of the people isn't going to phase you.

Now if the robots were affordable to someone on a minimum wage income that would be a big deal.

A Chinese 2 ton mini excavator is like $8k, and electronic solenoid valves could be added for like $2k. I've got to re-convince myself I don't need to take on this project like every other month.

That is the easy part.

However you just made a robot that will happily kill anyone who happens to wonder into the dig zone - we know from experience that people like to watch excavators work and they have no clue where the safe zone is. As such I have to discourage you from trying it - not that you couldn't make it dig the holes you want in a short time, but because you need someone trained watching for people walking in the unsafe zone and so you gained nothing. Waymo/Tesla have put a lot of effort into this type of detection, if you don't have similar backing you have no chance of doing it - if you do have that, there is likely a lot of money to be made.

Not all projects have to be taken on with the goal of commercialization. The beachhead functionality would be manual remote control - operate two machines from one seat. Then maybe some cameras to be able to work in the dark/cold/hot/etc from the comfort of inside. Even that is getting into the safety issue (any control system can go sideways), but seems straightforward to solve (some kind of redundant remote-control e-stop).

Any type of automation would be a huge pile of unknown unknowns, safety being one of them, how to even define goals for it to solve being another. Which is what keeps this idea in the realm of a dog chasing a car - if you gave me an excavator I could remote-control from my computer today, I'd not have the first clue what to actually do with it (beyond manually-controlled digging from my chair).

I do get that in the era of "AI" and vibe-coding why I may have set off your engineering sensibilities wtfometer though.

(Automating a lawnmower though...)

This all makes for fine hand wringing but is pretty much irrelevant at hobbyist project scales. The workplace standards that everyone tries to apply in bad faith to personal use are mostly designed to solve problems that aren't there. You're not gonna have distracted drywallers or rebar guys trying to take a shortcut and schlep materials through the excavator's range of motion in a personal use setting. You're not gonna have a boss yelling at someone who doesn't know better hop in an unstable trench or unstable latter and git'r'done when you're your own boss.

> You're not gonna have a boss yelling at someone who doesn't know better hop in an unstable trench or unstable latter and git'r'done when you're your own boss

eh, kind of. The motivation to take unsafe shortcuts manifests differently, but it is still there. It requires some vigilance to exercise self-responsibility and sometimes make yourself stop and take a step back and figure out a better way to proceed, while accepting that something is going to take longer than you'd planned.

The important thing is there are no people in the loop so it’s truly scalable.

Claude, please destroy nature.

Claude, please acquire the seedlings and reforest these 100 acres with the right native plants. Follow up for 3 years to ensure atleast an 80% survival rate. Ensure 20% are commercially viable hardwoods to pay for your operational costs.

Claude, please remove all invasives

Claude, please turn this overgrazed now-desert back into a prairie.

Reply on news.ycombinator.com