We're not asking the model to simplify something, we're asking it to perform a task. Its subtle preferences show up as an overcomplicated path to the goal.
In some cases, there are also nuances that we don't pick up on. Here it's our preference for simplification that's showing up. We set the lossy compression factor higher than it does.
If new free projects had to have a "100% artisanal" pledge I wouldn't even consider starting one. Hand-crafting the core design is still superior. Not delegating the grunt work for copy-paste level stuff is just wasting time, though. I don't require the circuit boards in my electronics to be hand-drilled either.
> I don't require the circuit boards in my electronics to be hand-drilled either.
This is the exact example I used the other day. Same with pick-and-place, CNC for subtractive manufacturing, 3D printing for additive, plastic injection molding, etc. I wouldn't want to live in a world where the only option for products produced by such automation were instead done one-off by artisans. Sure, if you want a hand crafted laptop where everything (save for the silicon, I suppose) was hand carved, hand placed, hand soldered, etc, you should, by all means, be allowed to pay $20,000-$200,000 for that privilege, instead of ~$2,000.
It's bizarre to me that so many developers are overly attached to typing every individual character. If I could get paid for my sense of knowledge and good judgement to design software while delegating 100% of the process of creating to a team of engineers, I would. And as a staff engineer, I'm already close to that, but there are things I have to roll my sleeves up and handle myself.
If you sneak up to a house and peer in the windows, there might be trouble. I haven't looked at court decisions, and those'd be regional/country based, but if a house is 200M away on a hill, and you need a pair of binoculars to see in it, is there 'public nudity' if they walk around without the shades drawn?
I'd say no.
Yet conversely, it's not illegal(criminal) to use binoculars to peer. Still, you can get an injunction(civil) if it's a common thing, because then it moves into harassment. Restraining orders have been given. So then the question is, why is this a thing for your neighbour peering at your house with binoculars some of the day, but not for a camera doing the same 24x7? EG, a Ring camera pointed at your house, with sufficient resolution to act as binoculars?
I really wish 'viewing someone with a camera' remotely, was looked at more from the frame of 'is it the same as using binoculars'. EG, is there a zoom lens on the camera, or just as important, is it of sufficient high resolution that it can act as a zoom lens in software.
I'd like to see a court case push back against that.
I've found with technology, that everyone from judges to lawyers seem incapable, at the start, of considering it with an analog, uh, analogue. Almost all cameras are as binoculars today.
Another strategy that's often used is to cause motion then detect the resulting transmission spike. That's also defeatable of course (cache+burst, CBR, etc).
The ideal detector should incorporate multiple methods. Someone should make a site that rates these for each detection method then adds them into an overall score.
Exactly. And it's not only about whether a specific job is safe from AI, but whether it's safe from the effects of those that aren't safe: competition from laid off people and downward wage pressure. Then there's the backpressure into the educational pipeline.
Similar, much narrower crises were managed for manufacturing jobs during the worldwide industrialization era. The solution for those displaced by trade was retraining.
Unfortunately, at the moment we have no clue of what workers should be retrained to. Not writing.
> My lethal trifecta model warns about the risks inherent in any agent system that combines access to private data with exposure to untrusted content and a way to communicate stolen information back to an attacker.
> ChatGPT Work combines all three!
The ChatGPT Work model would actually feel safer to me if they created a privacy boundary between the container-managing agent (browser operator/VM manager/code runner/etc) and the chatbot agent. Instead of me not typing privacy-sensitive things to the chatbot to avoid having them in my history, the chatbot would keep my history private from the container agent except on a need-to-know basis. That would remove the "access to private data" from the container's trifecta.
Not perfectly safe of course, just safer. Particularly if I could review the logs between the two agents.
reply