Because like the word 'intelligence' the word safety means a lot of things.
If your language model cyberbullies some kid into offing themselves could that fall under existing harassment laws?
If you hook a vision/LLM model up to a robot and the model decides it should execute arm motion number 5 to purposefully crush someone's head, is that an industrial accident?
Culpability means a lot of different things in different countries too.
I don't see bullying from a machine as a real thing, no more than people getting bullied from books or a TV show or movie. Bullying fundamentally requires a social interaction.
The real issue is more AI being anthropomorphized in general, like putting one in realistically human looking robot like the video game 'Detroit: Become Human'.
Simply put the last time we (as in humans) had full self autonomy was sometime we started agriculture. After that point the idea of ownership and a state has permeated human society and have had to engage in tradeoffs.
Quite often we are incapable of identifying what those different conditions were. When something you don't think is important is the actual cause of the failure you're unlikely to notice and misatrubute the cause.
This is more true for indie hackers or solo team founders i guess, if you're just a designer in a big corp, you don't usually handle marketing beyond trying to build/design a marketeable product, devrel and other positions are more marketing like
Honestly, that seems like a solvable problem. Certainly not easy, certainly tremendously difficult, but I'm not sure it is impossible nor that we can't make strides in that direction. We're fundamentally talking about a search algorithm, with specific criteria.
I doubt there would be good money in creating this, but certainly it would create a lot of value and benefit many just from the fact that if we channel limited resources to those more likely to create better things, then we all benefit. I'd imagine that even a poorly defined metric would be an improvement upon the current one: visibility. I'm sure any new metric will also be hacked but we're grossly misaligned right now and so even a poorly aligned system could be better. The bar is just really low.
I work with quite a few F100 companies. The actual amount of software most of them create is staggering. Tens of thousands of different applications. Most of it is low throughput and used by a small number of employees for a specific purpose with otherwise low impact to the business. This kind of stuff has been vibe coded long before there was AI around to do it for you.
At the same time human ran 'feature' applications like you're talking about often suffer from "let the programmer figure it out" problems where different teams start doing their own things.
While I'm not disagreeing with you, I would say you're engaging in the no true Scotsman fallacy in this case.
AI safety is: Ensuring your customer service bot does not tell the customer to fuck off.
AI safety is: Ensuring your bot doesn't tell 8 year olds to eat tide pods.
AI safety is: Ensuring your robot enabled LLM doesn't smash peoples heads in because it's system prompt got hacked.
AI safety is: Ensuring bots don't turn the world into paperclips.
All these fall under safety conditions that you as a biological general intelligence tend to follow unless you want real world repercussions.
reply