They absolutely lie and cheat. I recently had a problem where a process would die in a container. I told Claude to investigate. It came up with a hypothesis then I told it find a reproduction based on that. It spend many failed attempts until it found the "reproduction" to SSH into the container and `pkill` the process. Claude "knows" that this is cheating, because if I ask another instance to review that reproduction, it totally identifies that as nonsense.

Models sometimes blatantly lie and cheat. In a social context, where actors remember, that might work the first time but you get penalized in subsequent tasks with loss of trust.

> Everyone knows you can tackle those with heuristics, but you don't (ALWAYS) have to sacrifice optimality

> We absolutely have tools that can (OFTEN) find provably optimal solutions in reasonable time

An incentive based solution is: reduced pay on sick days. With _no pay_ you get sick people at work. With _full pay_ you get healthy people at home. With _reduced pay_ you can find a sweet spot. That adds no bureaucracy for doctors. For employers there is some but it directly translates into cost savings.

Don't like smoking? Try out right banning it. It's impossible. You have to fight the lobbyist, the addicted consumers, the workers employed by the industry, ... Slapping a tax on it is much easier. A soft NO that's giving everyone time to adjust.

Comments URL: https://news.ycombinator.com/item?id=49151933

Points: 1849

# Comments: 742

https://salmonnoodlekyoto.com/

I got the "Salmon Dandan Noodles". Good lord that broth. No idea what was in there. It wasn't clear; rather slightly thick and creamy. And the salmon was so tender. I don't even like pork but I also ate that without thinking. Man I still miss that dish.

- That's easier to get past the lobbies.

- It still surfaces unexpected consequences but reduces impact.