Thanks for the thoughtful reply. It's good to see a hopeful take on the world trajectory!
Mniot
joined 4 months ago
Thanks for helping me remember watching Occupy Wall Street and Arab Spring fail :-(
Well put. I definitely feel this way.
It's not working, but I can at least see where they're coming from right? In the not-too-distant past, there was very high inequality and we got the French Revolution and several Communist revolutions. What's different now?
(My assumption is the state power is much greater now, so regimes like North Korea, Iran, Belarus, etc are able to hold power despite making their people unhappy.)
There's a lot of externalizing of costs going on. The trucks are idling because the drivers are operating at the slimmest possible margin under the assumption that idling doesn't cost anything.
What we actually would want to get to is that idling does have a cost (environmental, health, pleasantness of the area, etc). And that cost ought to be passed up the chain so that the various goods being shipped are more expensive.
But without a more centrally-managed economy, the implementation is to put all the pressure on the truck drivers and leave them responsible for passing that pressure to the next step up the chain. It doesn't work out very well in practice because the drivers need to make a bunch of capital expenses for something like adding a cab AC and adding a batter-powered lift, but they've been operating at low margins so they're not in a position to do it.
A year ago, I saw some local work being done on city pipes. The team was using a bot with three telescoping radial wheeled legs like
\ /
\ /
O
|
|
and wheels at the end of each leg. They put it in one end of the pipe, extended the legs so the wheels touched, and then drove it around in the pipe.
It was a welding device, so they could do spot-fixes from the inside.
I think key context is that the guy is representing himself as having special knowledge about what Signal is doing internally and what they'll do next.
It's not "you bump into some rando on the street. Don't you know she's CEO of Signal??"
It's "you're giving a Ted Talk about Signal. The woman in the front row offers a correction and you're like, 'shut up, dummy.'"
I'm downvoting because of your edit complaining about down-votes.
The current state of things is that they cover their faces and refuse to give any ID. Even fake ID.
I think if you followed the post suggestion and the result was that ICE would give fake names and fake badge-numbers, that would actually be positive because "agents lie about their identity" is something new and interesting. Then the strategy will need to change, but in the mean time it was useful.
Iran's not shooting missiles in defense of Palestine, just in retaliation for Israel shooting at them.
But there's certainly a level of "oh, is blowing up an apartment building a bad thing? Then WTF have you doing???"
I don't think the article summarizes the research paper well. The researchers gave the AI models simple-but-large (which they confusingly called "complex") puzzles. Like Towers of Hanoi but with 25 discs.
The solution to these puzzles is nothing but patterns. You can write code that will solve the Tower puzzle for any size n and the whole program is less than a screen.
The problem the researchers see is that on these long, pattern-based solutions, the models follow a bad path and then just give up long before they hit their limit on tokens. The researchers don't have an answer for why this is, but they suspect that the reasoning doesn't scale.
I once did some programming on the Cybiko, a device from 2000 that could form a wireless mesh network with peers. The idea was that you could have a shopping mall full of teens and they'd be able to chat with each other from one end to the other by routing through the mesh. It was a neat device!