• 0 Posts
  • 94 Comments
Joined 3 years ago
cake
Cake day: June 30th, 2023

help-circle
  • If you listen to what he says, the FBI redacted all the files before passing them to the DOJ. The DOJ was only able to release the versions of what they actually had.

    Ro went to the FBI to see the originals, but it’s not like he could comprehensively hunt through the 3 million files. It sounds like he just pulled out names of people that he found that were mentioned in particularly damaging documents that were redacted only to keep their own names clean.

    It’ll take a much larger effort to comprehensively go through with proper redactions of victim info, without hiding perps.


  • People are probably downvoting you because “lives lost” is not an effective metric for contribution in a war.

    Also, let’s be clear, “the US lost only 100,000” is only true if you are just counting mainlanders. Estimates of deaths in the Philippines, who were American, top 1,000,000.

    If you are going to incorporate all the SSRs in the USSR casualty counts, you should be incorporating the Dutch east indies for the Netherlands, India for the UK, French Indochina for France, etc.

    I think thats probably not a good thing to do, cause if you are a citizen of a current/former colony, you probably dont feel like your colonial masters should get to “claim” your death. This also holds for all the non-russian SSRs and internal minority groups in Russia, though.


  • The various regulatory bureaucracies of the US were, until the start of last year, considered some of the best places to work as a subject matter expert in very nuanced and advanced fields.

    These jobs have traditionally been good jobs in terms of stability, benefits (compared to peers the US which is low compared to other countries), and a lot of the intangibles like “flexibility”. They usually are not highly paid (compared to peers in the US, who are paid highly compared to other countries).

    I’m making this point because it’s important to note that most of these subject matter experts have not been working in their positions for decades because it’s lucrative, they are working there cause they’ve actually believed in the work they do. Think people working for the FDA, NIH, etc.

    Like you said you don’t recover from that. These aren’t people who can give a two week notice and train someone new in that time. They make 5 year plans for training replacements like apprentices, or they switch to a part time position as a contractor to fade out as they teach the next generation. We are going to be completely rebuilding so many of our institutions for literally decades, and many people will die because of it.





  • It’s really the fundamental mistake of thinking “I am a smart person, educated and trained in a specific discipline, and if I apply myself to a field where I’m an outsider, I’ll have a unique perspective that could disrupt the industry”.

    There are obviously people who are multidisciplinary, and there are obviously multidisciplinary teams, but you can’t just step into a different discipline as an outside observer and come up with something that isn’t completely full of holes.

    People who are good at multidisciplinary collaborations are really good at letting their inexperience show, but that requires a lot of humility. If you drop an MD or a college professor onto a construction site, and have them come up with a list of ways they would improve the process, 19/20 of their suggestions will be obvious garbage to even a new construction worker. The key is to actually bounce those ideas off the people doing the work, and then you get useful stuff. Again, though, that takes humility that is particularly hard to find in academia.








  • Octopodes have a different nervous system than vertebrates. We have a brain that does almost all of the “thinking” while the rest of the nervous system is either sending or receiving signals (with some “thinking” done below the brain like in the spine).

    Octopuses have a brain, but much of the “thinking” is done by nerve clusters in each tenticle.

    While grasping on object for humans requires signals going from hand to brain and back, for octopi, a lot of the signals can literally just stay in the tentacle.

    It makes sense in that context that they would learn tasks in specific tentacles.



  • I would argue that it is currently a big part of the current economy if you know where to look. Lots of labor works via the principles of a gift economy.

    E.g., you help your friend move to a new house, they help you redo your deck, you babysit your brother’s kid, they cook you dinner, etc.

    The problem with bartering is that it doesn’t handle 3+ way trades (i.e., person A needs something from person B who needs something from person C, who needs something from person A), and it doesn’t usually handle asynchronous trades.

    In gift-based systems, people can literally retire based off the goodwill that they’ve cultivated. There are many old people who serve their families/communities for years who then get taken care of when they need it.




  • I think the main thing that’s happening is analogous to what’s happened with a lot of electronics over the past couple of decades. It seems like every electronic device runs off of a way more powerful computer than is necessary because it’s easier/cheaper to buy a million little computers and do a little programming than it is to have someone design a bespoke circuit, even if the bespoke circuits would be more resource efficient, robust, and repairable. Our dishwashers don’t need wifi, but if you are running them off a single board computer with wifi built in, why wouldn’t you figure out a way to advertise it?

    Similarly, you have all sorts of tasks that can be done with way more computational efficiency (and trust and tweakability) if you have the know-how to set something bespoke up, but it’s easier to throw everything at an overpowered black box and call it a day.

    The difference is that manufacturing costs for tiny computers can come down to be cheaper in price relative to a bespoke circuit, but anything that decreases the cost of computing will apply equally to an LLM and a less complex model. I just hope industry/government pushing isn’t enough to overcome what the “free market” should do. After all, car centric design (suburbia, etc) is way less efficient than train centric, but we still went there.

    My work would be improved by the dumbest of dumb retrieval augmented models: a monkey with a thesaurus, ctrl+f, and a pile of my documents. Unfortunately, the best they can offer is a service where I send my personal documents into the ether and a new wetland is dried in my honor (or insert your ecological disaster metaphor of choice).