That is an interesting concept. But actually, even in this extreme situation, relatively few Trumper officials have no friends on the other side of the aisle; that will change over time as things become even more impossible to defend, but the "it's just a gig" aspect of it has not entirely faded.
There is definitely more "control" at the executive branch level: FBI people have literally been told they will be fired if they don't cut off contact with former FBI Trump critics.
But even so, being told you will be fired is small beer compared to how proper cults deal with dissent, and the level of control (you won't get a job in this town) is different to the level of psychological control a proper cult wields.
Why did the original post for which these are the discussions disappear from HN?
I cannot even find it with a word search and it should not have been displaced from the top 30 yet.
What happened to a new JS front end library every week?
If this keeps up, we won't get to completely throw away all of our old code and retool every two years (the way we've been operating for the last 20 years)
How will we ever spend 85% of our time spinning up on new js front end libraries?
And don't even get me started on the back end.
If AI had been around in 2010, we probably still have some people writing apps in Rails.
OMG what a disaster that would be.
It's a good thing we just completely threw away all of the work that went into all of those gems. If people had continued using them, we wouldn't have had the chance to completely rewrite all of them in node and python from scratch.
This website is mostly full of very junior developers that just click on articles they think "sound smart" or are a hot take they agree with.
Don't get me wrong, it's also one of the few places where you find experts from all sorts of industry mingling. But the quality of commentary on HN has plummeted in the last 10 years.
As if web/css wasn't a new gui library no one asked for. Peak programming was VB6 & Delphi 6 (maybe 7). Everything after that was just treading water in increasingly degenerate ways.
Eh. Web is the best write-once-run-everywhere we've achieved so far, especially with the proliferation of WASM. I'd be lying if I said it was perfect, but it's better than Java.
Some real revisionist history as Rails cribbed most of those gems from Python. Now Python just rebranded for web and its doing everything Rails does and more.
I guess this signals the end of focusing on testing to prove one's "rank" as a student and actually start focusing on educating students rather than evaluating them.
The whole problem with factory mass-education is the almost exclusive focus on evaluation. A little bit of education happens as a side-effect of the unrelenting soul-crushing evaluation so we've put up with it and called it education for 100 years.
ChatGPT will make it nearly impossible to cause any educating to happen as a side effect of constant re-evaluation.
That means we are going to have to focus on education directly. Probably just train ChatGPT to be a really good tutor and everybody gets a 19th century rich boy's education instead of the junk we've had in feed-lot-style schools for the last 100 years.
Note that Waldorf schools are kind of immune to this problem since the kids don't get to touch a computer until they are in 9th grade and heck, they don't even read until 2nd grade.
In France the extreme majority of your grades comes from in-class exams that are essay-based or at least long format rather than multi-choice like in the US. Homeworks are just exercises to learn and aren't graded.
This entirely solves ALL the problems introduced by LLMs.
The US just needs to adapt their evaluation system.
At my high school, teachers were given a lot of freedom to decide their own grading, but had the rule that the final could not be more than 20% of your grade. Typically, homework ended up being 50%.
I failed my Geometry class despite getting 110% on the final (She had some hard extra-credit questions that went beyond what we learned) because I didn't do a single homework assignment.
Meanwhile, this high school made all sorts of claims about preparing students for college, and my first math class in college, homework was only 3% of your grade.
This is probably because you can fail catastrophically with the college way of doing things. I remember many people who failed one test out of three for the semester and essentially failed the class (given they did average on the other 2). No high school wants to deal with a system like that, it would be too much for kids to handle.
Cost is a huge factor. Asking a student to write what 7+8 is on a sheet of paper is a cheap substitute of high quality math education.
Not saying that this is the case in OP, it's very likely that the teacher will eventually catch up to the fact that a kid with A on their homework isn't quite a math genius. The same systems that would catch a child being aided by parents or cheating would catch this anomaly.
Waldorf schools work.
There is no testing, there are no grades until high School.
Lots of the silicon valley elite send their kids to Waldorf schools.
Are they all completely unaware of the effectiveness of deliberate practice?
I rather doubt it.
I think they are aware that motivation is what really matters.
My three kids all went to Waldorf School and never had grades until they entered high school at a public school.
None of them ever made anything less than an a+ in any class in a public school.
This was transitioning cold straight out of Waldorf with no testing to a public school.
They all got into very good colleges and are performing extraordinarily well at college. (My oldest daughter graduated from Stanford with a CS degree and straight A's)
Can you guys add huggingface transformers as one of the public demo repos?
I have some very specific use cases where I've seen ChatGPT with GPT4 totally fall on its face Dunning-Kruger style.
So specifically, you are saying that LLM coding assistant currently gets confused when working on a large source file but if it had room for more context, you could get better help in writing code because it would have understanding of the entire module. Correct?
Technically so, but not confused - it may hallucinate but that's unrelated to context window (and more due to dataset and number of parameters)
However, a longer context does mean you can give a lot more verbose instruction, maybe feed a whole code base over just a single script to give wider context over the current problem
This is one of the smallest scale cases I’ve heard of, but not nearly as weird or innovative as it sounds at first blush.
People have long been making analogous use of stomachs, intestines, even skulls if you go back far enough.