Quanta Magazine consistently explains mathematics/physics for an advanced lay audience in ways that don't terribly oversimplify / still expose you to the true ideas. It's really nice! I don't know of any other sources like this.
Unfortunately I have not found this to be true (though it is in this case). There are quite a number of articles that are misleading or flat out false.
The best example is the quantum wormhole article and video[0,1], because it is egregious and doesn't take much nuance or expertise see the issues. I'm glad they made a note and wrote a follow-up[2], but all this illustrates what is wrong with the picture. For one, the article and video were published the same day as it was published in Nature[3]. Sure, they are getting wind of the preprints, but in this case there was none! They're often acting as a PR firm for many of the big universities and companies, unfortunately so is Nature.
The article was published Nov 30th, but the note didn't come till March 29th![4] You might think, oh it took that much time to figure out that there were problems, but no, only a few days after the publication (Dec 2nd) even Ars Technica was posting about the misinformation. They even waited over a month after Kobrin, Schuster, and Yao placed their comment on ArXiv[6]. Scott Aaronson had already written about it[7]. There was so much dissent in that time frame that it is hard to explain it as an accident. A week or two and it wouldn't be an issue.
But I think Peter Woit explains it best[8] (published, yes, Nov 30th).
This work is getting the full-press promotional package: no preprint on the arXiv, embargoed info to journalists, with reveal at a press conference, a cover story in Nature, accompanied by a barrage of press releases. This is the kind of PR effort for a physics result I’ve only seen before for things like the Higgs and LIGO gravitational wave discoveries. It would be appropriate I suppose if someone actually had built a wormhole in a lab and teleported information through it, as advertised.
I hate to say it, but you need to be careful with Quanta and others that __should__ be respectable. And I don't think we should let these things go. They are unhealthy for science and fundamentally create more social distrust for science. Now science skeptics can point to these same things as if there isn't more nuance all because they were more willing to take money from Google and CIT than wait a day and get some comments from other third party sources. (The whole peer review thing is another problem, but that's a different rabbit hole).
(1) This is awesome. Feels like this wraps enough complexity that it won't just be a toy / for prototyping.
(2) When a schema is provided, is it fully enforced? Is there a way to do migrations?
Migrations are the only remaining challenge I can think of that could screw up this tool long-term unless a good approach gets baked in early. (They're critically important + very often done poorly or not supported.) When you're dealing with a lot of data in a production app, definitely want some means of making schema changes in a safe way. Also important for devex when working on a project with multiple people — need a way to sync migrations across developers.
Stuff like scalability — not worried about that — this tool seems fundamentally possible to scale and your team is smart :) Migrations though... hope you focus on it early if you haven't yet!
> When a schema is provided, is it fully enforced?
Right now the schema understands the difference between attributes and references. If you specify uniqueness constraints, they are also enforced. We haven’t supported string / number yet, but are actively working towards it. Once that’s supported, we can unlock sort by queries as well!
> Migrations though... hope you focus on it early if you haven't yet!
We don’t have first class support for migrations yet, but are definitely thinking about it. Currently folks use the admin SDK to write migration scripts.
Question: do you have any favorite systems for migrations?
That's a great list, thank you! We are thinking along similar lines; looking forward to when we can design this portion. If you have other feedback, please let us know.
I haven’t seen a better solution than remolacha’s #2 (create separate temporary state for the form).
Forms just inherently can have partially-finished/invalid states, and it feels wrong to try and kraal model objects into carrying intermediary/invalid data for them (and in some cases won’t work at all, eg if a single form field is parsed into structured data in the model)
In two books I've read recently (Conquistador about Hernan Cortes, and Civilized to Death which contrasts the lives of modern humans with ancient ones) the authors mention how fat was often poached from the dead -- animals and people alike. It was used like a grease for all sorts of things, including filling wounds to prevent infection. Pretty interesting stuff.
In all the cases in the article it looks like shame plays a big role. I wonder if hikikomori is caused by a loop of [adverse circumstances that cause the person to feel shame] -> withdrawal to avoid shame -> being ashamed of having withdrawn [loop]
Shame of educational pressure might be causing this, as mentioned in the article. But why do we as society place kids under so much stress? Let kids be kids and learn by exploring.
I'd bet it's because the parents are feeling a lot of stress. Especially without a robust community support network/string external role models, children tend to inherit their parents' emotional states.
As a parent I also thought “I won’t repeat these patterns” but reality is, that is so hard. Often parents want the best for their children and use any possible technique to make sure they are successful.
I am not giving an excuse but rather want to point to our society and our behaviour. When an expat at work asked me yesterday where to move to make sure that his 5 year old will have the best schools of the country… with such an elitist behaviour, I can only facepalm and see this is going to be much worse in the next 5-10years.
Problem is, the world is an economic slugfest today unlike it was at least when I grew up. When my High School class graduated a long time ago, most of us were competing for jobs with people in our own small town. At most, we were competing with the surrounding counties. There was university for A students, community college and/or middle class office work for B students, normal working class jobs for C students, and tougher lower-paying jobs for D students. As for university, we were competing for entrance with mostly other people in our state.
Today's kids are competing with the entire world, and the middle class is disappearing. So it's much higher stakes. And it's bimodal: You're either one of the few winners and get to live a comfortable life with a professional job, or you're off to WalMart or an Amazon warehouse, or Prison. The "kind of comfortable middle class life" is shrinking quickly. So it's not enough to just get straight A's. You need extra credit, get a 5.0 GPA, take all the "right" AP classes, have the "right" extracurriculars, and the "right" community service and so on. Otherwise you risk landing on the bad side of the career bimodal distribution.
> You're either one of the few winners and get to live a comfortable life with a professional job, or you're off to WalMart or an Amazon warehouse, or Prison. The "kind of comfortable middle class life" is shrinking quickly. So it's not enough to just get straight A's. You need extra credit, get a 5.0 GPA, take all the "right" AP classes, have the "right" extracurriculars, and the "right" community service and so on. Otherwise you risk landing on the bad side of the career bimodal distribution.
This seems a little hyperbolic. The requirements you describe are probably true enough for top 20 universities, but they aren't true for top 100, and there still seem to be plenty of random white collar office jobs that hire people from merely medium-ranked universities.
Problem is, with growing inequality, the 80th-percentile school and the 20th-percentile school is vastly different, whether that is in school resources or the life trajectories of graduates.
How’s PostgreSQL’s code quality? If projects have tons of technical debt or poor abstractions it can often be hard to make significant changes. Is that the case here, or no?
eh...you know, if you're in the right parts its actually pretty pleasant. there is alot of good design in Postgres.
otoh, there is some awful legacy stuff, and some really awfully crosscutting stuff around physical logging (I just looked at the locking around running queries on a replica, and that's clearly never going to be correct)
despite the fact that its in C, given a couple major refactors that will never happen, it could be really nice
It was certainly an unresolved question before they did this work!
Naively, it seems reasonable to believe that if you adjust all the weights of a neural net towards the behavior you want via SFT and RLHF, that it would compete with/mute/obscure undesired behavior like a back door. But it seems not to be so… Indeed the cute mask does not cover the entire shoggoth— it may still have tentacles (https://images.app.goo.gl/YW9g3BvwGqGwYTgd6)
An LLM can be executed in an “OODA” loop like in AutoGPT and given a goal towards which it takes agentic actions, especially if the LLM is fine-tuned for function calling. So, it can be the main component of an agent that does have goals/de-facto motives! The wrapper code can just be a couple hundred lines.
AutoGPT itself is pretty weak, but it’s possible to write wrapper code that leads to stronger agency. Also, agents formed this way with GPT4 are way stronger than with GPT3.5… so expect this trend to continue.
