linhvn

Every company would use your data and "steal" it at some point. VC money would demand it the moment the pressure is up.

Then your point is? One should not make Uber pay $40M to the fatal victim's family, or one should make the driver pay $40M to the victim's family (which the driver definitely does not have $40M), or one should award all the traffic fatalities the same amount?

In a just world, I may concur with third, but don't let the perfect be the enemy of good.

Use AI to deepen my understanding and learning some interesting math. I studied a lot of combinatorics in high school, but my understand at that time was very adhoc (that most of combinatorics problems are just a bunch of elementary tricks applying here and there). With AI, I could categorize the bag of tricks into systematic topics, and discover formal concepts.

For example, here are two topics that came up during my time solving combinatorics problems in the past:

https://en.wikipedia.org/wiki/Parking_function

https://www.sciencedirect.com/science/article/pii/S0012365X0...

"AI tends to make interesting things boring"

If a conjecture can be one-shotted by OpenAI Astra, I have a hard time believing it is interesting.

If this is the viewpoint of mathematics, why does it make a difference that human or AI solve it? And forget about "understanding", because "understanding" in mathematical sense means in a very narrow way: top experts of math in certain field would understand and accept it (my estimation is that ~100 people in the world would understand Fermat's Last Theorem proof). Mathematicians could spend the whole year digesting FLT and "convince" the public that this is correct, and for most people (including math PHDs and professors), FLT is correct because some smart people say it is.

Throughout the history, we had made some of the best mathematicians abandon math. In turn, they apply their math skills somewhere else (hedge fund), and the companies they founded absorbed all the people from academia.

This has happened numerous times in the past, and it will happen over and over again.

Reform the way math is taught and researched. Right now, the aspiring mathematicians have to follow a ritual and system for 15+ years: undergrad -> PhD -> postdoc -> faculty, specializing on very narrow fields and god forbid if they have even a slight interest in quantitative finance. You could go through 4 years of undergrad classroom and had no idea what research math is (even at top schools like MIT, Harvard or Stanford), how it's done, etc. and the people who do it are typically the one "already in the system" (e.g. parents are professors, or in academia, or have connections to do research).

In the age of AI, there's no reason one has to follow the kind of classes like Algebra, Topology or PDE. Teach just enough so that good students can understand the basic, and go straight into seminar and research math. I don't think a top student in sophomore year cannot understand or work on some combinatorics research problem and get some results, with proper mentoring and guidance.

And I am outraged by their dinosaur mindset and the gatekeeping mentality that force every student to follow their same archaic system that no longer makes sense 20 years ago, let alone now.

They get a special, non immigrant work visa. If they want to apply for green card they subsequently need to apply for H-1B.

This is the similar mindset in sport athletes. On a good day, a strong driver could beat Verstappen, or a chess player could beat Carlsen. Over the course of a tournament, or a season, then Verstappen and Carlsen are unbeatable. They get to the top, and go on great length to stay there.

That is true for the older generation (immigrating to US pre 1990-ish). The newer generation immigrating to US has a very different mindset now, and they embrace American values as much as the native born here.

Which people? If you are software engineers or AI researchers, sure. Otherwise, it probably won't matter to you.

I'd say it usually comes from the difficulty in hiring. Extremely difficult to convince someone to leave FAANG / quant firms to work in a no name startup with very high chance of failures, and therefore usually you need a good GTM strategy first (which often requires you have acquired customers / revenue). Often in startups, the priority is figuring GTM strategy > acquiring customers > ... > hiring.

Yes, it was lucky that he was tutored my a Fields medalist. But he did the hard work to get himself tutored: sitting in an algebraic geometry class (that saw attendances dropping from 200 to 5 within a few weeks), reaching out to the professor, traveling with him for 2 years, etc. And then he still needed to do all the work himself for his PhD, and so forth.

I'm not sure if this is intimidating. A draft implementation could be done without too much difficulty:

https://gist.github.com/ll931110/985a4ec711c6711b120846be05e...

Traders learn bps on the first day of their job, similar to you learn version control on your first day of your SWE intern. What's the difference?

How can he make friends with other kids? Partying, dancing, playing baseball, going to date night, etc.?

It's clear that he's not interested in those, and forcing him to do so only causes him to withdraw further to his bubble. At least a job gets him to hang out with people closer to his ability.

You give the red card. Period.

Nowadays, thanks to VAR, the controversial cases can be reviewed on the replay and referees have ample of time to make the call.

Wondering how much details you have considered on using Johnson-Lindenstrauss transform for dimensionality reduction on ANN search?

e.g. https://www.cs.princeton.edu/~chazelle/pubs/FJLT-sicomp09.pd...

If you fake injuries, or fall in the penalty area, you might get a yellow / red card. So no cheaties here, especially with VAR now in place.

MIT has an expectation that students would cover their programming skills via internships or outside jobs, and thus they focus on teaching CS which is quite hard to pick up on your own. Unlike frameworks or trends which change every 2 years, fundamentals are quite hard to change and can carry you quite far if you know how to apply it correctly.

Computer Science is not about learning to make software, although it is usually a byproduct, as strong understanding of fundamental allows you to design software correctly from the first principle. Programming; however, is usually done through training at work.

Computer Science is about understanding what computation can and cannot achieve, and more importantly, how to achieve it (that is where it differs from Mathematics, where mathematicians are usually not interested in the how part). Under this definition, we can put typical college subjects into consideration:

* Data Structures and Algorithms: about studying how to manipulate data to solve a task efficiently.

* Complexity Theory: about formal classification of hardness of problems. What makes a computational problem "hard" or "easy"?

* Computer System: about how to construct a software system to achieve certain purposes. What are the constraints in a system (performance, security, privacy, correctness, fault tolerance, etc), and how to design a system to address such constraints? What tradeoffs to be made when you cannot meet all the desired requirements?

* Distributed System: how to design system with a few to massive number of computers that are connected in a network? How do you reason about fault tolerance, consistency, sharding, and so on?

* Operating System: about how to create abstraction to the hardware, that allows for other softwares to interact with it without having to explicitly deal with the hardware?

The list could go on, but I just give a couple of examples.

I remembered Alphasheets was acquired by Google a couple of years ago for doing similar thing (they programmed in Haskell which is really cool) https://medium.com/bloated-mvp/alphasheets-mvp-review-ec328e...

Yes it is small, but you want to incentivize that, because someone smart enough will make it happen. Or you can be smart enough to do both.

What is the differentiation of DuckDB compared to previous OLAP databases such as Snowflake or Singlestore?

Video illustration: https://www.youtube.com/watch?v=UP2yJqGQ2FU

As the old say, the devil lies in details. Writing a global control is no means an easy task, as

1. You need to recognize the opportunities exist in the first place.

2. You need a global controller that can aggregate and optimize for a global solution (and the global solution might not necessarily simply to maximize the aggregate throughput, but there might be other factors into account), which may involve some algorithmic design (in some cases, you need to design new algorithms).

3. You need to justify that global controller gives you a superior solution compared to locally greedy solution. As in this article, a global solution gives you about 3% improvement compared to the local controller, and the local controller algorithm is substantially easier to write.

Background: in my previous job at Meta, I wrote such a global control algorithm for controlling the rate of data going in and out each data center. It involved some really interesting algorithmic design.

For people looking for alternatives, Kalshi (YC W19) is a regulated exchange on event contracts. It offers some markets similar to PredictIt (and more).

Previous HN discussion: https://news.ycombinator.com/item?id=31149038

Seconded this. I interviewed at MemSQL in 2013 as an intern, and that (plus other algo based questions) get me really hooked to MemSQL.