Ask HN: Anybody Started a Research Institute? • Hacker News Board

Re: Ask HN: Anybody Started a Research Institute?

#61

Post by robertlagrant » Thu, May 27, 2021, 1:25 PM UTC

montenegrohugo wrote: post #49

_gsgc wrote: post #44

Hi. I went through this journey myself. I have a CS master degree, and always wanted to transition into AI research (focus on probabilistic programming, deep learning and self-play, but also simulation, learning-to-learn, bayesian inference, genetic algorithms). The initial plan was to self-learn machine learning, deep learning and statistics. I did that for 1 year, with the hope of joining on of the big AI labs afte…

You turned 40.000 into [redacted]? That's incredibly impressive (even with the recent crypto bull run). Was this all achieved with time-series prediction through ML models? I had the impression those were hard to get right, but props to you for clearly doing so.

Everyone's a genius in a bull market :D

Re: Ask HN: Anybody Started a Research Institute?

#62

Post by _Nat_ » Thu, May 27, 2021, 1:27 PM UTC

tsbischof wrote: post #36

I am helping lead the first research group at a recently-founded institute in Germany. What is the mission of your institute? Is this a collection of individuals doing their own research, or a group of collaborators working towards a shared goal? How do your ideal members measure their own performance? Is it through papers, blog posts, tech demos, products, or something else? With academics you need to be especially…

> With academics you need to be especially vigilant about the difference between vanity metrics and real output.

What would you consider "real output", especially in contrast to "vanity metrics"?

Asking mostly because I'm curious how someone in your position would see it.

For a lot of folks, academia's a fun ecosystem to participate in: they enjoy it and it pays the bills. To participate, they need to maintain popularity, much as a living organism needs to maintain energy. So their focus is on optimizing their popularity. They'd be primarily interested in metrics that represent their popularity, especially if those metrics help them plan how to gain more of it to further advance their career.

But other folks participate in academia to do research for other reasons, e.g. advancing technology. They're more likely to focus on maintaining sustenance-level popularity rather than maximizing popularity. They'd tend to dismiss many popularity-centric metrics as frivolous.

So if you're helping lead a research-institution, then are you primarily concerned with "real output" in the sense of "stuff that makes us popular enough to garner continued support"? Or are you more interested in another goal, like advancing science/technology? And how do you assess the value of a researcher's work given these goals?

Re: Ask HN: Anybody Started a Research Institute?

#63

Post by lokl » Thu, May 27, 2021, 1:27 PM UTC

Have you considered using the facilities and proximity to academic infrastructure to host academic conferences instead of full-time research? That might be a more realistic way to attract top talent when you are first starting. You would obviously want to do this in discussion with relevant nearby academic institutions, but being able to offer free accommodations for attendees + maybe some travel expenses could help to lure people for a week event, and a conference with the right attendees coming together can definitely benefit academic research.

A similar option would be to host classes or seminars on cutting edge research, taught by one or a few top researchers and attended by post-docs. Not full-time, but a way to start.

If researchers find it beneficial, if it's a conducive environment and the financials are right, if you master the role of "host," then you might be able to increase the number of conferences/seminars per year and, eventually, with a high-enough frequency, add full-time support staff and longer-term functionality (e.g., researcher-in-residence) that can lead to year-round activity.

Maybe this is a slow approach, but even one well-run conference can be beneficial and you'll be able to learn how to do this and make connections along the way.

Re: Ask HN: Anybody Started a Research Institute?

#64

Post by _gsgc » Thu, May 27, 2021, 1:28 PM UTC

montenegrohugo wrote: post #49

_gsgc wrote: post #44

Hi. I went through this journey myself. I have a CS master degree, and always wanted to transition into AI research (focus on probabilistic programming, deep learning and self-play, but also simulation, learning-to-learn, bayesian inference, genetic algorithms). The initial plan was to self-learn machine learning, deep learning and statistics. I did that for 1 year, with the hope of joining on of the big AI labs afte…

You turned 40.000 into [redacted]? That's incredibly impressive (even with the recent crypto bull run). Was this all achieved with time-series prediction through ML models? I had the impression those were hard to get right, but props to you for clearly doing so.

I think it's both easy and hard, depending how you look at it. Hard because it's a tough ML problem where you have to combat overfitting and the messiness of the real world; easy because crypto is clearly one of the easiest markets to trade and there are opportunities all over the place (compared to equities for example). On top of my head: stat arb, yield arb, exchange arb, funding arb, market making / liquidity providing (CEX/DEX), doing/preventing sandwich attacks, good old yield farming...

Re: Ask HN: Anybody Started a Research Institute?

#65

Post by robertlagrant » Thu, May 27, 2021, 1:29 PM UTC

nerdponx wrote: post #21

NationalPark wrote: post #11

Having worked with academics before, the biggest problem you will face is not having enough prestige to attract "great minds". You're probably off on the money too - academics will expect to be able to buy a comfortable house and support a family on your stipend, not live on ramen like a startup. They're a fickle bunch, but very rewarding to work with!

I don't think it's fickle to want to make a decent living from your hard-earned and valuable skills.

Valuable skills will enable a decent living for sure.

Re: Ask HN: Anybody Started a Research Institute?

#66

Post by gryn » Thu, May 27, 2021, 1:30 PM UTC

bigth wrote: post #7

I started a research institute. Our goal was to find out how to center divs in css.

too hard, you might have a better chance in your endeavor by forking the web into 2 things.

- A web of apps: A (sand-boxed) VM environment that has the advantage of not having to install the programs locally and with a capability based model. this is already similar to what's already here but once you commit to it you can ditch the DOM and other things.

- A web of documents.

Re: Ask HN: Anybody Started a Research Institute?

#67

Post by kordlessagain » Thu, May 27, 2021, 1:33 PM UTC

_gsgc wrote: post #44

Hi. I went through this journey myself. I have a CS master degree, and always wanted to transition into AI research (focus on probabilistic programming, deep learning and self-play, but also simulation, learning-to-learn, bayesian inference, genetic algorithms). The initial plan was to self-learn machine learning, deep learning and statistics. I did that for 1 year, with the hope of joining on of the big AI labs afte…

And now we know how inflated the markets are.

Re: Ask HN: Anybody Started a Research Institute?

#68

Post by kordlessagain » Thu, May 27, 2021, 1:34 PM UTC

> End goal as of now: assemble a few great minds, give them the space to work rent-free in a region with good academic infrastructure, ideally let them collaborate, watch the beauty of it.

My advice is not to expect others to do the work for rent and a “stipend”.

This isn’t a game. AI done wrong will be the end of us.

Re: Ask HN: Anybody Started a Research Institute?

#69

Post by gryn » Thu, May 27, 2021, 1:34 PM UTC

donohoe wrote: post #8

bigth wrote: post #7

I started a research institute. Our goal was to find out how to center divs in css.

Sir, I find this of great interest and a noble pursuit. I would very much like to invest large sums of my newly created cryptocurrency in this. Have your people contact my people.

DivCoin ?

Re: Ask HN: Anybody Started a Research Institute?

#70

Post by _gsgc » Thu, May 27, 2021, 1:36 PM UTC

___luigi wrote: post #55

_gsgc wrote: post #44

Hi. I went through this journey myself. I have a CS master degree, and always wanted to transition into AI research (focus on probabilistic programming, deep learning and self-play, but also simulation, learning-to-learn, bayesian inference, genetic algorithms). The initial plan was to self-learn machine learning, deep learning and statistics. I did that for 1 year, with the hope of joining on of the big AI labs afte…

That's impressive. This reminds me of a story I read about being an independent researcher ( https://andreas-madsen.medium.com/becoming-an-independent-re... ). For those who want to do the same. When did you start?. How did you start?. What kind of setup did you have?. How did you stay focus?, and for how long did you work on that?. Any failure stories you want to share?.

Most of it comes down do conviction and sticking to my not-so-promissing setup for 1 year, while every single of my peers were either thinking WTF are you doing, or telling me why my thing would fail. The rest is luck where we could scale way past what we though was possible.

I stayed focused by creating a penalty system where if I worked for less than 7 hours that day, I'd burn 1$ for every missing hour. I still maintain that spreadsheet with the self-debt over time (that I have yet to pay, but I will by burning crypto). Also my screen saver message is "get back to work"