nimithryn

Solution to ... what? I work in this area - I don’t think a “monorepo” is going to solve the numerous difficult problems the US has with modern multi-domain ISR. How do you represent and compare different capabilities from different domains? How do you pick which assets to buy? How do you pick which theatre to deliver them to? How do you plan tactical deployment of your assets? How do you dynamically retask assets in the event of disruption by the enemy? How do you fuse the numerous different data formats to interpret the multi-modal data you receive from your ISR assets? How do you retrain users to use the new capabilities? How do you maintain the assets for a long period of time without cost overruns? How do you get the systems to interface with legacy systems? How do you deal with all the intra-military political issues? These are just a few issues faced by the DoD in the ISR space.

I think there’s a bit of Chesterton’s fence in general around the DoD - the military (and the government in general) has many inefficient, slow, and expensive practices, but while it sounds like a good idea on the surface to tear it up and start from scratch, people underestimate the difficulty and danger of hastily changing or defunding things without understanding why they were put in place.

Not to mention that there's way more video of Obama speaking than most people

If anything it's the reverse: "NASA: 90% of our rocket scientists use the rocket you designed, but you can't fix an AC so f* off"

This doesn't surprise me at all. I'd posit that it is only in the interest of these "gurus" to reveal their stock picks when they are ready to unload them - i.e. when they think the opposite of whatever they are predicting is going to come true.

Look into OpenCog - there's some sheaf-theoretic NLP stuff going on there. There's a recent high-level overview by Linas Vepstas you can find on the ArXiv somewhere. There's a project called SheafSystem also, which is a sheaf-based database for scientific computing (I've never used it). I have some ideas I'm working on in this area also (not affiliated with these parties in any way, and the ideas are not ready to share yet, unfortunately.

What's your background, out of curiosity?

Ah yes I am very familiar with ILP - thanks for sending these references!

>>>Don't expect the industry to lead this effort, though. The industry sees the reliance on large datasets as something to be exploited for a competitive advantage.

This is only true for the Facebooks and Googles of the world. There are definitely small companies (like the one I work for) trying very hard to figure out how to build models that use less data because we don't have access to those large datasets.

The industry is larger than just the Big N.

Try watching a movie version - I really like Branagh's Hamlet and I never expected to (long though)

Curious what you're working on and if you have any specific references - I'm also interested in fusion of classic symbolic AI with deep learning

There's also UMAP, which is new but looks promising.

Yeah it's most obvious that there is some natural/genetic talent at play in sports like swimming/track and field. I was pretty obsessed with running in high school - I trained damn hard, did 80 mile weeks in the offseasons, etc. I ended up being pretty good, but there were some guys who trained less hard and less smart than me that were just as good (or better), and I certainly was not capable of becoming an elite runner no matter how hard I trained. It is just not in my genes to run a sub-4:00 mile or whatever. While I think it's comforting to think that all natural talent is just hard work and we could all be geniuses and savants if we just put in the effort, I don't think it's true. That being said, I do think people overestimate the difficulty of picking up many skills to a proficient or superior level.

I am not convinced he is right about how susceptible people are to the black swan effect: it seems like your average person overvalues tail probabilities, not undervalues (hence why people play the lottery and buy insurance). It's been a while since I read anything by him, but I remember not being convinced by how he dealt with the problem of induction either, which seemed to lean dangerously close to intellectual nihilism.

Edit: grammar

Plenty of mathematicians are older I'm sure. I know Wiles was 40ish when he proved Fermat's Last Theorem.

This article argues that these companies want physicists to handle the abstract math and numerical programming that machine learning requires. But physicists don't have a monopoly on those skills; they are taught in plenty of engineering disciplines and applied math programs. I would even argue that as long as you can handle the math, you are better off with an education focusing on an engineering discipline so you can learn the design aspects of building large pieces of infrastructure as well.

Two methods I always use:

1. Google the article title and click on the search results to reach the article. 2. Refresh the webpage; as the page loads, hit the x in the browser to kill the loading. Sometimes this loads the article but not the paywall.

The paper (which admittedly I only skimmed, so someone correct me if I'm wrong) seems to indicate that only around 70% of the variance based on a simulation (and I believe historical data) was due to fishing. The remainder could very well be ontogenic. So it's entirely possible that fishing is only one factor in this phenomenon.

The problem with this is that the various cryptocurrencies are all very correlated with each other. So having an "index fund" won't achieve much diversification, unlike a real index fund that has stocks in different industries.

Or a confounding variable: high achieving men work hard in the gym and in their relationships.

Or people who are more likely to commit to long term exercise plans are also more likely to commit to long term relationships? Like, maybe it's a discipline thing. Are people with advanced degrees and skills that take discipline to maintain also more likely to be married?

You may be confusing the distribution of the means of the samples (which will be normally distributed if the Central Limit Theorem holds) with the sample itself (which will not generally be normally distributed).