Hacker News Reader

Comment by majos | Hacker News Reader

This post sorely lacks evidence for its big first claim:

> I've seen a similar pattern in many different fields: even though lots of people have worked hard in the field, only a small fraction of the space of possibilities has been explored, because they've all worked on similar things.

Anyone want to step in with some examples? Without them, the thrust of the essay seems to be: "If only other people understood what problems are worth working on! Especially in the well-studied areas of essays, Lisp, and venture funding! Too bad they do not. Well, goodbye."

View on HN

There is a lot of evidence for what Paul says once you dig into a specific field. Taking two fields I know well: in physics, there was decades of work on fundamental questions on systems in equilibrium, while many obviously important open questions in out-of-equilibrium systems went neglected until the last 10 years or so when there's been a huge upsurge. These questions were known to be open 20 and 30 years ago, but just weren't as fashionable as a topic. Anyone senior enough in those earlier decades knew there were tons of open questions but also that relatively few people were working on them for whatever reason.

In machine learning, there are currently a lot of people working with neural networks, but relatively fewer people exploring alternative model architectures. So much so that issues specific to neural networks sometimes get framed as fundamental to machine learning itself. I'm personally exploring an alternative class of models called tensor networks with many possibilities for research directions and lots of open questions but only a handful of people work on them. One reason for working on a popular idea is that it's nice to work on a topic where you have many colleagues and know in advance that your model is likely to give good results on challenging datasets.

I know next to nothing about physics, but I do know some about ML.

I think the reasons tensor networks are unexplored is interesting: The tools and techniques for dealing with them build on those for neural networks and the theoretical benefits are not clear cut enough to gain them a foothold over the practical results of neural networks.

Agreed. There is still a lot of work left to do!

Forgot to mention, but here is a recent theoretical result (prediction of generalization performance giving size of training set) based on tensor networks: https://itensor.org/miles/GenerativeMPS.pdf

I am wondering how far the essay would go if it were not from PG.

Probably not very. I get the distinct feeling that the last few essays were very light on content/interesting insights, and did well on name value alone.

Would be interesting to actually test that to be honest. Get Paul Graham to set up another site under a fake identity, post the next say, three essays there, then submit 'em to HN under the same identity and see the stats.

In some cases it’s because the territory is mostly mapped by his previous work. “Novelty and Heresy” was a belated sequel to “What You Can’t Say”, for instance.

OTOH, seeing pg write essays again pleases me in much the same way that one might be pleased if their favorite long-defunct rock band started recording a new album.

Is this even an essay, or am I missing something? The content behind the link is a few sentences at most. Hardly an essay.

This is basically my biggest critique of PG and various other YC leaders. Many of the conclusions, while likely be probably being more "right" than "wrong" in virtue, are derived from intuition and observations, not evidence based. It's even more ironic, given how much emphasis the firm places on evidence based thinking of its portfolio founders versus anecdotally thinking.

Great example:

"One quality that’s a really bad indication is a CEO with a strong foreign accent"[0]

The danger is that I think pg has been "right" about so many things that every time he pontificates about something it's treated as dogma. So when he's "wrong", it'll be ignored. Impressionable people (which likely fits the characteristic of many young tech entrepreneurs) will therefore be lead astray.

[0] - https://www.forbes.com/sites/knowledgewharton/2013/12/19/292...

To start with, it would be called a tweet and not an essay.

Haha, I was wondering the same. Probably not that far. But at the same time, it is important who says what, so you can't just ignore that.

Two areas where I have a little experience:

Economics: There are surprisingly little people studying very large, grand, topics such an inequality. You've probably heard of the handful of economists that do. It's unfashionable due to the influence of the Chicago school and it's focus on free market principles, as well as many other factors.

Physics: Good luck getting any interest or funding if you are not studying the currently dominant theory in your field (even if there has been no progress in decades).

Okay, but to play devil's advocate -- if you study something controversial and/or unfashionable in academia (such as economics of inequality) aren't you just going to be ignored by the community?

I think being unfashionable is more viable in fields that are meritocratic and don't require popular approval to succeed.

I have this suspicion that, in economics, trends and fashions have sometimes formed around principles whose primary attraction is that they make the math tractable.

There are a large number of people working on inequality.

The two most common fields studying it are: 1) Development 2) Labor Although there are a number of Macro-Economists who have studied the impact of inequality on growth. The reason you may not read about it is that the relationship between inequality and growth is a largely "solved" problem. It is "unfashionable" because of that.

Is it? Can you link me to an econometric review?

The fact that the field thinks it is a solved problem is a marvelous example of what the essay is talking about!

Not exactly. It is a political problem not an economic one.

I genuinely don't understand where these ideas about economics come from. Inequality is a hot topic, and it has been for years.

Both examples are wrong. Lots of economists working on inequality , it is even a hot topic. Your physics example is so vague that is meaningless. Name one promising theory/approach in physics which is not getting interest because it goes against "the currently dominant theory of the field".Tip: LQG does not count.

> Name one promising theory/approach in physics which is not getting interest because it goes against "the currently dominant theory of the field".Tip: LQG does not count.

Lack of such promising theories is precisely the point (I guess we are talking specifically about the problem of quantum gravity). Everyone is working with the same 1.5 old approaches and the progress has stalled. Of course individually this strategy is rational because if you try to think of something new and it doesn't work out (which is very probable) your career is toast. But imagine what the brainpower poured into string theory could achieve with a more breadth-first search strategy.

No, lots of economists don't. Especially considering how important of a problem it is. It's a "hot topic" because of public opinion at the moment. We have not been attacking it in new and creative ways for decades.

In mathematics I'd put forward the conjecture that "for every proven theorem you could ask at least 5 more similar questions which are unproven."

For example it is proven there are infinitely many primes. Are there infinitely primes that differ by 2? By n for any n? Are there infinitely many palindromic primes? Are there infinitely many primes of form n^2 + 1? Is there always a prime between n^2 and (n+1)^2?

If this is true then, assuming there are 200k proven theorems, there would be >1m unproven but readily stated theorems which would mean it wouldn't be too hard to find areas no one is looking into.

>Are there infinitely primes that differ by 2?

Pardon the digression, but the Twin Primes conjecture was proven by a Subway restaurant worker a few years ago.

https://www.google.com/search?q=Twin+Primes+conjecture+was+p...

I'm not sure, I'm not an expert, it says here the twin prime conjecture itself is still unproven.

"On April 17, 2013, Yitang Zhang announced a proof that for some integer N that is less than 70 million, there are infinitely many pairs of primes that differ by N."

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

It bears notice that the Subway worker, Yitang Zhang, had a math PhD from Purdue.

When I was reading the essay I was immediately reminded of the concept of paradigm shift (https://en.wikipedia.org/wiki/Paradigm_shift). The gist is that during periods of "normal science" most scientists are working within the framework of the dominant paradigm which, among other things, determines what kind of problems are worth working on. But every once in a while the paradigm shifts (which happens comparatively rarely, on the time scale of decades), old problems are deemed irrelevant and everybody piles on to work on the new stuff. The original example is physics but other fields are pretty similar (in our field current AI craze comes to mind).

I might be overgeneralizing, but I feel like 10-15 years ago everyone wanted to do wireless communications, now everyone wants to do AI, and probably in a few years everyone wants to do Blockchain or something else. I kinda find it hard to believe that those people are all passionate about the field, not the massive career opportunities.

> This post sorely lacks evidence for its big first claim

Agreed.

I think the premise is true, but PG has given no actual insight.

NGO's are a good anti-example. In developing countries we know of working solutions but fashionable rules. I like computers so I'll give them to the poor.

Elon is a good working example, rather than looking at existing tech in space craft he looked at how to make them cheaper using steel and methane.

There's a Just So Story about rapid prototyping that I think fits how to break the mold. I can't remember which competition, but the winner won by working out how to cheaply rapidly prototype working models rather than tackling the problem head on.

So they didn't so much not be fashionable but created a system that allowed for non fashionable ideas.

> This post sorely lacks evidence...

It’s a blog, not a peer reviewed scientific journal.

String theory.

Reply on news.ycombinator.com