Hacker News Reader

Comment by JoeAltmaier | Hacker News Reader

That's not enough. Learning what doesn't work is sort of useful, but only if it ever leads to something that DOES work. There's an infinite supply of dead ends out there. You need a higher bar than 'learned something' if that something is a negative.

View on HN

It depends what question you are asking. If the question is "should we fund projects even if we don't know if or when they will have demonstrable results?" then I would argue that, yes, some portion of research should go towards that.

How long is "long enough" to deem that a line of experimentation "didn't work out?" There is no period long enough. Sometimes a negative result is quite useful down the road. Some research just comes together when the right things are learned and tried. So any notion of a "dead end" is really just a tentative assessment, frozen in time.

Without harsh assessments, a massive project will eventually dedicate some of its resources into self-perpetuation, that is justifying its existence rather than producing results.

It's a risk we need to take (or mitigate) if we want to have a steady flow of new breakthroughs. Science, just as programming or any other creative discipline, is best done when you have more money than you need and nobody is looking at your hands.

Well, we are talking about a perpetual motion machine. Would only make sense to dogfood it at all levels.

I'm not a scientist, but I am a programmer, and I find that learning new things is very often a process of eliminating dead ends through trial and error. That's how you narrow it down to the thing that actually does work.

It sounds like you're saying "don't bother exploring dead ends, either do it right the first time or give up," which is a lot like saying "I just want a penthouse apartment, stop wasting time building a foundation."

Yes, I think what you are saying can work well for programming. But with certain scientific questions, there are many theories that could work. It isn't just a matter of ruling things out (though that helps too). One key part of the scientific process is generating a theoretical insight (perhaps) that lets you look at something in a new way. Another key part is figuring out how to test that idea.

> You need a higher bar than 'learned something' if that something is a negative.

Not really. The point of big scientific research is that at the moment a project is being conceived, no one has any idea where it will go and whether it will bring anything useful. Yet in time, it always brings breakthroughs.

I'll put it bluntly: we need more trust. The right way to do science is to keep throwing money at it, no questions asked[0]. IMO the reason we've seen so much progress during last two wars was not because war per se stimulates research, but because during a conflict the military has infinite budget and it starts throwing more money at research that can be used, so scientists can do whatever they want without justifying it for general public[1]. Free of money&management issues, creative minds could focus on doing fundamental research in whatever they felt like doing[2].

Come to think of it, science is best done in the same conditions the programming (or any creative endeavour) is - when you're free to think, don't feel the pressure of looming deadlines, don't have to deal with managerial bullshit, and when you have enough money to not have to think about it (both in life and in cost of work materials).

So yes, I advocate some kind of almost blind trust. Just throw some part of taxpayer money at R&D, no questions asked. It's a long-time investment.

[0] - Of course there's a need to build in a system to limit the amount of parasites who will feed on free money; but they ultimately cannot be avoided, we can only try to limit their numbers.

[1] - I think it was Feynman who pointed out that the problem with science in the 90's was that scientists suddenly found themselves accountable to general public and struggled with explaining why their research was useful.

[2] - If that last paragraph rings bad to you, please note that I'm skipping over a lot of points related to creativity needing to be free of trivial bullshit, intrinsic motivation beating monetary rewards/prestige in performance, etc.

No. Feynman actually criticised scientists who weren't prepared to explain and justify their work to the public. He says this in one of the bbc videos, can't remember which, as part of a critique of the social sciences.

Standing on the shoulders of giants is many times a means of saying that something that "DOES work" was derived from something that didn't work, or perhaps that it was derived from the difficult yet iterative steps of others.

Look at the Large Hadron Collider. Do you suppose they scrub from their data stores all tests that fail to produce a "working" result? Are you suggesting that trying and failing produces irrelevant knowledge?

>You need a higher bar than 'learned something' if that something is a negative.

Why?

I would assume that they are only working with a narrow subset of dead ends. They probably have promising leads and work on those. I think it's a constant process of refinement.

Reply on news.ycombinator.com