Climbing the wrong hill | Hacker News Reader

Climbing the wrong hill

cdixon.org

56 pointsprakash15 comments

That's a terrible analogy. Good analogies should provide an honest insight into a complex situation using a simple imaginary construction. In this case, there's no reason to assume that anyone's objection to founding a startup has anything to do with short-term versus long-term goals, and even if there was, the analogy of the "hill problem" doesn't address that objection at all.

I think it's much more reasonable to guess that this guy's reluctance to do the startup dance is the result of more complex considerations. His current job is safe; not glamorous, sure, but safe. He's probably making good money, and even if he doesn't really enjoy it, he's just found out that other people actually want him. That can be a powerful motivator by itself.

The startup thing is risky. Talking about it is all well and good, and we can make startups sound really sexy because they're what we prefer or need to do, but that doesn't mean that they're really for everyone. People might like the idea of startups, but when it comes to the actual risk, guesswork, and extreme hard work associated with them -- not so much.

I mean, hell, I'm working on my second one now, and I couldn't imagine working for anyone else ever again (or sitting in a cubicle). On the other hand, there are definitely days that I would love to not have to worry about whether or not I'll find enough business to pay the rent a couple of months from now. I haven't had a real get-away vacation in I-don't-know-how-long, and my average work day is probably around 11 hours long. I honestly don't know how my girlfriend puts up with it. And, I'm constantly thinking about one business or the other; it's not like I can just "punch out" at 5 or 6 o'clock each day and not think about work until the next morning.

So, yeah. Startups aren't all roses, and maybe some people are smart enough to figure that out before they jump into them.

I agree that the connection to startup v. career just isn't there, but I really like the analogy insofar as it concludes that it's worthwhile to "meander some in your walk (especially early on)," and "randomly drop yourself into new parts of the terrain."

That randomly drop, fits pretty well with some old advice to occasionally try something REALLY different, if only as a hobby or on a vacation, to help keep yourself flexible.

So is he proposing simulated annealing for finding the best occupation in one's life ? http://en.wikipedia.org/wiki/Simulated_annealing

A person looking for for purpose in life might have much more limitations than a hill climbing algorithm. He/she also has different advantages than such algorithm, so I do not think this is very good analogy.

One crucial advantage is that we have the examples of others to follow or avoid. In this respect, particle swarm optimization (as referred to in another comment in this post [1]) may be a better model.

[1] http://news.ycombinator.com/item?id=832163

The actual algorithms wouldn't be useful, but the general idea of a "field" with a lot of local optima that you have to choose between or navigate around is a useful visualization to help you think about where you are going versus where you want to be going.

Reminds me of this: http://sethgodin.typepad.com/seths_blog/2005/11/understandin...

This reminds me, I have written a fairly good, fairly generic hill-climbing "library" for Python. I did it because I needed something specific quickly, and it's gradually grown over time. I haven't found anything generally available that does the same thing, and I was wondering if anyone here can suggest some existing libraries/modules that I can compare/contrast with before thinking about tidying up and releasing.

Suggestions?

Yes, I can certainly point you in the right direction.

Toby Segaran, author of the highly esteemed and good-read Programming Collective Intelligence (published by O'Reilly, so you know it's good), released all of the source code that he talked about in the book. You can find it all here on his blog: http://blog.kiwitobes.com/?p=44.

He talks about hill climbing and variants (random restart, simulated annealing). He does a great job of explaining the algorithm and the math behind the hill climbing decisions. For optimization, something hill climbing is not even the best tool. Look at some of the other optimization strategies he employs (specifically the genetic mutation algorithm).

I highly recommend buying the book, even if only one chapter currently interests you. I bet you the other chapters will too. It's on Amazon, and at B&N if you want it this second.

Brilliant - thanks - I'll check that out.

I do know that hill-climbing isn't always the best strategy - I've got experience. What I'm lacking is knowledge of people from other disciplines doing things in an organised and more complete manner. This is a good start.

I've also found AIMA stuff on Google - http://code.google.com/p/aima-python/ - which claimsto be a Python implementation of algorithms from Russell and Norvig's 'Artificial Intelligence: A Modern Approach'

I haven't done much of a survey, but I stumbled into this and looks promising:

http://en.wikipedia.org/wiki/Particle_swarm_optimization

I found this book a very enjoyable read (edit: and a lot less intimidating than the Wikipedia article!). Be warned, though, that it's one or two chapters worth of technical detail. The rest goes to motivating the subject and exploring its philosophical implications. And that is OK, because the algorithm is actually quite simple.

http://tinyurl.com/lns85a

I liked the 'The Dip' explanation way better: there are some things that are worth doing and others that are not. Most people do a lot of worthless stuff just because doing the worthy stuff is harder, even though it might yield a higher return in the long run.

* Imagine you are dropped at a random spot on a hilly terrain, where you can only see a few feet in each direction (assume it’s foggy or something). The goal is to get to the highest hill.*

Who climbs a hill they don't know in the fog, starting already lost and without navigation kit? Silly computer scientists.

I assume the "you are dropped" part is involuntary. The hill-climbing part just makes sense, so you can get high enough to see where the heck you are.

This might actually be a viable analogy if the hills were "fuzzy" - i.e. changed in height and location over time. Because careers are like that - often the highest hills get worn down, or moved to other countries.

Reply on news.ycombinator.com