Hacker News Board Hacker News, browsed like a classic message board
Live data from Hacker News

Elon Musk on First Principles

blog.microlancer.com

Post reply on HN
Print view View original

41–50 of 105 posts

Re: Elon Musk on First Principles

#41

Post by Double_Cast »

Radim wrote: post #29
redmaverick wrote: post #21

Earlier quoted context omitted.

According to Kasparov, Carlsen has a knack for sensing the potential energy in each move, even if its ultimate effect is too far away for anyone — even a computer — to calculate. In the grand-master commentary room, where chess’s clerisy gather to analyze play, the experts did not even consider several of Carlsen’s moves during his game with Kramnik until they saw them and realized they were perfect. “It’s hard to ex…

Now, that concept I'm familiar with. In fact, I think that's a rather common feeling, not even exclusive to chess or pros. Intuition (whatever that is) plays a great role, especially at the top level.

"Noobs use approximately 95% calculation and 5% intuition, whereas a grand master has a ratio closer to 40% calculation and 60% intuition."

I vaguely paraphrased this from a paperback chess guide of mine (hopefully more credible than the internet). According to my guide, grand masters can effortlessly memorize the positions of realistic matches. But for "nonsensical chess matches" where the positions were randomly generated, the masters were just as bad at recall as the noobs. From this data, my guide inferred that masters rely on recalling attack-patterns from their experience more than brute-force calculation. Similar to how a tourney-strength player has openings memorized by heart, I imagine that a master recognizes common middle-game patterns like the back of his hand.

Re: Elon Musk on First Principles

#42

Post by Florin_Andrei »

aridiculous wrote: post #31

Reasoning from first principles is ahistorical, and that is why it so rarely works in the real world. Political and social constructs inherited through time (like, say, the law) will butt heads against a "pure" new solution. You need the resources (and determination, will, etc) to overpower colossal systems we've arrived at through the progression of history. If you're going to hit homeruns like Musk, I think you mus…

Some qualifiers:

Working from first principles does work if you're super, super smart. One could argue that Special Relativity was almost a redesign from scratch to the problem of the aether.

In any case, it's never really a pure first-principles construct. That's impossible. Your mind is contaminated anyway with all the knowledge from, you know, the shoulders of all the previous giants, and you can't really 100% suspend it at will no matter how hard you try.

What matters then is the effort to dive in towards the physical (for engineering) or mathematical (for computing) roots of the problem, and re-arrange the causal links from that depth up.

It's always a blended approach, but you may not always be aware that that's what you're doing.

Re: Elon Musk on First Principles

#43

Post by jonnathanson »

ChuckMcM wrote: post #4

I'm a big fan of reasoning from first principles, but in practice it tends to work out more like 'total honesty' does with your spouse. Which is to say, not always all that great. I tend to be a reasoned sort of guy, this doesn't mean I don't say stupid things, I do, but when called on those things I can tell you how I reasoned to them and you can help me see the error in my reasoning. When someone can't reason from…

"You could do ok with the Socratic Method (asking questions that might lead them to understand where they were mistaken) but even that was dangerous if done in a group context"

This is very true, and I've learned it the hard way.

I do not advise using the Socratic Method to dissect the reasoning of a "needs-to-be-right" personality in a group setting, particularly if that person happens to be your superior. It's been my experience that these personalities will feel under attack, and possibly even attribute personal/political motivations to you.

When removed from a group (i.e., to your point, an audience), the Socratic Method works a lot better. It allows the recipient of the questioning to save face. He or she will be much more likely to engage in a real dialogue.

Re: Elon Musk on First Principles

#44

Post by dnautics »

logicallee wrote: post #39
aridiculous wrote: post #31

Reasoning from first principles is ahistorical, and that is why it so rarely works in the real world. Political and social constructs inherited through time (like, say, the law) will butt heads against a "pure" new solution. You need the resources (and determination, will, etc) to overpower colossal systems we've arrived at through the progression of history. If you're going to hit homeruns like Musk, I think you mus…

I don't agree: law, political, and social constructs are also amenable to arguments from first principles.

there is still an activation barrier (which is a chemistry first principle)

Re: Elon Musk on First Principles

#45

Post by pkinsky »

wwkeyboard wrote: post #36
aridiculous wrote: post #31

Reasoning from first principles is ahistorical, and that is why it so rarely works in the real world. Political and social constructs inherited through time (like, say, the law) will butt heads against a "pure" new solution. You need the resources (and determination, will, etc) to overpower colossal systems we've arrived at through the progression of history. If you're going to hit homeruns like Musk, I think you mus…

I think "standing on the shoulders of giants" is the cliche you're looking for. - Looking over the examples from the blog post, those aren't really from 'first principles'. They are assuming cars and an internet and other huge innovations. I think the real message should be 'double check your assumptions'.

>- Looking over the examples from the blog post, those aren't really from 'first principles'. They are assuming cars and an internet and other huge innovations. I think the real message should be 'double check your assumptions'.

I think the point is that you should view cars as a physical object composed of engineered subsystems instead of as a cultural phenomenon.

Re: Elon Musk on First Principles

#46

Post by possibilistic »

xzephyr wrote: post #34

The so-called reason by analogy is just gradient descent to approach local optima in optmization. The "by first principle" is just providing an approximate (ideally convex) model and solve it analytically for global optima. The problem is generally very hard (#P hard), both for formulating the problem and solving it. Global optimum of course in theory is better, but the quality of your objective function and constrai…

I love this explanation. Do you happen to blog about optimization topics?

Re: Elon Musk on First Principles

#47

Post by stinkytaco »

thangalin wrote: post #32
Aqueous wrote: post #19

Earlier quoted context omitted.

This works if the ultimate goal is to sign them up for a free account. But what if the goal is to convert them to a paid account after a certain trial period? A user can just clear their cache or cookies and use it free indefinitely.

In freemium models, some features are free. People should be able to use the app without submitting contact details. I did not state that all features had to be free. Enough features should be available that users can quickly understand premium account benefits. Deciding what features to make free vs. premium can be a difficult business decision. Then, unlock premium features for registered (paying) accounts. (Trial…

For the life of me I can't find the link right now, but it was from Pat and Bingo Card Creator. He requires an account for troubleshooting and support purposes. The vast majority of support is from first time/free users and when they did not create an account, support was a headache at best and impossible at worst. Perhaps there were other solutions for him, but if you can get people to create an account and it makes your life easier, why not?

Re: Elon Musk on First Principles

#48

Post by bernardom »

Florin_Andrei wrote: post #42
aridiculous wrote: post #31

Reasoning from first principles is ahistorical, and that is why it so rarely works in the real world. Political and social constructs inherited through time (like, say, the law) will butt heads against a "pure" new solution. You need the resources (and determination, will, etc) to overpower colossal systems we've arrived at through the progression of history. If you're going to hit homeruns like Musk, I think you mus…

Some qualifiers: Working from first principles does work if you're super, super smart. One could argue that Special Relativity was almost a redesign from scratch to the problem of the aether. In any case, it's never really a pure first-principles construct. That's impossible. Your mind is contaminated anyway with all the knowledge from, you know, the shoulders of all the previous giants, and you can't really 100% sus…

> Working from first principles does work if you're super, super smart.

It depends on the problem you're solving! In a former job, I worked for a consultancy that would be called in to solve industrial problems. Our strategy was to use first principles to truly understand how to best affect the metric we were trying to change- and it worked beautifully, and we built our entire business around it. Yes, we had to hire smart engineers, but not necessarily geniuses- because the problems we were solving weren't "how do I lower the cost of car ownership in this country," they were more like "why has the efficiency of this naphtha recovery plant decreased by 25% over the last year and how can we fix it?"

(I'll note that the fact that we were being called in constitutes a type of selection bias- the "typical" methods of solving the problem weren't working.)

EDIT: formatting

Re: Elon Musk on First Principles

#49

Post by bsdetector »

technical innovations that many thought impossible, like producing battery packs more cheaply than ever before.

Really? Many people thought that cheaper battery packs were actually impossible?

Maybe the rest of the article says something important, but when the second sentence is tripe like that I'm not going to stick around to find out.

Re: Elon Musk on First Principles

#50

Post by thangalin »

saraid216 wrote: post #38
thangalin wrote: post #32

Earlier quoted context omitted.

In freemium models, some features are free. People should be able to use the app without submitting contact details. I did not state that all features had to be free. Enough features should be available that users can quickly understand premium account benefits. Deciding what features to make free vs. premium can be a difficult business decision. Then, unlock premium features for registered (paying) accounts. (Trial…

So really, the answer is to use payment details as user information. That seems like a great idea. /s

Where did I say to use payment details as user information?
Post reply on HN