-
Notifications
You must be signed in to change notification settings - Fork 1
/
Learning-Efficiently.html
216 lines (204 loc) · 20.3 KB
/
Learning-Efficiently.html
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
79
80
81
82
83
84
85
86
87
88
89
90
91
92
93
94
95
96
97
98
99
100
101
102
103
104
105
106
107
108
109
110
111
112
113
114
115
116
117
118
119
120
121
122
123
124
125
126
127
128
129
130
131
132
133
134
135
136
137
138
139
140
141
142
143
144
145
146
147
148
149
150
151
152
153
154
155
156
157
158
159
160
161
162
163
164
165
166
167
168
169
170
171
172
173
174
175
176
177
178
179
180
181
182
183
184
185
186
187
188
189
190
191
192
193
194
195
196
197
198
199
200
201
202
203
204
205
206
207
208
209
210
211
212
213
214
215
216
<?xml version="1.0" encoding="UTF-8" ?>
<!DOCTYPE html PUBLIC "-//W3C//DTD XHTML 1.0 Strict//EN" "http://www.w3.org/TR/xhtml1/DTD/xhtml1-strict.dtd">
<html xmlns="http://www.w3.org/1999/xhtml" xml:lang="en" lang="en">
<head>
<meta http-equiv="Content-Type" content="text/html; charset=UTF-8" />
<link rel="icon" type="image/png" href="./images/favicon-32x32.png" sizes="32x32" />
<link rel="icon" type="image/png" href="./images/favicon-16x16.png" sizes="16x16" />
<title>Learning Efficiently - SPK's Rationality Essays</title>
<link rel="stylesheet" type="text/css" href="./css/default.css" />
<link rel="stylesheet" type="text/css" href="./css/highlight.css" />
<!-- <script src="http://ajax.googleapis.com/ajax/libs/jquery/1.10.2/jquery.min.js"></script> -->
<!-- <script type="text/javascript" src="/js/header-links.js"></script> -->
<script type="text/javascript" src="http://cdn.mathjax.org/mathjax/latest/MathJax.js?config=TeX-AMS-MML_HTMLorMML"></script>
<link href="atom.xml" type="application/atom+xml" rel="alternate" title="Sitewide ATOM/RSS Feed" />
<!-- Google Analytics stuff -->
<!-- Google tag (gtag.js) -->
<script async src="https://www.googletagmanager.com/gtag/js?id=G-DEWF2J5BG8"></script>
<script>
window.dataLayer = window.dataLayer || [];
function gtag(){dataLayer.push(arguments);}
gtag('js', new Date());
gtag('config', 'G-DEWF2J5BG8');
</script>
<script type="text/javascript" src="https://fast.fonts.net/jsapi/f7f47a40-b25b-44ee-9f9c-cfdfc8bb2741.js"></script>
</head>
<body>
<div id="header">
<div id="logo">
<a href="./">SPK's Rationality Essays</a>
</div>
<div id="navigation">
<a href="./">Home</a>
<a href="./notes.html">Notes</a>
<!-- <a href="/about.html">About</a> -->
<a href="./archive.html">Archive</a>
<a href="./atom.xml" type="application/atom+xml" rel="alternate" title="Sitewide ATOM/RSS Feed">RSS</a>
</div>
</div>
<div id="content">
<h1 id="post-title">Learning Efficiently</h1>
<!-- <center><img src="https://fbcdn-sphotos-d-a.akamaihd.net/hphotos-ak-prn1/t31.0-8/p600x600/10257116_10202295769100492_2438594605053717342_o.jpg" height="400" width="300" class="sujeet-pic" alt="Sujeet pic" /></center> -->
<p><strong>Entry Question</strong>: How can we learn more efficiently?</p>
<hr />
<h1 id="making-learning-efficient">Making learning efficient</h1>
<p>Change the textbook format.</p>
<p>First off, why the f*ck do I need to read all of 125k words in order to get what I want from a textbook?</p>
<p>Question: How can we <em>shorten</em> the time necessary to get what I want from a textbook?</p>
<p>In fact, what do I really <em>want</em> from a textbook?</p>
<p>Can’t I get that directly somehow, like snorting cocaine, just zoom the content straight into my head?</p>
<p>We saw what we want: we want to know the best hypotheses in the field (i.e., the models that work) and be able to make predictions using these hypotheses to solve problems.</p>
<p>The previous section described what a good textbook should give us. So, it’s a matter of making each of those things more efficient.</p>
<hr />
<p>Pre-load your past knowledge so that you can install the new results effectively.</p>
<p>Faith in the problem-solving process: Make it clear that once you understand the hypotheses, you can easily solve the relevant problems. The answers will come out trivially. The problem-solving process shouldn’t be magic. You should see plenty of evidence that the solutions follow naturally from the hypotheses and intermediate results you have learnt.</p>
<h1 id="what-stops-us-from-learning-efficiently">What stops us from learning efficiently?</h1>
<p>What do we want our learning process to help us with?</p>
<ul>
<li><p>Too much information load makes it hard to think further.</p>
<p>We feel the need to “clear our head”. Our learning process should account for this.</p>
<p>This is major. For example, it was only after I split this ongoing essay into two parts (this is the second part) that I was able to think about new ideas for it. Till then, I was kinda fatigued.</p>
<p>Making things <a href>succinct</a> will give us huge payoffs.</p></li>
<li><p>Poor memory</p>
<p>Not just about not being able to recall stuff we’ve learnt in this domain. We can’t remember stuff we learnt a long time ago. We forget whole areas of knowledge somehow.</p>
<p>However, keep in mind my essay on memory.</p></li>
<li><p>Other cognitive factors</p></li>
<li><p>Identifying problem types</p>
<p>“What kind of problem is this?”</p>
<p>or rather, “Which of the theories I know will help me solve this problem?”</p>
<p>For example, in some cases, it would be good to think economically - think at the margin, for example. Other cases, think scientifically…</p>
<p>I dunno… maybe you should always think scientifically. Maybe this whole domain-breakup thing is not helpful. No. It is necessary. You can’t handle the raw complexity of the world. I guess you need to simplify it into categories.</p>
<p>Anyway, we do learn about stuff organized by domain. So, we need to know what kind of theories (hypotheses) we should use for a particular problem we face.</p></li>
</ul>
<p>This seems to be a design problem. We need to design a learning process that takes into account our learning difficulties and our various requirements and gives us great performance.</p>
<h1 id="what-parts-of-traditional-learning-do-we-ditch">What parts of traditional learning do we ditch?</h1>
<p>Which aspects are crap and inefficient?</p>
<ul>
<li><p>Maybe mere fact-learning, without thinking about the hypothesis and hence about the anticipation-constraints. When you learn just facts (“gravity” -> balls reach ground together), you can’t constrain anticipation for other events.</p></li>
<li><p>Fact-learning = just learning about certain events and the factually correct predictions for those events. Or maybe some techniques for handling certain events. Plus, a “because” added in to make you think you understand the concept.</p>
<p>Far more important to learn about the hypothesis and learn to constrain anticipation for various events.</p>
<p>Mainly, a change of focus.</p></li>
<li><p>Remove the Dark Arts</p>
<p>Try not to bring in unnecessary emotions and other politics.</p>
<p>Don’t try to install false beliefs using Dark Arts (for example, by ridiculing some belief but not disproving it. See: <a href>How to Kill an Idea</a>)</p>
<p>Just show the hypotheses and the evidence for them. This is both for their good and for yours. <strong>You</strong> want to get to the truth, first and foremost. Forget about convincing people to get over to your side. Focus on finding the truth and winning.</p></li>
</ul>
<h2 id="what-new-actions-do-we-take-on-and-amplify">What new actions do we take on and amplify?</h2>
<hr />
<p>Not every change is an improvement, but every improvement is a change.</p>
<h1 id="how-can-you-test-your-learning">How can you test your Learning?</h1>
<p>See section above about true test of learning.</p>
<h1 id="how-is-learning-different-from-deliberate-practice">How is Learning different from Deliberate Practice?</h1>
<p>I think Deliberate Practice will help greatly in Learning (aka correct predictions from some hypothesis) by helping you cache and strengthen the intermediate results and techniques. This way you can infer the answers really quickly, which matters a lot when you need to make a sequence of correct inferences before you get to the answer.</p>
<h1 id="how-to-discover-the-most-useful-intermediate-results-to-cache">How to discover the most useful intermediate results to cache?</h1>
<h1 id="can-we-use-the-scientific-method-bayesian-thinking-to-help-in-solving-problems">Can we use the Scientific Method / Bayesian Thinking to help in solving problems?</h1>
<p>Or is normal problem-solving just about applying the rules and intermediate results correctly to reach the solution?</p>
<p>Can’t you use the Scientific Method to speed it up?</p>
<p>Ditto for Research. How do we use the Scientific Method there?</p>
<h1 id="the-theory-of-learning">The Theory of Learning</h1>
<p>Test all existing learning techniques against my theory.</p>
<ul>
<li><p>Thinking concretely</p>
<p>How does that help?</p></li>
</ul>
<h1 id="the-practice-of-learning-the-techniques-steps-whatnot">The Practice of Learning (the techniques, steps, whatnot)</h1>
<p>When encountering a new chapter, ask:</p>
<ul>
<li><p>What are the events for which I need to make predictions?</p></li>
<li><p>Can’t I do that with my existing knowledge?</p></li>
<li><p>What rent does this new concept (hypothesis-part) pay (on its own)?</p></li>
</ul>
<p>Well, if you’re just learning (not looking to do research), aren’t the first and third questions enough? Why strain to see if it can be done with existing knowledge? (Deliberate Practice side-benefits)</p>
<h1 id="practical-test">Practical test</h1>
<p>Test my theory out by going ahead and learning from some textbook</p>
<p>If you’re stuck on some problem, <strong>all</strong> it means is that the chain of inference is too long (or you’re confused).</p>
<p>No need to be ashamed (if you’re not confused).</p>
<p>So, build up the language to the problem. Build intermediate results (lemmas).</p>
<p>Find out exactly why you can’t solve it and build in that direction.</p>
<h1 id="make-it-easy-policy">Make It Easy Policy</h1>
<p>A good textbook should make explicit <em>all</em> the techniques experts use to solve problems. There shouldn’t be much mystery about how to solve the relevant problems. The issue is that we immediately assume we are <em>dumb</em> if we aren’t able to solve an exercise problem (or is that just me?). It seems like everybody else was able to solve the problem and only we were stuck. Again, the question of <a href>magic</a> comes up.</p>
<p>There should be <em>nothing</em> magical about how to solve the problem. Others were able to solve the problem because they had all the intermediate results loaded in their head and were able to make an easy inferential jump to the final solution. That’s all it is. Yes, it’s pretty hard to get those things ready at hand, but it’s not a mysterious process.</p>
<p>Why not give the readers the precise techniques you use to go all the way to the end of the solution? Yes, I know that the people who got the answers feel really smug about how they are better. (Maybe they don’t know how they did it. Cool. But try to find out. Don’t just let it slide so easily.)</p>
<p>Will it make the topic too boring for readers? Well, I’d bet there are more people who get disheartened and leave because they kept getting stuck on the problems than those who left the topic because it was too easy. Yo, if it’s too easy, f*cking level up and go solve some harder problems. Best case, if you think you’re so good, go solve research problems. People will thank you.</p>
<p>I don’t think we should hesitate to make the stuff as easy as f*cking possible for the reader. Save your displays of heroic intelligence for the research problems. Maybe not even then - “There is nothing honorable…”</p>
<p>Make the problem-solving as easy as possible!</p>
<p>Most textbooks will have problem sets at the end of each chapter. These will contain some easy, a lot of medium-difficult questions, and a few hard problems. These problems are all meant to be solvable if you “understand” the chapter’s concepts.</p>
<p>But how many of us can ever solve all the problems? Ditto for the problems in an exam. How many of us can solve all the problems in the paper, even after going back home and reading up on the topic?</p>
<p>From my experience, we just leave it at that. We assume, and I think the profs assume, that this is what it is.</p>
<p>Some people can solve most of the problems - they get the top grades. The others can’t - and they get the other grades.</p>
<p>Why don’t we aim to get everyone - every single person in the class - to be able to solve every damn last problem in the paper? I mean, it’s not like it has some hitherto unsolved research problems in there. They are all meant to be solvable. But they weren’t. It wasn’t all 100/100 in the class. Perhaps nobody got 100/100.</p>
<p>(Leave aside the time limit thing for a while. Assume you have lots of time. Given even the whole day, I wouldn’t have been able to solve all the problems on my college exams. Sometimes, you just have no f*cking clue how to solve a problem.)</p>
<p>Ok. Forget about everybody in the class. Let’s just focus on ourselves.</p>
<p>Why don’t we bring the concepts and techniques to such a goddamn understandable level that we are able to solve every problem in the textbook?</p>
<p><strong>Claim</strong>: You haven’t understood a concept well unless you are able to solve every problem in the textbook <em>with utter ease</em>.</p>
<p>The last qualifier is the key part. It should be a piece of cake. Not “I had to think hard for an hour and then <em>somehow</em> I got the answer”. Or “I did this and then that and then… some magic happened… and I got the answer”</p>
<p>You should be able to see exactly how you got the answer. It should not be a mystery.</p>
<h1 id="so-easy-it-should-be-banned">So easy it should be banned</h1>
<p>Here’s a helpful clue: when you’ve got it all loaded, it will seem so easy it will be like you’re <strong>cheating</strong>.</p>
<p>What would you think of someone who brought a slip of paper into the exam hall, containing the solution to a problem?</p>
<p>You would call him a cheater.</p>
<p>Ok. What if someone brought a slip of paper with the necessary results for the last step of the solution process? That is, they have to make one inference from these results to get to the final answer.</p>
<p>What would you say? Hey… that sounds quite like cheating. I mean, it’s not full cheating, like the previous guy, but still… not on the up and up.</p>
<p>Well, when you’ve got all the concepts fully loaded and freshly available in your mind, you would be like the second guy, if not directly like the first guy.</p>
<p>The answer to any question would be pretty much at one or two short jumps. It would seem like child’s play.</p>
<h1 id="effort-and-results">Effort and Results</h1>
<p>Do you feel that there is something not quite right with the above arguments?</p>
<p>I certainly do.</p>
<p>It feels like it <em>should not be so easy</em> to get great results.</p>
<p>It feels like it should take hard work, immense courage, and the perseverance of a champion to scale the mountains of hard problems. I mean, if there turns to be an easy way to learn things, then anybody and his great-grandmother could go around and solve hard problems.</p>
<p>I think this is called the Protestant work ethic. You believe that you will get great results <em>only</em> if you do great amounts of hard work. Somehow, results that you get easily will not last or give you what you want.</p>
<p>This is crap. The universe doesn’t check for the amount of effort you put in before it gives you the results as defined by the laws of physics.</p>
<p>If you do the right things, easy or hard, you will get the results you want. Effort does not matter directly.</p>
<p>I have this vague feeling in my mind that if I want to get great results, like the great scientists and thinkers, then I <em>need</em> to do tremendous amounts of hard work.</p>
<p>It feels like if I have to solve hard problems, I need to create insanely long and hard chains of inference. I need to have <strong>super-duper memory power</strong> and hold all the 153 results at the same time in my mind to arrive at the final answer. I need to have super-fast inference skills where I immediately see the answer to some question I come across. And then, I need to have epic Will Power to sit at my desk and plough through problem after problem from some hard exercise set. And, worst of all, I need to have insane amounts of “creativity” and “intuition” so that I can pluck correct hypotheses out of thin air.</p>
<p>I have none of those things. No wonder it seems impossible that I will ever be able to solve problems easily.</p>
<p>No. I don’t think that’s how it works. We all have limited brain-power. Our mind can only run 100 different steps for a given operation.</p>
<blockquote>
<p>In neurology, the “hundred-step rule” is that any postulated operation has to complete in at most 100 sequential steps.</p>
<p>– <a href="http://lesswrong.com/lw/k5/cached_thoughts/">Cached Thoughts</a>, Eliezer Yudkowsky.</p>
</blockquote>
<p>So, you’re not gonna run a much bigger inference chain when you’re an expert. You’re just gonna do it more efficiently. You’re gonna cache the important results so that you just have to run the last few steps of the inference chain. You’re gonna do what our friend, the cheater, did on his exam.</p>
<p>Yes, experts remember more things and have more knowledge of the field than n00bs, but it’s not like they can hold a thousand things in their mind either. The way I understand it, your mind chunks up data so that you manipulate the same <em>number</em> of chunks as you always did, just that the chunks are way bigger in themselves. So, you don’t need to have the super-duper memory power of remembering the entire sequence of cards during blackjack or whatever. Not even the experts card-players do it like that. They do it far smarter.</p>
<p>Creativity and intuition are mysteries to me now, so I won’t comment on them. As for Will Power, I don’t think the great achievers procrastinate any less than normal people, most probably more. They just <a href="http://paulgraham.com/procrastination.html">procrastinate on crap</a> and focus on the important stuff. And it’s not their Will Power pushing them up the hill, it’s the <a href="./The-Hacker-Way.html">delight</a> of solving the problem and learning new things.</p>
<p>So, your assumptions were all false there, SPK.</p>
<p>Use your learning process to make problem-solving so easy that it feels like cheating.</p>
<hr />
<h1 id="notes">Notes</h1>
<ul>
<li><p>Perhaps check out the Samford Prof Chow’s videos on Deep Processing and stuff</p></li>
<li><p>What is “intuition”? How can we use that to shape our textbooks and learning processes?</p></li>
<li><p>I think “creativity” will play a huge part in both making inferences from existing hypotheses and in searching for a new hypothesis to solve some problem.</p>
<p>How to deal with this? How <em>exactly</em> does “creativity” help out here? How can we get more of it?</p></li>
<li><p>What about Attention and Motivation?</p>
<p>I’ve heard Attention matters big time for learning something - it helps build those circuits in your brain.</p></li>
<li><p>Don’t try to make a Plan for learning efficiently. Remember the <a href>Problem with Plans</a>.</p></li>
<li><p>See if we can use ideas from the <a href>Hacker Way</a> in our learning process.</p></li>
</ul>
<div class="info">Created: November 22, 2014</div>
<div class="info">Last modified: August 6, 2015</div>
<div class="info">Status: finished</div>
<div class="info"><b>Tags</b>: learning, anticipation constraint</div>
<br />
<div id="disqus_thread"></div>
<script type="text/javascript">
/* * * CONFIGURATION VARIABLES: EDIT BEFORE PASTING INTO YOUR WEBPAGE * * */
var disqus_shortname = 'spkrationalitytrainingground'; // required: replace example with your forum shortname
var disqus_identifier = '/Learning-Efficiently.html';
var disqus_title = 'Learning Efficiently';
/* * * DON'T EDIT BELOW THIS LINE * * */
(function() {
var dsq = document.createElement('script'); dsq.type = 'text/javascript'; dsq.async = true;
dsq.src = '//' + disqus_shortname + '.disqus.com/embed.js';
(document.getElementsByTagName('head')[0] || document.getElementsByTagName('body')[0]).appendChild(dsq);
})();
</script>
<script type="text/javascript" src="https://fast.fonts.net/jsapi/f7f47a40-b25b-44ee-9f9c-cfdfc8bb2741.js"></script>
<noscript>Please enable JavaScript to view the <a href="http://disqus.com/?ref_noscript">comments powered by Disqus.</a></noscript>
<a href="http://disqus.com" class="dsq-brlink">comments powered by <span class="logo-disqus">Disqus</span></a>
</div>
<div id="footer">
Site proudly generated by
<a href="http://jaspervdj.be/hakyll">Hakyll</a>
</div>
</body>
</html>