-
Notifications
You must be signed in to change notification settings - Fork 0
/
Copy pathinsights2.html
196 lines (181 loc) · 9.47 KB
/
insights2.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
<!DOCTYPE html>
<html lang="en">
<head>
<title>Revisiting the Insights model - Median Group</title>
<link rel="stylesheet" type="text/css" href="/theme/css/styles.css" />
<link href="https://fonts.googleapis.com/css?family=IBM+Plex+Serif"
rel="stylesheet" />
<link rel="shortcut icon" type="image/x-icon" href="/theme/images/favicon.ico">
<meta charset="utf-8">
<meta name="viewport" content="width=device-width, initial-scale=1">
<meta name="description" content="Note: This demo is in beta, and you may experience issues such as strange numerical behavior at this time. Last year, we released our insights-based model that generated a projected timeline using historical data and a prior distribution. We’ve revisited it to address its limitations and improve the data …"/>
</head>
<body>
<div class="container">
<div class="header">
<div class="logo-wrapper">
<a href="/">
<img src="/theme/images/logo.png" class="logo" alt="Median Group logo"/>
</a>
</div>
<div class="headline">
<div class="title">Median Group</div>
</div>
</div>
<div class="main">
<nav class="navigation">
<div class="links-div">
<ul class="links">
<li class="nav-item"><a href="/">Home</a></li>
<li class="nav-separator"></li>
<li class="nav-item"><a href="/research">Research</a></li>
<li class="nav-separator"></li>
<li class="nav-item"><a href="/archives">Blog</a></li>
<li class="nav-separator"></li>
<li class="nav-item"><a href="/team">Team</a></li>
<li class="nav-separator"></li>
<li class="nav-item"><a href="/contact">Contact</a></li>
<li class="nav-separator"></li>
<li class="nav-item"><a href="/donate">Donate</a></li>
</ul>
</div>
</nav>
<main class="content-container">
<div class="content">
<section id="content" class="body">
<header>
<h1 class="entry-title">Revisiting the Insights model </h1>
</header>
<div class="entry-content">
<link rel="stylesheet" type="text/css" href="/theme/css/nouislider.min.css">
<script async src="/scripts/nouislider.min.js"></script>
<script async src="/scripts/insights2.js" type="module"></script>
<section id="intro">
<p>
<small>
Note: This demo is in beta, and you may experience issues such as strange numerical behavior at this time.
</small>
</p>
<p>
Last year, we released our <a href="http://mediangroup.org/insights">insights-based model</a> that generated a projected timeline using historical data and a prior distribution. We’ve revisited it to address its limitations and improve the data it draws from.
</p>
<p>
The model relies on the assumption that progress in <span class="caps">AI</span> relies on accumulating <em>insights</em>, fundamental advances in our understanding, that allow for improvements in capacity without increase in resources expended. This choice makes an attempt to separate out the effects of true technological advancement from the effects of an increase in computing power devoted to a problem, both of which can increase the capacity of machine intelligence to solve complex problems. Computational power is an expensive, finite resource, and without a paradigm-shifting improvement in computing itself, precise allocation of that power alone will not be enough to continue advancing <span class="caps">AI</span>’s problem-solving capabilities.
</p>
<p>
The interactive model below provides two methods of capturing a prior about how many more advances in understanding are required to achieve human-level machine intelligence. Based on that prior, and on the pace of insight discovery during a particular historical period, we compute a probability distribution over time of the likelihood humans will develop human-level <span class="caps">AI</span>. Results of this calculation are shown in the “Implied timeline” graph below.
</p>
</section>
<section id="dist">
<h2>Step 1: Specify a prior for current progress</h2>
<h3>Option A: Draw a distribution</h3>
<p>
For various “percentages of the way” done <span class="caps">AI</span> research could be, in terms of percentage of the necessary insights discovered, what is the probability that <span class="caps">AI</span> research is not yet that percentage done?
</p>
<p>
The graph below allows you to draw a distribution of how likely it is we have achieved a particular portion of the insights required for human-level machine intelligence.
</p>
<canvas width="500" height="500" id="progress_dist"></canvas>
<div style="width: 500px; text-align: center;">
<button id="reset">Reset</button>
</div>
<h3>Option B: Pre-set priors from Pareto distribution</h3>
<p>
Instead of drawing a cumulative distribution function, you can instead use a pre-set prior based on a <a href="https://en.wikipedia.org/wiki/Pareto_distribution">Pareto distribution</a>.
</p>
<p>
To make the choice of Pareto distribution more intuitive, we parameterize the distribution in terms of a probability <em>q</em>, equal to the probability that a doubling in number of insights (starting from the minimum number of insights) would result in a sufficient set of insights. <em>q</em> can be set directly, or we can sample from a mixture of Pareto distirbutions, where the <em>q</em> parameters are sampled from a uniform distribution or a beta distribution.
</p>
<input type="number" value="10" min="1" id="min_insights" hidden="true">
<div>
<h4><label for="num_samples">Number of samples to take when running the simulation</label></h4>
<input type="number" value="5000" min="1000" max="20000" step="1000" id="num_samples">
</div>
<div style="display:flex; justify-content: space-between; flex-wrap: wrap; text-align: center;">
<div>
<h4>Set <em>q</em> directly</h4>
<label for="pareto_q"><em>q</em></label>
<input type="number" value="0.2" id="pareto_q" step="0.05" min="0.05" max="0.95">
<br>
<button id="pareto_btn">
Set <em>q</em>
</button>
</div>
<div>
<h4>Sample <em>q</em> uniformly over (0,1)</h4>
<button id="uniform_pareto_btn">
Sample <em>q</em>
</button>
</div>
<div>
<h4>Sample <em>q</em> with Beta(α, β)</h4>
<label for="alpha"><em>α</em></label><span>:</span>
<input type="number" value="1" id="alpha">
<br>
<label for="beta"><em>β</em></label><span>:</span>
<input type="number" value="1" id="beta">
<br>
<button id="beta_pareto_btn">
Sample <em>q</em>
</button>
</div>
</div>
<p>
<small>
Note: The simulator can be very slow for larger values of <em>q</em>, as most of the samples need to be thrown away.
</small>
</p>
</section>
<section id="insights">
<h2>Step 2: Specify pace of progress</h2>
<p>
Which period in history is most representative of the future pace of <span class="caps">AI</span> insight discovery?
<p>
The graph below plots the aggregate of insights discovered over time and allows selection of a particular period of history in <span class="caps">AI</span> research. The curve fit to that period (linear, exponential, or sigmoidal) is used to project the future distribution of discoveries.
</p>
<canvas width="500" height="300" id="insights_plot"></canvas>
<div style="width: 500px;">
<div id="year_slider"></div>
<output id="year_range" style="display: block; text-align: center; margin-top: .5em;">
</output>
</div>
<br>
<label for="regression">Regression mode:</label>
<select id="regression">
<option value="Linear">Linear</option>
<option value="Exponential">Exponential</option>
<option value="Sigmoidal">Sigmoidal</option>
</select>
</section>
<section id="timeline" style="margin-top: 3em">
<h2>Result: Implied timeline</h2>
<canvas id="timeline_plot" width="500" height="300"></canvas>
<div id="last_year" style="width: 500px;">
</div>
</section>
<section id="Sources">
<h2> Sources </h2>
<p>
The data used in this model is available as a <a href="http://mediangroup.org/docs/insights.json"><span class="caps">JSON</span> file</a>.
The <a href="https://github.com/Median-Group/insights2">source code</a> for the demo can be found on the Median Group <a href="https://github.com/Median-Group/">github</a>.
</p>
</section>
</div><!-- /.entry-content -->
</section>
</div>
</main>
</div>
<div class="footer">
<div>
© 2018 Median Group
</div>
<div class="ornament-div">
<img class="ornament" src="/theme/images/ornament1black.svg" />
</div>
<div>
<a href="mailto:info@mediangroup.org">info@mediangroup.org</a>
</div>
</div>
</div>
</body>
</html>