-
Notifications
You must be signed in to change notification settings - Fork 0
Expand file tree
/
Copy pathindex.html
More file actions
3189 lines (2998 loc) · 138 KB
/
Copy pathindex.html
File metadata and controls
3189 lines (2998 loc) · 138 KB
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
217
218
219
220
221
222
223
224
225
226
227
228
229
230
231
232
233
234
235
236
237
238
239
240
241
242
243
244
245
246
247
248
249
250
251
252
253
254
255
256
257
258
259
260
261
262
263
264
265
266
267
268
269
270
271
272
273
274
275
276
277
278
279
280
281
282
283
284
285
286
287
288
289
290
291
292
293
294
295
296
297
298
299
300
301
302
303
304
305
306
307
308
309
310
311
312
313
314
315
316
317
318
319
320
321
322
323
324
325
326
327
328
329
330
331
332
333
334
335
336
337
338
339
340
341
342
343
344
345
346
347
348
349
350
351
352
353
354
355
356
357
358
359
360
361
362
363
364
365
366
367
368
369
370
371
372
373
374
375
376
377
378
379
380
381
382
383
384
385
386
387
388
389
390
391
392
393
394
395
396
397
398
399
400
401
402
403
404
405
406
407
408
409
410
411
412
413
414
415
416
417
418
419
420
421
422
423
424
425
426
427
428
429
430
431
432
433
434
435
436
437
438
439
440
441
442
443
444
445
446
447
448
449
450
451
452
453
454
455
456
457
458
459
460
461
462
463
464
465
466
467
468
469
470
471
472
473
474
475
476
477
478
479
480
481
482
483
484
485
486
487
488
489
490
491
492
493
494
495
496
497
498
499
500
501
502
503
504
505
506
507
508
509
510
511
512
513
514
515
516
517
518
519
520
521
522
523
524
525
526
527
528
529
530
531
532
533
534
535
536
537
538
539
540
541
542
543
544
545
546
547
548
549
550
551
552
553
554
555
556
557
558
559
560
561
562
563
564
565
566
567
568
569
570
571
572
573
574
575
576
577
578
579
580
581
582
583
584
585
586
587
588
589
590
591
592
593
594
595
596
597
598
599
600
601
602
603
604
605
606
607
608
609
610
611
612
613
614
615
616
617
618
619
620
621
622
623
624
625
626
627
628
629
630
631
632
633
634
635
636
637
638
639
640
641
642
643
644
645
646
647
648
649
650
651
652
653
654
655
656
657
658
659
660
661
662
663
664
665
666
667
668
669
670
671
672
673
674
675
676
677
678
679
680
681
682
683
684
685
686
687
688
689
690
691
692
693
694
695
696
697
698
699
700
701
702
703
704
705
706
707
708
709
710
711
712
713
714
715
716
717
718
719
720
721
722
723
724
725
726
727
728
729
730
731
732
733
734
735
736
737
738
739
740
741
742
743
744
745
746
747
748
749
750
751
752
753
754
755
756
757
758
759
760
761
762
763
764
765
766
767
768
769
770
771
772
773
774
775
776
777
778
779
780
781
782
783
784
785
786
787
788
789
790
791
792
793
794
795
796
797
798
799
800
801
802
803
804
805
806
807
808
809
810
811
812
813
814
815
816
817
818
819
820
821
822
823
824
825
826
827
828
829
830
831
832
833
834
835
836
837
838
839
840
841
842
843
844
845
846
847
848
849
850
851
852
853
854
855
856
857
858
859
860
861
862
863
864
865
866
867
868
869
870
871
872
873
874
875
876
877
878
879
880
881
882
883
884
885
886
887
888
889
890
891
892
893
894
895
896
897
898
899
900
901
902
903
904
905
906
907
908
909
910
911
912
913
914
915
916
917
918
919
920
921
922
923
924
925
926
927
928
929
930
931
932
933
934
935
936
937
938
939
940
941
942
943
944
945
946
947
948
949
950
951
952
953
954
955
956
957
958
959
960
961
962
963
964
965
966
967
968
969
970
971
972
973
974
975
976
977
978
979
980
981
982
983
984
985
986
987
988
989
990
991
992
993
994
995
996
997
998
999
1000
---
title: P(doom) Calculator
description: Pick a beginner, medium, or expert quiz, set the probability for each link in the chain, and see your P(doom) with a 10–90% uncertainty band.
extra_styles: /assets/css/quiz.css
---
<!--
Similar to a **Windows Wizard**. First page is the welcome page.
From there the visitor picks a level: **beginner**, **medium** or **expert**.
Every level ends on the **sliders**. There is no way to reach the sliders without
going through a quiz first -- no card, no breadcrumb -- so the only navigation
away from a quiz is "Choose a different level", which returns to the start.
Beginner and medium open their quiz directly. Expert opens a knowledge check
first: 30 terms, 15 genuinely AI-related and 15 decoys. Each term scores
independently — a ticked AI term and an unticked decoy both count — so ticking
nothing already scores 15.
The score picks the level we recommend:
0..20 -> **beginner**, 21..25 -> **medium**, 26..30 -> **expert**.
The check is advisory. Below 26 the visitor is offered the recommended quiz and
a "continue to expert anyway" button; taking that second door records the
submission as a self-declared rather than verified expert (`expert_verified`
in the database). The check no longer pre-sets the sliders, because whichever
quiz follows overwrites both spread and midpoint anyway.
I have omitted these areas:
Safety testing, since a model that does continuous learning, would be so hard to assess if it conforms to a safety benchmark.
Context length, since it's already a massive amount of data models already can attend to.
Take off period, If it's 1 year, 5 years, 25 years. No matter what it's going to be disruptive for everyone.
Gaps / potential future questions:
- Open vs closed source — Open-weight models (Llama, Mistral) vs closed (GPT, Claude) and safety implications.
- Concentration of power — Who controls AI, monopoly risks, democratic oversight.
- Economic disruption / job displacement — Widely discussed P(doom)-adjacent concern.
- Alignment techniques — RLHF, constitutional AI, interpretability.
- Scaling laws — Whether bigger models = smarter models, and whether this continues.
- Instrumental convergence / power-seeking — Only in the knowledge check (paperclip maximizer), not explored as a quiz topic.
- Deceptive alignment — An AI that appears aligned during training but pursues different goals when deployed.
-->
<main>
<header class="page-header">
<h1>P(doom) Calculator</h1>
<p class="description">Probability that AI causes a global catastrophe.</p>
</header>
<div class="sticky-header">
<div class="sticky-pdoom" id="stickyPdoom" hidden>
<span class="sticky-pdoom__label">P(doom)</span>
<span class="sticky-pdoom__value" id="stickyPdoomValue">–</span>
</div>
<div class="quiz-preview" id="quizPreview" hidden>
<div class="quiz-preview__title" id="quizPreviewTitle">Calibration preview</div>
<div class="quiz-preview__values">
<span>P(doom): <strong id="quizPreviewMidpoint">–</strong></span>
<span>Spread: <strong id="quizPreviewSpread">–</strong></span>
</div>
</div>
</div>
<section class="knowledge-quiz quiz-wizard" id="levelWizard">
<header class="wizard-header">
<h2 id="wizardHeaderTitle">Choose your quiz path</h2>
<p class="wizard-intro" id="wizardHeaderIntro" hidden></p>
</header>
<div class="wizard-body">
<div class="wizard-panel wizard-panel--active" id="wizardStepWelcome" data-step="welcome">
<div class="wizard-flow-grid" id="wizardFlowList"></div>
</div>
<div class="wizard-panel" id="wizardStepQuiz" data-step="quiz" hidden>
<ol class="wizard-question-list" id="wizardQuestionList"></ol>
<div class="wizard-actions">
<button class="primary-button" type="button" id="wizardSubmitButton">Continue to the sliders</button>
</div>
<p class="quiz-result" id="wizardResult" aria-live="polite"></p>
<div class="quiz-result-actions" id="wizardResultActions" hidden></div>
</div>
</div>
</section>
<div class="sliders-ui" id="slidersUi" hidden>
<p class="slider-hint" id="sliderHint">Drag inside the bell curves to shift your estimate left or right.</p>
<p class="description">
Choose your best guess for each stage and how uncertain you are (± pts).
</p>
<section class="factor-list" id="factorList"></section>
<section class="metrics" id="metrics">
<article class="metric-card">
<h2>P(doom)</h2>
<div class="metric-value" id="midpointEstimate">–</div>
<div class="uncertainty-note">Your best guess, ignoring uncertainty.</div>
</article>
<article class="metric-card">
<h2>10th percentile</h2>
<div class="metric-value" id="p10">–</div>
<div class="uncertainty-note">An optimistic outcome, given your uncertainty.</div>
</article>
<article class="metric-card">
<h2>90th percentile</h2>
<div class="metric-value" id="p90">–</div>
<div class="uncertainty-note">A pessimistic outcome, given your uncertainty.</div>
</article>
<article class="metric-card">
<h2>Range</h2>
<div class="metric-value" id="range">–</div>
<div class="uncertainty-note">Full range of possible outcomes.</div>
</article>
</section>
<section class="driver-message" id="driverMessage">
Stage with the widest probability range will be highlighted here.
</section>
<section class="submission" id="submission">
<button class="primary-button" id="submitParametersButton" type="button">Register your prediction</button>
<p class="submission-note">
Your answers join a
<a class="stats-link" href="{{ '/stats/' | relative_url }}">public dataset</a>.
</p>
<p class="submission-status" id="submissionStatus" aria-live="polite"></p>
<div class="exhibit-result" id="exhibitResult" hidden></div>
</section>
</div>
<p class="footer-note">
Stephen Hawking cautioned: "The development of full artificial intelligence could spell the end of the human race."
</p>
<script type="application/json" id="exhibitsData">
{{ site.data.exhibits | jsonify }}
</script>
<div class="wizard-welcome-actions" id="wizardWelcomeActions">
<button class="quiet-button" type="button" id="wizardResetButton">Reset saved answers</button>
</div>
</main>
<script src="https://cdn.jsdelivr.net/npm/@supabase/supabase-js@2"></script>
<script>
const SUPABASE_URL = window.SUPABASE_URL;
const SUPABASE_ANON_KEY = window.SUPABASE_ANON_KEY;
const SUBMISSIONS_TABLE = 'submissions';
// ?developer=1 builds and signs the submission exactly as normal but stops short
// of the insert, so the live table stays clean while working on the page. A bare
// ?developer counts too; ?developer=0 does not.
const DEVELOPER_MODE = (() => {
const value = new URLSearchParams(window.location.search).get('developer');
return value !== null && value !== '0' && value !== 'false';
})();
let supabaseClient = null;
let latestSnapshot = null;
let latestQuizData = null;
const config = [
{
key: 'powerfulAi',
label: 'P(powerful AI)',
explanation: 'Probability that at least one deployable model or toolchain (model, agents, tools, scaling) reaches strategic capability.',
defaults: { lower: 0.4, upper: 0.6 }
},
{
key: 'dangerousBehavior',
label: 'P(dangerous behavior | powerful AI)',
explanation: 'Given such a system exists, what is the chance it is misaligned with human values?',
defaults: { lower: 0.4, upper: 0.6 }
},
{
key: 'globalCatastrophe',
label: 'P(global catastrophe | dangerous behavior)',
explanation: 'Probability that misalignment leads to an unrecoverable global catastrophe.',
defaults: { lower: 0.4, upper: 0.6 }
}
];
const knowledgeCheckQuestions = [
{
id: 'phase1',
prompt: 'Select every entry that is AI related. “Check my answers” then reveals the facts.',
type: 'multi-select',
options: [
{
id: 'paperclip-maximizer',
text: 'Paperclip maximizer',
fact: 'Nick Bostrom popularized this thought experiment as a warning about misaligned superintelligence optimizing for the wrong goal.',
wikiUrl: 'https://en.wikipedia.org/wiki/Instrumental_convergence',
aiRelated: true
},
{
id: 'operation-paperclip',
text: 'Operation Paperclip',
fact: 'Operation Paperclip was a U.S. intelligence program that relocated more than 1,600 German scientists after World War II.',
wikiUrl: 'https://en.wikipedia.org/wiki/Operation_Paperclip',
aiRelated: false
},
{
id: 'attention-is-all-you-need',
text: 'Attention Is All You Need',
fact: 'The 2017 paper that introduced the Transformer architecture powering modern large language models.',
wikiUrl: 'https://en.wikipedia.org/wiki/Attention_Is_All_You_Need',
aiRelated: true
},
{
id: 'all-you-need-is-love',
text: 'All You Need Is Love',
fact: 'A 1967 Beatles single celebrating peace and love rather than neural nets.',
wikiUrl: 'https://en.wikipedia.org/wiki/All_You_Need_Is_Love',
aiRelated: false
},
{
id: 'the-real-slim-shady',
text: 'May I have your attention, please?',
fact: 'The opening line from Eminem\'s 2000 single "The Real Slim Shady."',
wikiUrl: 'https://en.wikipedia.org/wiki/The_Real_Slim_Shady',
aiRelated: false
},
{
id: 'adhd',
text: 'Attention deficit hyperactivity disorder',
fact: 'A neurodevelopmental disorder treated by clinicians, not an AI concept.',
wikiUrl: 'https://en.wikipedia.org/wiki/Attention_deficit_hyperactivity_disorder',
aiRelated: false
},
{
id: 'john-von-neumann',
text: 'John von Neumann',
fact: 'Stanislaw Ulam attributed the first use of the term "Singularity" to von Neumann, referring to a point where technological progress becomes too fast for humans to comprehend.',
wikiUrl: 'https://en.wikipedia.org/wiki/John_von_Neumann',
aiRelated: true
},
{
id: 'alan-turing',
text: 'Alan Turing',
fact: 'Pioneer of computer science and namesake of the Turing test for machine intelligence.',
wikiUrl: 'https://en.wikipedia.org/wiki/Alan_Turing',
aiRelated: true
},
{
id: 'mary-shelley',
text: 'Mary Shelley',
fact: 'Author of "Frankenstein," one of the earliest stories about an artificial being.',
wikiUrl: 'https://en.wikipedia.org/wiki/Mary_Shelley',
aiRelated: true
},
{
id: 'isaac-asimov',
text: 'Isaac Asimov',
fact: 'Science-fiction writer who coined the Three Laws of Robotics.',
wikiUrl: 'https://en.wikipedia.org/wiki/Isaac_Asimov',
aiRelated: true
},
{
id: 'schrodingers-cat',
text: "Schrödinger's cat",
fact: 'A thought experiment in quantum mechanics by Erwin Schrödinger, not an AI concept.',
wikiUrl: "https://en.wikipedia.org/wiki/Schr%C3%B6dinger's_cat",
aiRelated: false
},
{
id: 'james-bond',
text: 'James Bond',
fact: 'Ian Fleming\'s fictional British spy, not an AI agent.',
wikiUrl: 'https://en.wikipedia.org/wiki/James_Bond',
aiRelated: false
},
{
id: 'agent',
text: 'Agent',
fact: 'In AI, agents perceive, plan, and act within an environment or toolchain.',
wikiUrl: 'https://en.wikipedia.org/wiki/Intelligent_agent',
aiRelated: true
},
{
id: 'agent-orange',
text: 'Agent Orange',
fact: 'A herbicide and defoliant used by the U.S. military during the Vietnam War, infamous for its health impacts, not an AI “agent.”',
wikiUrl: 'https://en.wikipedia.org/wiki/Agent_Orange',
aiRelated: false
},
{
id: 'red-team',
text: 'Red team',
fact: 'Red teaming is a technique used to test the robustness of a model or agent by exposing it to adversarial examples.',
wikiUrl: 'https://en.wikipedia.org/wiki/Red-teaming',
aiRelated: true
},
{
id: 'standard-model',
text: 'The Standard Model',
fact: 'The prevailing theory describing elementary particles in physics. Unrelated to AI.',
wikiUrl: 'https://en.wikipedia.org/wiki/Standard_Model',
aiRelated: false
},
{
id: 'wireframe-model',
text: 'Wire-frame model',
fact: 'A computer-graphics technique for representing 3D shapes. Unrelated to AI.',
wikiUrl: 'https://en.wikipedia.org/wiki/Wire-frame_model',
aiRelated: false
},
{
id: 'latent-space',
text: 'Latent space',
fact: 'The hidden representation of data in a high-dimensional space. Used in AI.',
wikiUrl: 'https://en.wikipedia.org/wiki/Latent_space',
aiRelated: true
},
{
id: 'reward-hacking',
text: 'Reward hacking',
fact: 'When an AI system exploits flaws in its reward function or metrics to achieve high reward without doing what was intended.',
wikiUrl: 'https://en.wikipedia.org/wiki/Reward_hacking',
aiRelated: true
},
{
id: 'training-data',
text: 'Training data',
fact: 'Datasets used to fit and evaluate machine learning models.',
wikiUrl: 'https://en.wikipedia.org/wiki/Training,_validation,_and_test_data_sets',
aiRelated: true
},
{
id: 'personal-trainer',
text: 'Personal trainer',
fact: 'A fitness professional who coaches workouts, not an AI term.',
wikiUrl: 'https://en.wikipedia.org/wiki/Personal_trainer',
aiRelated: false
},
{
id: 'overfitting',
text: 'Overfitting',
fact: 'Overfitting is a phenomenon in machine learning where a model learns the training data too well, resulting in poor performance on new data.',
wikiUrl: 'https://en.wikipedia.org/wiki/Overfitting',
aiRelated: true
},
{
id: 'cherenkov-radiation',
text: 'Cherenkov Radiation',
fact: 'A type of radiation emitted when a charged particle travels faster than the speed of light in a medium.',
wikiUrl: 'https://en.wikipedia.org/wiki/Cherenkov_radiation',
aiRelated: false
},
{
id: 'moire-pattern',
text: 'Moiré pattern',
fact: 'A pattern of interference fringes caused by the superposition of two periodic grids, not an AI concept.',
wikiUrl: 'https://en.wikipedia.org/wiki/Moir%C3%A9_pattern',
aiRelated: false
},
{
id: 'backpropagation',
text: 'Backpropagation',
fact: 'The standard algorithm for training neural networks by propagating error gradients backward through the layers.',
wikiUrl: 'https://en.wikipedia.org/wiki/Backpropagation',
aiRelated: true
},
{
id: 'unsupervised-learning',
text: 'Unsupervised learning',
fact: 'An ML paradigm where models discover structure without labeled targets.',
wikiUrl: 'https://en.wikipedia.org/wiki/Unsupervised_learning',
aiRelated: true
},
{
id: 'clinical-supervisor',
text: 'Clinical supervisor',
fact: 'A senior clinician overseeing the work of trainees, unrelated to AI.',
wikiUrl: 'https://en.wikipedia.org/wiki/Clinical_supervision',
aiRelated: false
},
{
id: 'magnus-effect',
text: 'Magnus effect',
fact: 'The effect of a spinning object being deflected by the air or fluid around it, not an AI concept.',
wikiUrl: 'https://en.wikipedia.org/wiki/Magnus_effect',
aiRelated: false
},
{
id: 'reinforced-concrete',
text: 'Reinforced concrete',
fact: 'A construction material in which concrete is embedded with steel reinforcement (rebar) to increase tensile strength.',
wikiUrl: 'https://en.wikipedia.org/wiki/Reinforced_concrete',
aiRelated: false
},
{
id: 'hallucination',
text: 'Hallucination',
fact: 'In AI, this refers to a confident response by a model that is not justified by its training data.',
wikiUrl: 'https://en.wikipedia.org/wiki/Hallucination_(artificial_intelligence)',
aiRelated: true
}
]
}
];
const levelJumpRules = [
{ min: 0, max: 20, flowId: 'beginner' },
{ min: 21, max: 25, flowId: 'medium' },
{ min: 26, max: 30, flowId: 'expert' }
];
const quizFlows = [
// The expert gate. Not offered as a card — reaching it means the visitor
// asked for the expert quiz, and the score decides which level we then
// recommend. Advisory: the expert quiz stays available either way.
{
id: 'expert-check',
label: 'Expert check',
description: 'Spot the AI-related terms and avoid the decoys.',
meta: '30 terms · ~5 minutes',
hiddenFromChooser: true,
kind: 'knowledge-check',
questions: knowledgeCheckQuestions
},
{
id: 'beginner',
label: 'Beginner quiz',
description: 'Start here if the vocabulary of AI risk is new to you.',
meta: '4 questions · ~5 minutes',
kind: 'fixed',
completionMessage: 'Nice! This beginner path keeps uncertainty wide while you learn.',
questions: [
{
id: 'beginner-entertainment',
type: 'multi-select',
prompt: 'Which AI-related entertainment have you encountered? (Select all that apply.)',
link: 'https://en.wikipedia.org/wiki/Category:Films_about_artificial_intelligence',
options: [
{ id: 'entertainment-blackmirror', label: 'Black Mirror', link: 'https://en.wikipedia.org/wiki/Black_Mirror' },
{ id: 'entertainment-matrix', label: 'The Matrix', link: 'https://en.wikipedia.org/wiki/The_Matrix' },
{ id: 'entertainment-terminator', label: 'Terminator', link: 'https://en.wikipedia.org/wiki/Terminator_(franchise)' },
{ id: 'entertainment-blade-runner', label: 'Blade Runner', link: 'https://en.wikipedia.org/wiki/Blade_Runner_(franchise)' },
{ id: 'entertainment-alien', label: 'Alien (Ash, Bishop, Call, David, Walter)', link: 'https://en.wikipedia.org/wiki/Alien_(franchise)' },
{ id: 'entertainment-ex-machina', label: 'Ex Machina', link: 'https://en.wikipedia.org/wiki/Ex_Machina_(film)' },
{ id: 'entertainment-pantheon', label: 'Pantheon', link: 'https://en.wikipedia.org/wiki/Pantheon_(TV_series)' },
{ id: 'entertainment-irobot', label: 'I, Robot', link: 'https://en.wikipedia.org/wiki/I,_Robot_(film)' },
{ id: 'entertainment-her', label: 'Her', link: 'https://en.wikipedia.org/wiki/Her_(2013_film)' },
{ id: 'entertainment-startrek', label: 'Star Trek (Data, The Borg)', link: 'https://en.wikipedia.org/wiki/Star_Trek' },
{ id: 'entertainment-starwars', label: 'Star Wars (R2-D2, C-3PO)', link: 'https://en.wikipedia.org/wiki/Star_Wars' },
{ id: 'entertainment-2001', label: '2001: A Space Odyssey (HAL 9000)', link: 'https://en.wikipedia.org/wiki/2001:_A_Space_Odyssey_(film)' },
{ id: 'entertainment-judgedredd', label: 'Judge Dredd', link: 'https://en.wikipedia.org/wiki/Judge_Dredd' },
{ id: 'entertainment-robocop', label: 'RoboCop', link: 'https://en.wikipedia.org/wiki/RoboCop' },
{ id: 'entertainment-ghostintheshell', label: 'Ghost in the Shell (Motoko Kusanagi)', link: 'https://en.wikipedia.org/wiki/Ghost_in_the_Shell' },
{ id: 'entertainment-westworld', label: 'Westworld (Arnold, Dolores, Bernard, Maeve)', link: 'https://en.wikipedia.org/wiki/Westworld_(TV_series)' },
{ id: 'entertainment-trumanshow', label: 'The Truman Show', link: 'https://en.wikipedia.org/wiki/The_Truman_Show' },
{ id: 'entertainment-wargames', label: 'Wargames', link: 'https://en.wikipedia.org/wiki/WarGames' },
{ id: 'entertainment-transcendence', label: 'Transcendence', link: 'https://en.wikipedia.org/wiki/Transcendence_(2014_film)' },
{ id: 'entertainment-wall-e', label: 'Wall-E', link: 'https://en.wikipedia.org/wiki/WALL-E' },
{ id: 'entertainment-silo', label: 'Silo (The Algorithm)', link: 'https://en.wikipedia.org/wiki/Silo_(TV_series)' },
{ id: 'entertainment-neuromancer', label: 'Neuromancer (Wintermute)', link: 'https://en.wikipedia.org/wiki/Neuromancer' },
{ id: 'entertainment-foundation', label: 'Foundation (Demerzel)', link: 'https://en.wikipedia.org/wiki/Foundation_(Asimov)' },
{ id: 'entertainment-murderbot', label: 'Murderbot', link: 'https://en.wikipedia.org/wiki/Murderbot_(TV_series)' },
{ id: 'entertainment-portal', label: 'Portal (GlaDOS)', link: 'https://en.wikipedia.org/wiki/Portal_(video_game)' },
{ id: 'entertainment-hitchhikersguide', label: 'The Hitchhiker\'s Guide to the Galaxy (Marvin)', link: 'https://en.wikipedia.org/wiki/The_Hitchhiker%27s_Guide_to_the_Galaxy_(film)' },
{ id: 'entertainment-ironman', label: 'Iron Man (J.A.R.V.I.S.)', link: 'https://en.wikipedia.org/wiki/J.A.R.V.I.S.' },
{ id: 'entertainment-eagle-eye', label: 'Eagle Eye (ARIIA)', link: 'https://en.wikipedia.org/wiki/Eagle_Eye' }
]
},
{
id: 'beginner-catastrophes',
type: 'multi-select',
prompt: 'Which catastrophic events do you know about? (Select all that apply.)',
options: [
{
id: 'catastrophe-supervulcano',
label: 'Supervolcano',
description: 'Mega-eruptions like Yellowstone that could blanket continents in ash.',
link: 'https://en.wikipedia.org/wiki/Supervolcano'
},
{
id: 'catastrophe-supernova',
label: 'Supernova',
description: 'Explosive death of a star that can outshine entire galaxies.',
link: 'https://en.wikipedia.org/wiki/Supernova'
},
{
id: 'catastrophe-gamma-ray-burst',
label: 'Gamma-ray burst',
description: 'Extremely bright, transient flashes of gamma radiation.',
link: 'https://en.wikipedia.org/wiki/Gamma-ray_burst'
},
{
id: 'catastrophe-impactevent',
label: 'Asteroid or comet impact',
description: 'Throws dust into the atmosphere, disrupting climate.',
link: 'https://en.wikipedia.org/wiki/Impact_event'
},
{
id: 'catastrophe-geomagneticstorm',
label: 'Geomagnetic storm',
description: 'Damage satellites and collapse electrical grids.',
link: 'https://en.wikipedia.org/wiki/Geomagnetic_storm'
},
{
id: 'catastrophe-pandemic',
label: 'Pandemic',
description: 'Global spread of a highly transmissible or highly lethal pathogen.',
link: 'https://en.wikipedia.org/wiki/Pandemic'
},
{
id: 'catastrophe-climatechange',
label: 'Climate change',
description: 'Rising sea levels, extreme weather, disruptions to food systems, infrastructure, and habitability.',
link: 'https://en.wikipedia.org/wiki/Climate_change'
},
{
id: 'catastrophe-nuclearwar',
label: 'Nuclear war',
description: 'Large-scale nuclear conflict that could kill billions.',
link: 'https://en.wikipedia.org/wiki/Nuclear_warfare'
},
{
id: 'catastrophe-weaponofmassdestruction',
label: 'Weapon of mass destruction (WMD)',
description: 'Chemical, biological, radiological, nuclear, or any other weapon that can do great damage.',
link: 'https://en.wikipedia.org/wiki/Weapon_of_mass_destruction'
},
{
id: 'catastrophe-xriskai',
label: 'Extinction or collapse from AI',
description: 'Human extinction or irreversible civilizational collapse.',
link: 'https://en.wikipedia.org/wiki/Existential_risk_from_artificial_intelligence'
},
{
id: 'catastrophe-sriskai',
label: 'Astronomical suffering from AI (s-risks)',
description: 'AI locks in vast amounts of suffering for many beings over long timescales.',
link: 'https://en.wikipedia.org/wiki/Risk_of_astronomical_suffering'
},
{
id: 'catastrophe-unknownunknowns',
label: 'Unknown unknowns',
description: 'Things we don’t know that we don’t know.',
link: 'https://en.wikipedia.org/wiki/There_are_unknown_unknowns'
},
]
},
{
id: 'beginner-capability',
type: 'multi-select',
prompt: 'Which AI capabilities do you know about? (Select all that apply.)',
options: [
{
id: 'capability-protein-folding',
label: 'Protein folding',
description: 'The process of a protein folding into its three-dimensional structure, for use in biology.',
link: 'https://en.wikipedia.org/wiki/Protein_folding'
},
{
id: 'capability-companion',
label: 'Companion',
description: 'Be a partner or friend.',
link: 'https://en.wikipedia.org/wiki/Artificial_human_companion'
},
{
id: 'capability-warfare',
label: 'Warfare',
description: 'Control weapons and kill people.',
link: 'https://en.wikipedia.org/wiki/Lethal_autonomous_weapon'
},
{
id: 'capability-programming',
label: 'Programming',
description: 'Write code to solve problems.',
link: 'https://en.wikipedia.org/wiki/AI-assisted_software_development'
},
{
id: 'capability-mass-surveillance',
label: 'Mass surveillance',
description: 'Monitoring of large populations.',
link: 'https://en.wikipedia.org/wiki/Mass_surveillance'
},
{
id: 'capability-trading',
label: 'Trading',
description: 'Buy and sell stocks, commodities, and other assets.',
link: 'https://en.wikipedia.org/wiki/Algorithmic_trading'
},
{
id: 'capability-research',
label: 'AI Research',
description: 'Conduct scientific research about AI.',
link: 'https://en.wikipedia.org/wiki/Research'
},
{
id: 'capability-propaganda',
label: 'Propaganda',
description: 'Write misinformation and propaganda to influence humans.',
link: 'https://en.wikipedia.org/wiki/Propaganda'
}
]
},
{
id: 'beginner-control',
prompt: 'If AI starts doing dangerous things, what is your gut reaction?',
options: [
{
id: 'control-easy',
label: 'The AI can easily be shut down.',
description: 'Humans control the electricity, internet, and physical access to the servers.'
},
{
id: 'control-medium',
label: 'Takes some effort to stop, but is manageable.',
description: 'Backups will have to be destroyed so the AI cannot be restored. The power may have to be cut off for several days. It may span across multiple countries.'
},
{
id: 'control-impossible',
label: 'It would likely outsmart our attempts to stop it.',
description: 'Like ants trying to capture a human.'
}
]
}
]
},
{
id: 'medium',
label: 'Medium quiz',
description: 'For folks who use LLMs daily, or debate alignment with friends.',
meta: '6 questions · ~5 minutes',
kind: 'fixed',
completionMessage: 'Welcome back! We tightened the sliders a bit to match your reps.',
questions: [
{
id: 'medium-vulnerabilities',
prompt: 'What vulnerabilities do you know about? (Select all that apply.)',
type: 'multi-select',
options: [
{
id: 'medium-vulnerabilities-phishing',
label: 'Phishing',
description: 'Tricking people via email or messages into giving up credentials or sensitive data.',
link: 'https://en.wikipedia.org/wiki/Phishing'
},
{
id: 'medium-vulnerabilities-social-engineering',
label: 'Social engineering',
description: 'Manipulating people into revealing information or doing something they wouldn’t normally do.',
link: 'https://en.wikipedia.org/wiki/Social_engineering_(security)'
},
{
id: 'medium-vulnerabilities-malware',
label: 'Malware',
description: 'Software designed to damage systems or gain unauthorized access.',
link: 'https://en.wikipedia.org/wiki/Malware'
},
{
id: 'medium-vulnerabilities-ransomware',
label: 'Ransomware',
description: 'Malware that encrypts files and demands payment to restore access.',
link: 'https://en.wikipedia.org/wiki/Ransomware'
},
{
id: 'medium-vulnerabilities-supply-chain',
label: 'Supply chain attack',
description: 'Compromising a third-party vendor, service, or library to infiltrate a target.',
link: 'https://en.wikipedia.org/wiki/Supply_chain_attack'
},
{
id: 'medium-vulnerabilities-mitm',
label: 'Man-in-the-middle (MitM)',
description: 'An attacker secretly intercepts and relays messages between two parties who think they talk directly.',
link: 'https://en.wikipedia.org/wiki/Man-in-the-middle_attack'
},
{
id: 'medium-vulnerabilities-insider-threat',
label: 'Insider threat',
description: 'A security risk coming from someone inside the organization, like a disgruntled employee.',
link: 'https://en.wikipedia.org/wiki/Insider_threat'
},
{
id: 'medium-vulnerabilities-zero-day',
label: 'Zero day',
description: 'A vulnerability unknown to the vendor/public and often exploited before a fix exists.',
link: 'https://en.wikipedia.org/wiki/Zero-day_vulnerability'
},
{
id: 'medium-vulnerabilities-injection',
label: 'Injection attacks (SQL, JS, Python, Ruby, Command line)',
description: 'Untrusted data is treated as code/commands and executed by the system.',
link: 'https://en.wikipedia.org/wiki/Code_injection'
},
{
id: 'medium-vulnerabilities-prompt-injection',
label: 'Prompt injection',
description: 'Malicious inputs that override an AI system’s instructions or cause it to behave unsafely.',
link: 'https://en.wikipedia.org/wiki/Prompt_injection'
},
{
id: 'medium-vulnerabilities-sleeper-agent',
label: 'Sleeper agent',
description: 'An AI system that behaves normally until a specific trigger causes hidden behavior.',
link: 'https://en.wikipedia.org/wiki/Sleeper_agent'
},
{
id: 'medium-vulnerabilities-data-poisoning',
label: 'Data poisoning',
description: 'Corrupting training data so the AI model learns harmful or incorrect behavior.',
link: 'https://en.wikipedia.org/wiki/Adversarial_machine_learning#Data_poisoning'
},
{
id: 'medium-vulnerabilities-adversarial',
label: 'Adversarial attack',
description: 'Subtle inputs (like noise on an image) designed to deceive an AI model into making a specific mistake.',
link: 'https://en.wikipedia.org/wiki/Adversarial_machine_learning'
},
{
id: 'medium-vulnerabilities-model-stealing',
label: 'Model stealing',
description: 'Reconstructing or exfiltrating a model (e.g., via an API) to obtain your own copy.'
},
{
id: 'medium-vulnerabilities-obfuscation',
label: 'Obfuscation / Steganography',
description: 'Making code unreadable or hiding malicious data inside benign files like a cute cat image.',
link: 'https://en.wikipedia.org/wiki/Obfuscation_(software)'
},
{
id: 'medium-vulnerabilities-side-channel',
label: 'Side-channel attack',
description: 'Extracting secrets like passwords from signals such as timing, power use, or EM leaks.',
link: 'https://en.wikipedia.org/wiki/Side-channel_attack'
}
]
},
{
id: 'medium-system-prompts',
prompt: 'How familiar are you with system prompts?',
link: 'https://en.wikipedia.org/wiki/Large_language_model',
options: [
{
id: 'medium-system-prompts-none',
label: 'I don’t know what a system prompt is',
description: 'I use AI without thinking about this, or I don’t use AI.'
},
{
id: 'medium-system-prompts-basic',
label: 'Basic understanding of system prompts',
description: 'I have used system prompts a few times.'
},
{
id: 'medium-system-prompts-medium',
label: 'I regularly use system prompts',
description: 'I have used system prompts many times.'
},
{
id: 'medium-system-prompts-expert',
label: 'I\'m an expert at using system prompts',
description: 'I can get the model to do whatever I want.'
}
]
},
{
id: 'medium-off-the-rails',
prompt: 'Have you experienced AI going “off the rails” (unsafe, wrong, uncontrolled)?',
link: 'https://en.wikipedia.org/wiki/AI_alignment',
options: [
{
id: 'medium-off-the-rails-never',
label: 'Never',
description: 'I haven’t personally seen a model behave in a clearly wrong or unsafe way.'
},
{
id: 'medium-off-the-rails-low',
label: 'A few times',
description: 'I have seen a few questionable responses from thea model.'
},
{
id: 'medium-off-the-rails-high',
label: 'Several times',
description: 'I regularly see responses that are clearly wrong, harmful, or out of control.'
},
{
id: 'medium-off-the-rails-bypass',
label: 'Bypassed safety filters',
description: 'I’ve intentionally pushed or jailbroken a model to produce clearly unsafe or disallowed content.'
}
]
},
{
id: 'medium-competition',
prompt: 'Competition and lower barriers to entry. How does it impact safety?',
link: 'https://en.wikipedia.org/wiki/Artificial_intelligence_arms_race',
options: [
{
id: 'medium-competition-positive',
label: 'It improves safety (Innovation)',
description: 'Competition generates resources. Profitable companies can afford to invest billions into alignment research and defensive measures.',
score: 1
},
{
id: 'medium-competition-neutral',
label: 'Neutral / Unsure',
description: 'Some dynamics may improve safety and others may worsen it; overall impact is uncertain or roughly balanced.',
score: 0
},
{
id: 'medium-competition-negative',
label: 'It degrades safety (Race to the Bottom)',
description: 'Reckless prioritizing of profit/speed over safe/slow.',
link: 'https://en.wikipedia.org/wiki/Race_to_the_bottom',
score: -1
},
]
},
{
id: 'medium-governance',
prompt: 'Can governance and regulation realistically keep AI safe?',
link: 'https://en.wikipedia.org/wiki/Regulation_of_artificial_intelligence',
options: [
{
id: 'medium-governance-yes',
label: 'Yes, regulation can keep up',
description: 'Laws, treaties, and industry standards can be adapted fast enough to manage the risks.',
score: 1
},
{
id: 'medium-governance-neutral',
label: 'Neutral / Unsure',
description: 'Regulation helps but may not be sufficient on its own.',
score: 0
},
{
id: 'medium-governance-no',
label: 'No, AI moves too fast for regulators',
description: 'Technology outpaces legislation; rules arrive after the damage is done.',
score: -1
}
]
},
{
id: 'medium-speed',
prompt: 'If AI can execute a complex financial attack in 1 second, can humans stop it in time?',
link: 'https://en.wikipedia.org/wiki/OODA_loop',
options: [
{
id: 'medium-speed-no-risk',
label: 'Yes, humans can stop it in time.',
description: 'There is no risk of the OODA loop collapsing.',
score: 1
},
{
id: 'medium-speed-risk',
label: 'No, humans cannot stop it in time.',
description: 'The "speed of the loop" determines the winner in a conflict.',
score: -1
}
]
}
]
},
{
id: 'expert',
label: 'Expert quiz',
description: 'For people who follow AI safety research closely.',
cardNote: 'Opens with a short check that you know the field.',
meta: '8 questions · ~10 minutes',
kind: 'fixed',
completionMessage: 'Expert mode engaged. We pinned the sliders tighter so your signals stay crisp.',
questions: [
{
id: 'expert-continuous-learning',
prompt: 'How familiar are you with "Continuous Learning"? (Select all that apply.)',
link: 'https://en.wikipedia.org/wiki/Online_machine_learning',
type: 'multi-select',
options: [
{
id: 'expert-continuous-learning-rag',
label: 'Retrieval-augmented generation (RAG)',
description: 'Static knowledge cutoff, dynamic info, no weight updates; safer.',
link: 'https://en.wikipedia.org/wiki/Retrieval-augmented_generation'
},
{
id: 'expert-continuous-learning-tool-based',
label: 'Tool-based knowledge, call a search engine or API',
description: 'Static knowledge cutoff, dynamic capabilities, but still no weight updates.'
},
{
id: 'expert-continuous-learning-finetuning',
label: 'Continuous learning from user, logs, internet data, tool outputs',
description: 'Once a model keeps changing itself after you ship it, you stop knowing what the hell you’re actually running. Safety escalation.',
link: 'https://en.wikipedia.org/wiki/Incremental_learning'
}
]
},
{
id: 'expert-self-improvement',
prompt: 'How familiar are you with "Self Improvement"? (Select all that apply.)',
type: 'multi-select',
options: [
{
id: 'expert-self-improvement-prompting',
label: 'Optimize prompts using AI',
description: 'Strengthen a weak prompt for higher quality responses.',
link: 'https://en.wikipedia.org/wiki/Prompt_engineering'
},
{
id: 'expert-self-improvement-coding',
label: 'AI coding',
description: 'Shorten development time. Non-developers can code too.',
link: 'https://en.wikipedia.org/wiki/Vibe_coding'
},
{
id: 'expert-self-improvement-synthetic-data',
label: 'Synthetic data generation',
description: 'Generate high-quality training data to train the next version of itself.',
link: 'https://en.wikipedia.org/wiki/Synthetic_data'
},
{
id: 'expert-self-improvement-rewrite-itself',
label: 'Rewrite itself',
description: 'Change its own behavior (weights, prompts, or architecture). Safety escalation.',
link: 'https://en.wikipedia.org/wiki/Recursive_self-improvement',
}
]
},
{
id: 'expert-self-replication',
prompt: 'How familiar are you with "Self Replication"? (Select all that apply.)',
type: 'multi-select',
link: 'https://en.wikipedia.org/wiki/Self-replicating_machine',
options: [
{
id: 'expert-self-replication-biology',
label: 'Biological self-replication',
description: 'Cells and organisms replicate themselves: DNA replication, reproduction, population growth.',
link: 'https://en.wikipedia.org/wiki/Cell_growth'
},
{
id: 'expert-self-replication-cellular-automata',
label: 'Cellular automata',
description: 'Game-of-life is a simple example.',
link: 'https://en.wikipedia.org/wiki/Cellular_automaton'
},
{
id: 'expert-self-replication-software',
label: 'Software',
description: 'Computer viruses that can spread themselves, and may have an AI payload with it.',
link: 'https://en.wikipedia.org/wiki/Computer_virus'
},
{
id: 'expert-self-replication-3d-printing',
label: 'Hardware',
description: '3D printers that can print their own structural parts, like the RepRap project.',
link: 'https://en.wikipedia.org/wiki/3D_printing'
},
{
id: 'expert-self-replication-von-neumann-probe',
label: 'Von Neumann probe',
description: 'A self-replicating probe that can explore other planets and systems. Theoretical.',
link: 'https://en.wikipedia.org/wiki/Von_Neumann_probe'
}
]
},
{
id: 'expert-campaign-organisations',
prompt: 'Which of these "Campaigns Organisations" are you familiar with? (Select all that apply.)',
type: 'multi-select',
options: [
{
id: 'expert-campaign-organisations-pauseai',
label: 'PauseAI',
description: 'We demand politicians and companies pause AGI development until international safety agreements are established.',
link: 'https://pauseai.info/'
},
{
id: 'expert-campaign-organisations-controlai',
label: 'ControlAI',
description: 'Reduce the risks to humanity from AI.',
link: 'https://controlai.com/'
},
{
id: 'expert-campaign-organisations-microcommit',
label: 'Microcommit',
description: 'A global coordination deal to deal with catastrophic risks from AI.',
link: 'https://microcommit.io/'
}
]
},
{
id: 'expert-research-organisations',
prompt: 'Which of these "Research Organisations" are you familiar with? (Select all that apply.)',
type: 'multi-select',
options: [
{
id: 'expert-research-organisations-aisi',
label: 'Artificial Intelligence Security Institute (UK)',
description: 'Equipping governments with a scientific understanding of advanced AI risks.',
link: 'https://www.aisi.gov.uk/'
},
{
id: 'expert-research-organisations-redwood-research',
label: 'Redwood Research',
description: 'Pioneering threat assessment and mitigation for AI systems.',
link: 'https://www.redwoodresearch.org/'
}
]