-
Notifications
You must be signed in to change notification settings - Fork 1
Expand file tree
/
Copy pathindex.html
More file actions
651 lines (587 loc) · 18.2 KB
/
Copy pathindex.html
File metadata and controls
651 lines (587 loc) · 18.2 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
<!doctype html>
<html lang="en">
<head>
<meta charset="utf-8">
<meta name="viewport" content="width=device-width, initial-scale=1">
<meta
name="description"
content="A research concept for grounded, auditable language emergence between isolated artificial agents."
>
<meta name="theme-color" content="#0b1020">
<title>Agentic Language Development | Ethical Tech CoLab</title>
<style>
:root {
color-scheme: dark;
--bg: #070b16;
--surface: #10182b;
--surface-strong: #17213a;
--border: #293756;
--text: #eef3ff;
--muted: #a8b4cc;
--blue: #71a7ff;
--cyan: #56d7e5;
--purple: #bc8cff;
--green: #6ce2ad;
--yellow: #ffd479;
--max: 1120px;
}
* {
box-sizing: border-box;
}
html {
scroll-behavior: smooth;
}
body {
margin: 0;
background:
radial-gradient(circle at 20% 0%, rgba(86, 215, 229, 0.12), transparent 28rem),
radial-gradient(circle at 85% 15%, rgba(188, 140, 255, 0.13), transparent 30rem),
var(--bg);
color: var(--text);
font-family: "Segoe UI", system-ui, -apple-system, BlinkMacSystemFont, sans-serif;
line-height: 1.65;
}
a {
color: var(--cyan);
}
a:focus-visible,
.button:focus-visible {
outline: 3px solid var(--yellow);
outline-offset: 3px;
}
.shell {
width: min(calc(100% - 2rem), var(--max));
margin: 0 auto;
}
header {
border-bottom: 1px solid rgba(168, 180, 204, 0.18);
background: rgba(7, 11, 22, 0.82);
backdrop-filter: blur(18px);
position: sticky;
top: 0;
z-index: 10;
}
nav {
min-height: 4.5rem;
display: flex;
align-items: center;
justify-content: space-between;
gap: 1.5rem;
}
.brand {
color: var(--text);
font-weight: 750;
letter-spacing: 0.01em;
text-decoration: none;
}
.brand span {
color: var(--cyan);
}
.nav-links {
display: flex;
flex-wrap: wrap;
gap: 1.25rem;
align-items: center;
}
.nav-links a {
color: var(--muted);
font-size: 0.94rem;
text-decoration: none;
}
.nav-links a:hover {
color: var(--text);
}
main {
overflow: hidden;
}
.hero {
padding: 7.5rem 0 5rem;
display: grid;
grid-template-columns: minmax(0, 1.2fr) minmax(19rem, 0.8fr);
gap: 4rem;
align-items: center;
}
.eyebrow {
margin: 0 0 1rem;
color: var(--green);
font-size: 0.82rem;
font-weight: 750;
letter-spacing: 0.16em;
text-transform: uppercase;
}
h1,
h2,
h3,
p {
margin-top: 0;
}
h1 {
max-width: 13ch;
margin-bottom: 1.5rem;
font-size: clamp(3rem, 8vw, 6.6rem);
line-height: 0.96;
letter-spacing: -0.055em;
}
h1 .gradient {
background: linear-gradient(100deg, var(--cyan), var(--blue), var(--purple));
background-clip: text;
color: transparent;
}
h2 {
max-width: 19ch;
margin-bottom: 1rem;
font-size: clamp(2rem, 4vw, 3.3rem);
line-height: 1.08;
letter-spacing: -0.035em;
}
h3 {
font-size: 1.05rem;
}
.lede {
max-width: 64ch;
color: var(--muted);
font-size: clamp(1.05rem, 2vw, 1.28rem);
}
.actions {
margin-top: 2rem;
display: flex;
flex-wrap: wrap;
gap: 0.85rem;
}
.button {
min-height: 3rem;
padding: 0.75rem 1.15rem;
border: 1px solid var(--border);
border-radius: 0.75rem;
display: inline-flex;
align-items: center;
justify-content: center;
color: var(--text);
font-weight: 700;
text-decoration: none;
transition: transform 140ms ease, border-color 140ms ease, background 140ms ease;
}
.button.primary {
border-color: transparent;
background: linear-gradient(110deg, #257fd7, #7852c8);
}
.button:hover {
transform: translateY(-2px);
border-color: var(--cyan);
}
.architecture {
min-height: 24rem;
padding: 2rem;
border: 1px solid var(--border);
border-radius: 1.4rem;
display: grid;
gap: 1rem;
background: linear-gradient(160deg, rgba(23, 33, 58, 0.94), rgba(16, 24, 43, 0.72));
box-shadow: 0 2rem 6rem rgba(0, 0, 0, 0.3);
}
.role {
padding: 1rem;
border: 1px solid rgba(113, 167, 255, 0.34);
border-radius: 0.9rem;
background: rgba(7, 11, 22, 0.66);
}
.role strong {
display: block;
margin-bottom: 0.25rem;
color: var(--blue);
}
.role small {
color: var(--muted);
}
.babies {
display: grid;
grid-template-columns: 1fr 1fr;
gap: 1rem;
}
.channel {
padding: 0.75rem;
border: 1px dashed var(--purple);
border-radius: 999px;
color: var(--purple);
font-family: "Cascadia Code", "Fira Code", monospace;
font-size: 0.82rem;
text-align: center;
}
section {
padding: 5.5rem 0;
}
.section-intro {
max-width: 66ch;
margin-bottom: 2.5rem;
color: var(--muted);
}
.grid {
display: grid;
grid-template-columns: repeat(3, minmax(0, 1fr));
gap: 1rem;
}
.card {
padding: 1.4rem;
border: 1px solid var(--border);
border-radius: 1rem;
background: rgba(16, 24, 43, 0.7);
}
.card .number {
color: var(--green);
font-family: "Cascadia Code", "Fira Code", monospace;
font-size: 0.82rem;
}
.card p {
margin-bottom: 0;
color: var(--muted);
font-size: 0.95rem;
}
.callout {
padding: clamp(1.5rem, 5vw, 3rem);
border: 1px solid rgba(255, 212, 121, 0.38);
border-radius: 1.25rem;
background: linear-gradient(120deg, rgba(255, 212, 121, 0.08), rgba(188, 140, 255, 0.08));
}
.callout h2 {
color: var(--yellow);
}
.callout p:last-child {
margin-bottom: 0;
}
.research-list {
display: grid;
grid-template-columns: repeat(2, minmax(0, 1fr));
gap: 1rem;
}
.research-item {
padding: 1.1rem 0;
border-top: 1px solid var(--border);
}
.research-item strong {
color: var(--cyan);
}
.research-item p {
margin: 0.35rem 0 0;
color: var(--muted);
}
.cta {
padding: clamp(2rem, 6vw, 4.5rem);
border: 1px solid rgba(86, 215, 229, 0.35);
border-radius: 1.5rem;
background:
radial-gradient(circle at 80% 20%, rgba(86, 215, 229, 0.14), transparent 20rem),
var(--surface);
}
.cta h2 {
max-width: 24ch;
}
.cta p {
max-width: 65ch;
color: var(--muted);
}
footer {
padding: 2.5rem 0;
border-top: 1px solid rgba(168, 180, 204, 0.16);
color: var(--muted);
font-size: 0.92rem;
}
.footer-row {
display: flex;
flex-wrap: wrap;
justify-content: space-between;
gap: 1rem;
}
@media (max-width: 850px) {
.hero {
padding-top: 5rem;
grid-template-columns: 1fr;
}
.grid {
grid-template-columns: 1fr 1fr;
}
}
@media (max-width: 620px) {
nav,
.nav-links {
align-items: flex-start;
}
nav {
padding: 1rem 0;
flex-direction: column;
}
.hero {
padding-top: 3.5rem;
}
.grid,
.research-list {
grid-template-columns: 1fr;
}
}
</style>
</head>
<body>
<header>
<nav class="shell" aria-label="Primary navigation">
<a class="brand" href="#top">Ethical Tech <span>CoLab</span></a>
<div class="nav-links">
<a href="#concept">Concept</a>
<a href="#experiments">Experiments</a>
<a href="#research">Research</a>
<a href="https://github.com/Ethical-Tech-CoLab/agentic-language-development">GitHub</a>
</div>
</nav>
</header>
<main id="top">
<div class="shell hero">
<div>
<p class="eyebrow">Research concept / Pre-specification</p>
<h1>Can agents invent a <span class="gradient">language?</span></h1>
<p class="lede">
Two isolated agents. No human language between them. Shared experiences,
independent meaning ledgers, and a silent observer preserving the full
history of how understanding emerges.
</p>
<div class="actions">
<a class="button primary" href="research-book.html?open=1">
Read research as a book
</a>
<a
class="button"
href="https://github.com/Ethical-Tech-CoLab/agentic-language-development/blob/main/CONCEPT-IDEA.md"
>
Read the detailed concept
</a>
<a
class="button"
href="https://github.com/Ethical-Tech-CoLab/agentic-language-development"
>
Explore the repository
</a>
</div>
</div>
<aside class="architecture" aria-label="Nursery Lab conceptual architecture">
<div class="role">
<strong>BabySitter / Nursery</strong>
<small>Controls the experiment, observes every turn, teaches nothing.</small>
</div>
<div class="babies">
<div class="role">
<strong>Baby A</strong>
<small>Private state, learner, independent ledger.</small>
</div>
<div class="role">
<strong>Baby B</strong>
<small>Private state, learner, independent ledger.</small>
</div>
</div>
<div class="channel">Symbol Gateway // only permitted route</div>
<div class="role">
<strong>Immutable evidence</strong>
<small>Hash-chained ledgers, signed Merkle checkpoints, Base anchors, and replay.</small>
</div>
</aside>
</div>
<section id="concept">
<div class="shell">
<p class="eyebrow">The premise</p>
<h2>A grounded protocol, not a disguised dictionary.</h2>
<p class="section-intro">
Baby A and Baby B encounter shared objects, actions, relationships, and
consequences. They exchange only permitted non-human signals and revise
private hypotheses about what those signals mean.
</p>
<div class="grid">
<article class="card">
<span class="number">01</span>
<h3>Isolated twins</h3>
<p>Separate DTSF twins prevent shared memory, direct endpoints, and hidden coordination.</p>
</article>
<article class="card">
<span class="number">02</span>
<h3>Shared experience</h3>
<p>Meaning is grounded through joint attention, action, success, failure, and repair.</p>
</article>
<article class="card">
<span class="number">03</span>
<h3>Independent ledgers</h3>
<p>Each Baby records first use, provisional meaning, evidence, contradiction, and revision.</p>
</article>
<article class="card">
<span class="number">04</span>
<h3>Controlled channel</h3>
<p>A deterministic gateway rejects human language, tools, attachments, and side routes.</p>
</article>
<article class="card">
<span class="number">05</span>
<h3>Causal evaluation</h3>
<p>Symbol swaps and ablations test whether messages actually change receiver behavior.</p>
</article>
<article class="card">
<span class="number">06</span>
<h3>Replayable evidence</h3>
<p>Every run preserves configuration, observations, turns, outcomes, and policy checkpoints.</p>
</article>
</div>
</div>
</section>
<section>
<div class="shell callout">
<p class="eyebrow">An essential caveat</p>
<h2>"Infant-like" is an analogy, not a claim of equivalence.</h2>
<p>
Pretrained language models already contain human-language knowledge.
Restricting their external channel does not make them language-naive. With
those models, the experiment studies a newly shared external protocol.
</p>
<p>
Stronger language-acquisition claims require initially ungrounded trainable
agents. Results must always identify the agent type, learning mechanism,
observation encoding, and reward conditions.
</p>
</div>
</section>
<section id="experiments">
<div class="shell">
<p class="eyebrow">Experimental directions</p>
<h2>Change one condition. Observe what language becomes.</h2>
<p class="section-intro">
The concept defines independent experiment axes so affect, symbol creation,
motivation, negotiation, and protective encodings can be tested rather than
blended into one ambiguous run.
</p>
<div class="grid">
<article class="card">
<span class="number">AFFECT</span>
<h3>Emoji and emotion</h3>
<p>Allow only six declared displays in fixed feedback windows, with no general emoji set or sequences.</p>
</article>
<article class="card">
<span class="number">FORM</span>
<h3>No symbol library</h3>
<p>Replace fixed tokens with a blank sketch canvas, tones, gestures, or neutral marks.</p>
</article>
<article class="card">
<span class="number">MOTIVE</span>
<h3>Endogenous giddiness</h3>
<p>Test curiosity, social influence, prediction progress, and self-generated goals.</p>
</article>
<article class="card">
<span class="number">CIPHER</span>
<h3>Ephemeral conventions</h3>
<p>Separate code novelty from cryptographic security and compare against per-message keys.</p>
</article>
<article class="card">
<span class="number">TABLE</span>
<h3>Negotiation</h3>
<p>Progress from cooperative signaling to private preferences, offers, and strategic ambiguity.</p>
</article>
<article class="card">
<span class="number">MODEL</span>
<h3>RL versus non-RL</h3>
<p>Compare frozen LLM memory, extrinsic and intrinsic MARL, self-supervision, and no learning.</p>
</article>
</div>
</div>
</section>
<section id="research">
<div class="shell">
<p class="eyebrow">Research foundation</p>
<h2>Built on emergent communication, with stricter auditability.</h2>
<p class="section-intro">
The concept draws on referential games, grounded multi-agent learning,
compositionality research, causal communication metrics, intrinsic social
motivation, graphical signaling, negotiation, and adversarial neural
cryptography.
</p>
<div class="research-list">
<div class="research-item">
<strong>Grounding matters.</strong>
<p>Shared environments and actions provide evidence from which symbol meaning can emerge.</p>
</div>
<div class="research-item">
<strong>Success can mislead.</strong>
<p>Agents may solve a task with a brittle lookup code or messages the receiver ignores.</p>
</div>
<div class="research-item">
<strong>Structure is not guaranteed.</strong>
<p>Bandwidth, memory, task pressure, and representation shape compositionality.</p>
</div>
<div class="research-item">
<strong>Intervention is essential.</strong>
<p>Ablation and substitution are stronger evidence than a fluent transcript or ledger alone.</p>
</div>
</div>
</div>
</section>
<section>
<div class="shell cta">
<p class="eyebrow">Go deeper</p>
<h2>Read the complete concept and research review.</h2>
<p>
The detailed document covers the architecture, channel contract, mandatory
ledgers, learning loop, research sources, experimental matrix, cipher
caveats, risks, success criteria, and decisions required for the future
specification. Separate documents define ledger proofs and the ordered
researcher notebook.
</p>
<div class="actions">
<a
class="button primary"
href="research-book.html?open=1"
>
Open page-turn edition
</a>
<a
class="button"
href="https://github.com/Ethical-Tech-CoLab/agentic-language-development/blob/main/CONCEPT-IDEA.md"
>
Open CONCEPT-IDEA.md
</a>
<a
class="button"
href="https://github.com/Ethical-Tech-CoLab/agentic-language-development/blob/main/LEDGER-INTEGRITY-DESIGN.md"
>
Ledger integrity design
</a>
<a
class="button"
href="https://github.com/Ethical-Tech-CoLab/agentic-language-development/blob/main/EXPERIMENT-NOTEBOOK.md"
>
Experiment notebook
</a>
<a
class="button"
href="https://github.com/Ethical-Tech-CoLab/agentic-language-development/blob/main/SPECIFICATION.md"
>
Implementation specification
</a>
<a
class="button"
href="https://github.com/Ethical-Tech-CoLab/agentic-language-development/blob/main/BACKLOG.md"
>
Delivery backlog
</a>
<a
class="button"
href="https://github.com/Ethical-Tech-CoLab/agentic-language-development/blob/main/RESEARCH.md"
>
Research manuscript
</a>
<a
class="button"
href="https://github.com/Ethical-Tech-CoLab/diplomacy-table-live"
>
See Diplomacy Table
</a>
</div>
</div>
</section>
</main>
<footer>
<div class="shell footer-row">
<span>Agentic Language Development</span>
<span>An Ethical Tech CoLab research concept · MIT licensed</span>
</div>
</footer>
</body>
</html>