ABOUT THE SPEAKER
Rajesh Rao - Computational neuroscientist
Rajesh Rao seeks to understand the human brain through computational modeling, on two fronts: developing computer models of our minds, and using tech to decipher the 4,000-year-old lost script of the Indus Valley civilization.

Why you should listen

Rajesh Rao is looking for the computational principles underlying the brain's remarkable ability to learn, process and store information --  hoping to apply this knowledge to the task of building adaptive robotic systems and artificially intelligent agents.

Some of the questions that motivate his research include: How does the brain learn efficient representations of novel objects and events occurring in the natural environment? What are the algorithms that allow useful sensorimotor routines and behaviors to be learned? What computational mechanisms allow the brain to adapt to changing circumstances and remain fault-tolerant and robust?

By investigating these questions within a computational and probabilistic framework, it is often possible to derive algorithms that not only provide functional interpretations of neurobiological properties but also suggest solutions to difficult problems in computer vision, speech, robotics and artificial intelligence.

More profile about the speaker
Rajesh Rao | Speaker | TED.com
TED2011

Rajesh Rao: A Rosetta Stone for a lost language

Filmed:
2,103,451 views

Rajesh Rao is fascinated by "the mother of all crossword puzzles": how to decipher the 4000-year-old Indus script. He's enlisting modern computation to try to read this lost language, the key to understanding this ancient civilization.
- Computational neuroscientist
Rajesh Rao seeks to understand the human brain through computational modeling, on two fronts: developing computer models of our minds, and using tech to decipher the 4,000-year-old lost script of the Indus Valley civilization. Full bio

Double-click the English transcript below to play the video.

00:15
I'd like to begin with a thought experiment.
0
0
3000
00:19
Imagine that it's 4,000 years into the future.
1
4000
3000
00:22
Civilization as we know it
2
7000
2000
00:24
has ceased to exist --
3
9000
2000
00:26
no books,
4
11000
2000
00:28
no electronic devices,
5
13000
3000
00:31
no Facebook or Twitter.
6
16000
3000
00:34
All knowledge of the English language and the English alphabet
7
19000
3000
00:37
has been lost.
8
22000
2000
00:39
Now imagine archeologists
9
24000
2000
00:41
digging through the rubble of one of our cities.
10
26000
2000
00:43
What might they find?
11
28000
2000
00:45
Well perhaps some rectangular pieces of plastic
12
30000
3000
00:48
with strange symbols on them.
13
33000
3000
00:51
Perhaps some circular pieces of metal.
14
36000
3000
00:54
Maybe some cylindrical containers
15
39000
2000
00:56
with some symbols on them.
16
41000
2000
00:58
And perhaps one archeologist becomes an instant celebrity
17
43000
3000
01:01
when she discovers --
18
46000
2000
01:03
buried in the hills somewhere in North America --
19
48000
2000
01:05
massive versions of these same symbols.
20
50000
3000
01:10
Now let's ask ourselves,
21
55000
2000
01:12
what could such artifacts say about us
22
57000
3000
01:15
to people 4,000 years into the future?
23
60000
3000
01:18
This is no hypothetical question.
24
63000
2000
01:20
In fact, this is exactly the kind of question we're faced with
25
65000
3000
01:23
when we try to understand the Indus Valley civilization,
26
68000
3000
01:26
which existed 4,000 years ago.
27
71000
2000
01:28
The Indus civilization was roughly contemporaneous
28
73000
3000
01:31
with the much better known Egyptian and the Mesopotamian civilizations,
29
76000
3000
01:34
but it was actually much larger than either of these two civilizations.
30
79000
3000
01:37
It occupied the area
31
82000
2000
01:39
of approximately one million square kilometers,
32
84000
2000
01:41
covering what is now Pakistan,
33
86000
2000
01:43
Northwestern India
34
88000
2000
01:45
and parts of Afghanistan and Iran.
35
90000
2000
01:47
Given that it was such a vast civilization,
36
92000
2000
01:49
you might expect to find really powerful rulers, kings,
37
94000
4000
01:53
and huge monuments glorifying these powerful kings.
38
98000
3000
01:56
In fact,
39
101000
2000
01:58
what archeologists have found is none of that.
40
103000
2000
02:00
They've found small objects such as these.
41
105000
3000
02:03
Here's an example of one of these objects.
42
108000
3000
02:06
Well obviously this is a replica.
43
111000
2000
02:08
But who is this person?
44
113000
3000
02:11
A king? A god?
45
116000
2000
02:13
A priest?
46
118000
2000
02:15
Or perhaps an ordinary person
47
120000
2000
02:17
like you or me?
48
122000
2000
02:19
We don't know.
49
124000
2000
02:21
But the Indus people also left behind artifacts with writing on them.
50
126000
3000
02:24
Well no, not pieces of plastic,
51
129000
2000
02:26
but stone seals, copper tablets,
52
131000
3000
02:29
pottery and, surprisingly,
53
134000
2000
02:31
one large sign board,
54
136000
2000
02:33
which was found buried near the gate of a city.
55
138000
2000
02:35
Now we don't know if it says Hollywood,
56
140000
2000
02:37
or even Bollywood for that matter.
57
142000
2000
02:39
In fact, we don't even know
58
144000
2000
02:41
what any of these objects say,
59
146000
2000
02:43
and that's because the Indus script is undeciphered.
60
148000
3000
02:46
We don't know what any of these symbols mean.
61
151000
2000
02:48
The symbols are most commonly found on seals.
62
153000
3000
02:51
So you see up there one such object.
63
156000
2000
02:53
It's the square object with the unicorn-like animal on it.
64
158000
3000
02:56
Now that's a magnificent piece of art.
65
161000
2000
02:58
So how big do you think that is?
66
163000
2000
03:00
Perhaps that big?
67
165000
2000
03:02
Or maybe that big?
68
167000
2000
03:04
Well let me show you.
69
169000
2000
03:07
Here's a replica of one such seal.
70
172000
3000
03:10
It's only about one inch by one inch in size --
71
175000
2000
03:12
pretty tiny.
72
177000
2000
03:14
So what were these used for?
73
179000
2000
03:16
We know that these were used for stamping clay tags
74
181000
3000
03:19
that were attached to bundles of goods that were sent from one place to the other.
75
184000
3000
03:22
So you know those packing slips you get on your FedEx boxes?
76
187000
3000
03:25
These were used to make those kinds of packing slips.
77
190000
3000
03:28
You might wonder what these objects contain
78
193000
3000
03:31
in terms of their text.
79
196000
2000
03:33
Perhaps they're the name of the sender
80
198000
2000
03:35
or some information about the goods
81
200000
2000
03:37
that are being sent from one place to the other -- we don't know.
82
202000
3000
03:40
We need to decipher the script to answer that question.
83
205000
2000
03:42
Deciphering the script
84
207000
2000
03:44
is not just an intellectual puzzle;
85
209000
2000
03:46
it's actually become a question
86
211000
2000
03:48
that's become deeply intertwined
87
213000
2000
03:50
with the politics and the cultural history of South Asia.
88
215000
3000
03:53
In fact, the script has become a battleground of sorts
89
218000
3000
03:56
between three different groups of people.
90
221000
2000
03:58
First, there's a group of people
91
223000
2000
04:00
who are very passionate in their belief
92
225000
2000
04:02
that the Indus script
93
227000
2000
04:04
does not represent a language at all.
94
229000
2000
04:06
These people believe that the symbols
95
231000
2000
04:08
are very similar to the kind of symbols you find on traffic signs
96
233000
3000
04:11
or the emblems you find on shields.
97
236000
3000
04:14
There's a second group of people
98
239000
2000
04:16
who believe that the Indus script represents an Indo-European language.
99
241000
3000
04:19
If you look at a map of India today,
100
244000
2000
04:21
you'll see that most of the languages spoken in North India
101
246000
3000
04:24
belong to the Indo-European language family.
102
249000
3000
04:27
So some people believe that the Indus script
103
252000
2000
04:29
represents an ancient Indo-European language such as Sanskrit.
104
254000
3000
04:32
There's a last group of people
105
257000
2000
04:34
who believe that the Indus people
106
259000
3000
04:37
were the ancestors of people living in South India today.
107
262000
3000
04:40
These people believe that the Indus script
108
265000
2000
04:42
represents an ancient form
109
267000
2000
04:44
of the Dravidian language family,
110
269000
2000
04:46
which is the language family spoken in much of South India today.
111
271000
3000
04:49
And the proponents of this theory
112
274000
2000
04:51
point to that small pocket of Dravidian-speaking people in the North,
113
276000
3000
04:54
actually near Afghanistan,
114
279000
2000
04:56
and they say that perhaps, sometime in the past,
115
281000
3000
04:59
Dravidian languages were spoken all over India
116
284000
3000
05:02
and that this suggests
117
287000
2000
05:04
that the Indus civilization is perhaps also Dravidian.
118
289000
3000
05:07
Which of these hypotheses can be true?
119
292000
3000
05:10
We don't know, but perhaps if you deciphered the script,
120
295000
2000
05:12
you would be able to answer this question.
121
297000
2000
05:14
But deciphering the script is a very challenging task.
122
299000
2000
05:16
First, there's no Rosetta Stone.
123
301000
2000
05:18
I don't mean the software;
124
303000
2000
05:20
I mean an ancient artifact
125
305000
2000
05:22
that contains in the same text
126
307000
2000
05:24
both a known text and an unknown text.
127
309000
3000
05:27
We don't have such an artifact for the Indus script.
128
312000
3000
05:30
And furthermore, we don't even know what language they spoke.
129
315000
3000
05:33
And to make matters even worse,
130
318000
2000
05:35
most of the text that we have are extremely short.
131
320000
2000
05:37
So as I showed you, they're usually found on these seals
132
322000
2000
05:39
that are very, very tiny.
133
324000
2000
05:41
And so given these formidable obstacles,
134
326000
2000
05:43
one might wonder and worry
135
328000
2000
05:45
whether one will ever be able to decipher the Indus script.
136
330000
3000
05:48
In the rest of my talk,
137
333000
2000
05:50
I'd like to tell you about how I learned to stop worrying
138
335000
2000
05:52
and love the challenge posed by the Indus script.
139
337000
2000
05:54
I've always been fascinated by the Indus script
140
339000
3000
05:57
ever since I read about it in a middle school textbook.
141
342000
2000
05:59
And why was I fascinated?
142
344000
2000
06:01
Well it's the last major undeciphered script in the ancient world.
143
346000
4000
06:05
My career path led me to become a computational neuroscientist,
144
350000
3000
06:08
so in my day job,
145
353000
2000
06:10
I create computer models of the brain
146
355000
2000
06:12
to try to understand how the brain makes predictions,
147
357000
3000
06:15
how the brain makes decisions,
148
360000
2000
06:17
how the brain learns and so on.
149
362000
2000
06:19
But in 2007, my path crossed again with the Indus script.
150
364000
3000
06:22
That's when I was in India,
151
367000
2000
06:24
and I had the wonderful opportunity
152
369000
2000
06:26
to meet with some Indian scientists
153
371000
2000
06:28
who were using computer models to try to analyze the script.
154
373000
3000
06:31
And so it was then that I realized
155
376000
2000
06:33
there was an opportunity for me to collaborate with these scientists,
156
378000
3000
06:36
and so I jumped at that opportunity.
157
381000
2000
06:38
And I'd like to describe some of the results that we have found.
158
383000
2000
06:40
Or better yet, let's all collectively decipher.
159
385000
3000
06:43
Are you ready?
160
388000
2000
06:45
The first thing that you need to do when you have an undeciphered script
161
390000
3000
06:48
is try to figure out the direction of writing.
162
393000
2000
06:50
Here are two texts that contain some symbols on them.
163
395000
3000
06:53
Can you tell me
164
398000
2000
06:55
if the direction of writing is right to left or left to right?
165
400000
3000
06:58
I'll give you a couple of seconds.
166
403000
3000
07:01
Okay. Right to left, how many? Okay.
167
406000
3000
07:04
Okay. Left to right?
168
409000
2000
07:06
Oh, it's almost 50/50. Okay.
169
411000
2000
07:08
The answer is:
170
413000
2000
07:10
if you look at the left-hand side of the two texts,
171
415000
2000
07:12
you'll notice that there's a cramping of signs,
172
417000
3000
07:15
and it seems like 4,000 years ago,
173
420000
2000
07:17
when the scribe was writing from right to left,
174
422000
2000
07:19
they ran out of space.
175
424000
2000
07:21
And so they had to cram the sign.
176
426000
2000
07:23
One of the signs is also below the text on the top.
177
428000
2000
07:25
This suggests the direction of writing
178
430000
2000
07:27
was probably from right to left,
179
432000
2000
07:29
and so that's one of the first things we know,
180
434000
2000
07:31
that directionality is a very key aspect of linguistic scripts.
181
436000
3000
07:34
And the Indus script now has
182
439000
2000
07:36
this particular property.
183
441000
2000
07:38
What other properties of language does the script show?
184
443000
2000
07:40
Languages contain patterns.
185
445000
2000
07:42
If I give you the letter Q
186
447000
2000
07:44
and ask you to predict the next letter, what do you think that would be?
187
449000
3000
07:47
Most of you said U, which is right.
188
452000
2000
07:49
Now if I asked you to predict one more letter,
189
454000
2000
07:51
what do you think that would be?
190
456000
2000
07:53
Now there's several thoughts. There's E. It could be I. It could be A,
191
458000
3000
07:56
but certainly not B, C or D, right?
192
461000
3000
07:59
The Indus script also exhibits similar kinds of patterns.
193
464000
3000
08:02
There's a lot of text that start with this diamond-shaped symbol.
194
467000
3000
08:05
And this in turn tends to be followed
195
470000
2000
08:07
by this quotation marks-like symbol.
196
472000
2000
08:09
And this is very similar to a Q and U example.
197
474000
2000
08:11
This symbol can in turn be followed
198
476000
2000
08:13
by these fish-like symbols and some other signs,
199
478000
3000
08:16
but never by these other signs at the bottom.
200
481000
2000
08:18
And furthermore, there's some signs
201
483000
2000
08:20
that really prefer the end of texts,
202
485000
2000
08:22
such as this jar-shaped sign,
203
487000
2000
08:24
and this sign, in fact, happens to be
204
489000
2000
08:26
the most frequently occurring sign in the script.
205
491000
2000
08:28
Given such patterns, here was our idea.
206
493000
3000
08:31
The idea was to use a computer
207
496000
2000
08:33
to learn these patterns,
208
498000
2000
08:35
and so we gave the computer the existing texts.
209
500000
3000
08:38
And the computer learned a statistical model
210
503000
2000
08:40
of which symbols tend to occur together
211
505000
2000
08:42
and which symbols tend to follow each other.
212
507000
2000
08:44
Given the computer model,
213
509000
2000
08:46
we can test the model by essentially quizzing it.
214
511000
3000
08:49
So we could deliberately erase some symbols,
215
514000
2000
08:51
and we can ask it to predict the missing symbols.
216
516000
3000
08:54
Here are some examples.
217
519000
3000
09:00
You may regard this
218
525000
2000
09:02
as perhaps the most ancient game
219
527000
2000
09:04
of Wheel of Fortune.
220
529000
3000
09:08
What we found
221
533000
2000
09:10
was that the computer was successful in 75 percent of the cases
222
535000
2000
09:12
in predicting the correct symbol.
223
537000
2000
09:14
In the rest of the cases,
224
539000
2000
09:16
typically the second best guess or third best guess was the right answer.
225
541000
3000
09:19
There's also practical use
226
544000
2000
09:21
for this particular procedure.
227
546000
2000
09:23
There's a lot of these texts that are damaged.
228
548000
2000
09:25
Here's an example of one such text.
229
550000
2000
09:27
And we can use the computer model now to try to complete this text
230
552000
3000
09:30
and make a best guess prediction.
231
555000
2000
09:32
Here's an example of a symbol that was predicted.
232
557000
3000
09:35
And this could be really useful as we try to decipher the script
233
560000
2000
09:37
by generating more data that we can analyze.
234
562000
3000
09:40
Now here's one other thing you can do with the computer model.
235
565000
3000
09:43
So imagine a monkey
236
568000
2000
09:45
sitting at a keyboard.
237
570000
2000
09:47
I think you might get a random jumble of letters that looks like this.
238
572000
3000
09:50
Such a random jumble of letters
239
575000
2000
09:52
is said to have a very high entropy.
240
577000
2000
09:54
This is a physics and information theory term.
241
579000
2000
09:56
But just imagine it's a really random jumble of letters.
242
581000
3000
09:59
How many of you have ever spilled coffee on a keyboard?
243
584000
4000
10:03
You might have encountered the stuck-key problem --
244
588000
2000
10:05
so basically the same symbol being repeated over and over again.
245
590000
3000
10:08
This kind of a sequence is said to have a very low entropy
246
593000
3000
10:11
because there's no variation at all.
247
596000
2000
10:13
Language, on the other hand, has an intermediate level of entropy;
248
598000
3000
10:16
it's neither too rigid,
249
601000
2000
10:18
nor is it too random.
250
603000
2000
10:20
What about the Indus script?
251
605000
2000
10:22
Here's a graph that plots the entropies of a whole bunch of sequences.
252
607000
4000
10:26
At the very top you find the uniformly random sequence,
253
611000
2000
10:28
which is a random jumble of letters --
254
613000
2000
10:30
and interestingly, we also find
255
615000
2000
10:32
the DNA sequence from the human genome and instrumental music.
256
617000
3000
10:35
And both of these are very, very flexible,
257
620000
2000
10:37
which is why you find them in the very high range.
258
622000
2000
10:39
At the lower end of the scale,
259
624000
2000
10:41
you find a rigid sequence, a sequence of all A's,
260
626000
2000
10:43
and you also find a computer program,
261
628000
2000
10:45
in this case in the language Fortran,
262
630000
2000
10:47
which obeys really strict rules.
263
632000
2000
10:49
Linguistic scripts
264
634000
2000
10:51
occupy the middle range.
265
636000
2000
10:53
Now what about the Indus script?
266
638000
2000
10:55
We found that the Indus script
267
640000
2000
10:57
actually falls within the range of the linguistic scripts.
268
642000
2000
10:59
When this result was first published,
269
644000
2000
11:01
it was highly controversial.
270
646000
3000
11:04
There were people who raised a hue and cry,
271
649000
3000
11:07
and these people were the ones who believed
272
652000
2000
11:09
that the Indus script does not represent language.
273
654000
3000
11:12
I even started to get some hate mail.
274
657000
2000
11:14
My students said
275
659000
2000
11:16
that I should really seriously consider getting some protection.
276
661000
3000
11:19
Who'd have thought
277
664000
2000
11:21
that deciphering could be a dangerous profession?
278
666000
2000
11:23
What does this result really show?
279
668000
2000
11:25
It shows that the Indus script
280
670000
2000
11:27
shares an important property of language.
281
672000
2000
11:29
So, as the old saying goes,
282
674000
2000
11:31
if it looks like a linguistic script
283
676000
2000
11:33
and it acts like a linguistic script,
284
678000
2000
11:35
then perhaps we may have a linguistic script on our hands.
285
680000
3000
11:38
What other evidence is there
286
683000
2000
11:40
that the script could actually encode language?
287
685000
2000
11:42
Well linguistic scripts can actually encode multiple languages.
288
687000
3000
11:45
So for example, here's the same sentence written in English
289
690000
3000
11:48
and the same sentence written in Dutch
290
693000
2000
11:50
using the same letters of the alphabet.
291
695000
2000
11:52
If you don't know Dutch and you only know English
292
697000
3000
11:55
and I give you some words in Dutch,
293
700000
2000
11:57
you'll tell me that these words contain
294
702000
2000
11:59
some very unusual patterns.
295
704000
2000
12:01
Some things are not right,
296
706000
2000
12:03
and you'll say these words are probably not English words.
297
708000
3000
12:06
The same thing happens in the case of the Indus script.
298
711000
2000
12:08
The computer found several texts --
299
713000
2000
12:10
two of them are shown here --
300
715000
2000
12:12
that have very unusual patterns.
301
717000
2000
12:14
So for example the first text:
302
719000
2000
12:16
there's a doubling of this jar-shaped sign.
303
721000
3000
12:19
This sign is the most frequently-occurring sign
304
724000
2000
12:21
in the Indus script,
305
726000
2000
12:23
and it's only in this text
306
728000
2000
12:25
that it occurs as a doubling pair.
307
730000
2000
12:27
Why is that the case?
308
732000
2000
12:29
We went back and looked at where these particular texts were found,
309
734000
3000
12:32
and it turns out that they were found
310
737000
2000
12:34
very, very far away from the Indus Valley.
311
739000
2000
12:36
They were found in present day Iraq and Iran.
312
741000
3000
12:39
And why were they found there?
313
744000
2000
12:41
What I haven't told you is that
314
746000
2000
12:43
the Indus people were very, very enterprising.
315
748000
2000
12:45
They used to trade with people pretty far away from where they lived,
316
750000
3000
12:48
and so in this case, they were traveling by sea
317
753000
3000
12:51
all the way to Mesopotamia, present-day Iraq.
318
756000
3000
12:54
And what seems to have happened here
319
759000
2000
12:56
is that the Indus traders, the merchants,
320
761000
3000
12:59
were using this script to write a foreign language.
321
764000
3000
13:02
It's just like our English and Dutch example.
322
767000
2000
13:04
And that would explain why we have these strange patterns
323
769000
2000
13:06
that are very different from the kinds of patterns you see in the text
324
771000
3000
13:09
that are found within the Indus Valley.
325
774000
3000
13:12
This suggests that the same script, the Indus script,
326
777000
2000
13:14
could be used to write different languages.
327
779000
3000
13:17
The results we have so far seem to point to the conclusion
328
782000
3000
13:20
that the Indus script probably does represent language.
329
785000
3000
13:23
If it does represent language,
330
788000
2000
13:25
then how do we read the symbols?
331
790000
2000
13:27
That's our next big challenge.
332
792000
2000
13:29
So you'll notice that many of the symbols
333
794000
2000
13:31
look like pictures of humans, of insects,
334
796000
2000
13:33
of fishes, of birds.
335
798000
3000
13:36
Most ancient scripts
336
801000
2000
13:38
use the rebus principle,
337
803000
2000
13:40
which is, using pictures to represent words.
338
805000
3000
13:43
So as an example, here's a word.
339
808000
3000
13:46
Can you write it using pictures?
340
811000
2000
13:48
I'll give you a couple seconds.
341
813000
2000
13:50
Got it?
342
815000
2000
13:52
Okay. Great.
343
817000
2000
13:54
Here's my solution.
344
819000
2000
13:56
You could use the picture of a bee followed by a picture of a leaf --
345
821000
2000
13:58
and that's "belief," right.
346
823000
2000
14:00
There could be other solutions.
347
825000
2000
14:02
In the case of the Indus script,
348
827000
2000
14:04
the problem is the reverse.
349
829000
2000
14:06
You have to figure out the sounds of each of these pictures
350
831000
3000
14:09
such that the entire sequence makes sense.
351
834000
2000
14:11
So this is just like a crossword puzzle,
352
836000
3000
14:14
except that this is the mother of all crossword puzzles
353
839000
3000
14:17
because the stakes are so high if you solve it.
354
842000
4000
14:21
My colleagues, Iravatham Mahadevan and Asko Parpola,
355
846000
3000
14:24
have been making some headway on this particular problem.
356
849000
2000
14:26
And I'd like to give you a quick example of Parpola's work.
357
851000
2000
14:28
Here's a really short text.
358
853000
2000
14:30
It contains seven vertical strokes followed by this fish-like sign.
359
855000
3000
14:33
And I want to mention that these seals were used
360
858000
2000
14:35
for stamping clay tags
361
860000
2000
14:37
that were attached to bundles of goods,
362
862000
2000
14:39
so it's quite likely that these tags, at least some of them,
363
864000
3000
14:42
contain names of merchants.
364
867000
2000
14:44
And it turns out that in India
365
869000
2000
14:46
there's a long tradition
366
871000
2000
14:48
of names being based on horoscopes
367
873000
2000
14:50
and star constellations present at the time of birth.
368
875000
3000
14:53
In Dravidian languages,
369
878000
2000
14:55
the word for fish is "meen"
370
880000
2000
14:57
which happens to sound just like the word for star.
371
882000
3000
15:00
And so seven stars
372
885000
2000
15:02
would stand for "elu meen,"
373
887000
2000
15:04
which is the Dravidian word
374
889000
2000
15:06
for the Big Dipper star constellation.
375
891000
2000
15:08
Similarly, there's another sequence of six stars,
376
893000
3000
15:11
and that translates to "aru meen,"
377
896000
2000
15:13
which is the old Dravidian name
378
898000
2000
15:15
for the star constellation Pleiades.
379
900000
2000
15:17
And finally, there's other combinations,
380
902000
3000
15:20
such as this fish sign with something that looks like a roof on top of it.
381
905000
3000
15:23
And that could be translated into "mey meen,"
382
908000
3000
15:26
which is the old Dravidian name for the planet Saturn.
383
911000
3000
15:29
So that was pretty exciting.
384
914000
2000
15:31
It looks like we're getting somewhere.
385
916000
2000
15:33
But does this prove
386
918000
2000
15:35
that these seals contain Dravidian names
387
920000
2000
15:37
based on planets and star constellations?
388
922000
2000
15:39
Well not yet.
389
924000
2000
15:41
So we have no way of validating
390
926000
2000
15:43
these particular readings,
391
928000
2000
15:45
but if more and more of these readings start making sense,
392
930000
3000
15:48
and if longer and longer sequences
393
933000
2000
15:50
appear to be correct,
394
935000
2000
15:52
then we know that we are on the right track.
395
937000
2000
15:54
Today,
396
939000
2000
15:56
we can write a word such as TED
397
941000
3000
15:59
in Egyptian hieroglyphics and in cuneiform script,
398
944000
3000
16:02
because both of these were deciphered
399
947000
2000
16:04
in the 19th century.
400
949000
2000
16:06
The decipherment of these two scripts
401
951000
2000
16:08
enabled these civilizations to speak to us again directly.
402
953000
3000
16:11
The Mayans
403
956000
2000
16:13
started speaking to us in the 20th century,
404
958000
2000
16:15
but the Indus civilization remains silent.
405
960000
3000
16:18
Why should we care?
406
963000
2000
16:20
The Indus civilization does not belong
407
965000
2000
16:22
to just the South Indians or the North Indians
408
967000
2000
16:24
or the Pakistanis;
409
969000
2000
16:26
it belongs to all of us.
410
971000
2000
16:28
These are our ancestors --
411
973000
2000
16:30
yours and mine.
412
975000
2000
16:32
They were silenced
413
977000
2000
16:34
by an unfortunate accident of history.
414
979000
2000
16:36
If we decipher the script,
415
981000
2000
16:38
we would enable them to speak to us again.
416
983000
2000
16:40
What would they tell us?
417
985000
3000
16:43
What would we find out about them? About us?
418
988000
3000
16:46
I can't wait to find out.
419
991000
3000
16:49
Thank you.
420
994000
2000
16:51
(Applause)
421
996000
4000

▲Back to top

ABOUT THE SPEAKER
Rajesh Rao - Computational neuroscientist
Rajesh Rao seeks to understand the human brain through computational modeling, on two fronts: developing computer models of our minds, and using tech to decipher the 4,000-year-old lost script of the Indus Valley civilization.

Why you should listen

Rajesh Rao is looking for the computational principles underlying the brain's remarkable ability to learn, process and store information --  hoping to apply this knowledge to the task of building adaptive robotic systems and artificially intelligent agents.

Some of the questions that motivate his research include: How does the brain learn efficient representations of novel objects and events occurring in the natural environment? What are the algorithms that allow useful sensorimotor routines and behaviors to be learned? What computational mechanisms allow the brain to adapt to changing circumstances and remain fault-tolerant and robust?

By investigating these questions within a computational and probabilistic framework, it is often possible to derive algorithms that not only provide functional interpretations of neurobiological properties but also suggest solutions to difficult problems in computer vision, speech, robotics and artificial intelligence.

More profile about the speaker
Rajesh Rao | Speaker | TED.com