1
00:00:00,305 --> 00:00:01,824
A mathematical theorem proves

2
00:00:01,825 --> 00:00:03,334
AI will never be conscious

3
00:00:03,705 --> 00:00:04,533
In 2024

4
00:00:04,865 --> 00:00:06,373
the whole world was debating one question

5
00:00:07,025 --> 00:00:08,454
Does AI actually have consciousness

6
00:00:08,984 --> 00:00:12,734
Large language models sound more and more like beings with souls

7
00:00:13,185 --> 00:00:14,794
Many people began to have doubts

8
00:00:14,945 --> 00:00:15,993
But one man

9
00:00:16,024 --> 00:00:18,474
gave his answer thirty years ago

10
00:00:18,625 --> 00:00:19,013
He said

11
00:00:19,425 --> 00:00:20,114
No

12
00:00:20,225 --> 00:00:21,274
Never

13
00:00:21,344 --> 00:00:22,634
And he could prove it

14
00:00:23,024 --> 00:00:24,333
This man was not a philosopher

15
00:00:24,625 --> 00:00:25,993
or a religious leader

16
00:00:26,065 --> 00:00:29,143
He won the 2020 Nobel Prize in Physics

17
00:00:29,144 --> 00:00:33,514
the mathematical physicist Roger Penrose a peer of Hawking

18
00:00:33,982 --> 00:00:34,330
Hello

19
00:00:34,581 --> 00:00:35,050
everyone

20
00:00:35,382 --> 00:00:36,431
I am Leo Wang

21
00:00:36,662 --> 00:00:39,431
Todays episode came from a viewers comment

22
00:00:39,901 --> 00:00:42,391
He had read many books about consciousness

23
00:00:42,581 --> 00:00:46,071
but still found Penroses theory the most convincing

24
00:00:46,302 --> 00:00:47,671
He asked me to explain

25
00:00:47,821 --> 00:00:49,211
Penroses central claim

26
00:00:49,662 --> 00:00:51,711
Consciousness is noncomputable

27
00:00:51,861 --> 00:00:55,230
Conventional AI and Turing machines cannot simulate it

28
00:00:55,622 --> 00:00:57,620
and the theory Penrose and Hameroff

29
00:00:57,622 --> 00:01:00,431
proposed together called orchestrated objective reduction

30
00:01:00,982 --> 00:01:02,791
I thought this was a wonderful topic

31
00:01:02,901 --> 00:01:05,060
because this theory truly reaches

32
00:01:05,062 --> 00:01:07,370
a question we keep approaching on this channel

33
00:01:07,702 --> 00:01:10,060
but have never confronted head on

34
00:01:10,062 --> 00:01:10,610
the central question

35
00:01:11,381 --> 00:01:13,491
Is consciousness a form of computation

36
00:01:14,861 --> 00:01:17,571
Penroses main claim fits into three words

37
00:01:17,941 --> 00:01:19,470
Consciousness is noncomputable

38
00:01:19,821 --> 00:01:21,420
If those words are right

39
00:01:21,422 --> 00:01:24,071
they matter more than any AI regulation

40
00:01:24,262 --> 00:01:25,420
because they mean

41
00:01:25,422 --> 00:01:29,470
no matter how much compute data or parameters AI gets

42
00:01:29,502 --> 00:01:31,390
no matter how conscious it appears

43
00:01:31,502 --> 00:01:32,890
to be

44
00:01:33,182 --> 00:01:35,030
it is only performing

45
00:01:35,062 --> 00:01:37,250
It will never truly wake up

46
00:01:37,982 --> 00:01:39,970
This conclusion rests neither on intuition

47
00:01:40,142 --> 00:01:41,071
nor faith

48
00:01:41,142 --> 00:01:43,181
but on one of the twentieth centurys deepest

49
00:01:43,182 --> 00:01:44,090
mathematical theorems

50
00:01:44,542 --> 00:01:46,449
Gödels incompleteness theorem

51
00:01:47,041 --> 00:01:48,089
In 1931

52
00:01:48,120 --> 00:01:51,089
the twenty five year old Kurt Gödel published a paper

53
00:01:51,200 --> 00:01:53,289
that shattered the foundations of mathematics

54
00:01:53,601 --> 00:01:54,709
He proved one thing

55
00:01:55,161 --> 00:01:58,720
Any sufficiently powerful consistent mathematical system

56
00:01:58,720 --> 00:02:02,410
must contain propositions that are true

57
00:02:02,441 --> 00:02:05,449
but can never be proved within that system

58
00:02:05,838 --> 00:02:07,726
Let me put that plainly

59
00:02:08,077 --> 00:02:10,047
Imagine being locked in a room

60
00:02:10,118 --> 00:02:12,366
with a complete set of reasoning rules

61
00:02:12,477 --> 00:02:14,767
You can only use those rules to judge truth and falsehood

62
00:02:15,037 --> 00:02:15,866
Gödel tells you

63
00:02:16,198 --> 00:02:18,637
that some truths must exist in that room

64
00:02:18,638 --> 00:02:20,406
that those rules can never reach

65
00:02:20,557 --> 00:02:23,276
However you upgrade the rules or add axioms

66
00:02:23,277 --> 00:02:26,246
new unreachable truths will always await you

67
00:02:26,277 --> 00:02:27,806
You cannot escape this bind

68
00:02:28,198 --> 00:02:29,627
What did that mean at the time

69
00:02:30,117 --> 00:02:31,027
Before Gödel

70
00:02:31,437 --> 00:02:33,007
the mathematical world believed in

71
00:02:33,158 --> 00:02:34,146
Hilberts dream

72
00:02:34,557 --> 00:02:36,317
that one day we could find

73
00:02:36,318 --> 00:02:38,116
a perfect axiomatic system

74
00:02:38,117 --> 00:02:40,967
that could derive every mathematical truth

75
00:02:41,397 --> 00:02:44,086
One paper by Gödel shattered that dream

76
00:02:44,478 --> 00:02:46,586
Mathematics can never be perfectly contained

77
00:02:47,078 --> 00:02:50,646
Every powerful enough system has blind spots it cannot see

78
00:02:51,158 --> 00:02:54,367
In his 1989 book The Emperors New Mind

79
00:02:54,397 --> 00:02:57,326
Penrose linked Gödels theorem to consciousness

80
00:02:57,517 --> 00:02:59,027
His chain of reasoning goes like this

81
00:02:59,438 --> 00:03:01,516
Computers all computers

82
00:03:01,517 --> 00:03:04,487
however powerful are essentially Turing machines

83
00:03:04,718 --> 00:03:07,326
And a Turing machine is a rule based system

84
00:03:07,517 --> 00:03:10,286
It executes instructions algorithmically step by step

85
00:03:10,598 --> 00:03:12,286
Gödels theorem tells us

86
00:03:12,357 --> 00:03:15,446
every rule system has true statements beyond its reach

87
00:03:15,758 --> 00:03:18,883
Yet human mathematicians can see those truths

88
00:03:18,973 --> 00:03:19,674
Notice

89
00:03:19,674 --> 00:03:23,163
this seeing does not come from a stronger algorithm

90
00:03:23,313 --> 00:03:25,082
You cannot write a program

91
00:03:25,114 --> 00:03:26,742
to discover why a Gödel statement

92
00:03:27,074 --> 00:03:28,312
is true

93
00:03:28,313 --> 00:03:32,802
because Gödel proved such a program cannot exist in principle

94
00:03:33,114 --> 00:03:37,362
Human mathematicians can understand why these statements are true

95
00:03:37,514 --> 00:03:38,793
through an ability

96
00:03:38,794 --> 00:03:40,143
that transcends algorithms

97
00:03:40,514 --> 00:03:42,073
Penrose calls it

98
00:03:42,074 --> 00:03:42,943
mathematical intuition

99
00:03:43,394 --> 00:03:44,622
or more precisely

100
00:03:44,913 --> 00:03:45,943
understanding

101
00:03:46,554 --> 00:03:48,943
This is the fulcrum of Penroses entire argument

102
00:03:49,473 --> 00:03:53,282
If the human mind can do what a Turing machine cannot

103
00:03:53,313 --> 00:03:55,883
then the human mind is not a Turing machine

104
00:03:56,193 --> 00:03:57,883
Consciousness is not computation

105
00:03:58,541 --> 00:03:59,491
You might object

106
00:03:59,941 --> 00:04:02,130
Perhaps humans are merely more complex computers

107
00:04:02,622 --> 00:04:04,190
We think we understand

108
00:04:04,222 --> 00:04:06,651
but are only performing more advanced symbol manipulation

109
00:04:07,222 --> 00:04:09,021
In his 1994

110
00:04:09,022 --> 00:04:11,220
second book Shadows of the Mind

111
00:04:11,222 --> 00:04:12,171
Penrose spent more than three hundred pages

112
00:04:12,381 --> 00:04:13,671
answering these objections

113
00:04:13,821 --> 00:04:15,500
His argument is extremely rigorous

114
00:04:15,502 --> 00:04:17,570
but its core can be reduced to one sentence

115
00:04:18,022 --> 00:04:20,591
If you insist that humans are Turing machines

116
00:04:20,621 --> 00:04:21,570
you must accept

117
00:04:21,942 --> 00:04:23,380
that human mathematicians abilities

118
00:04:23,382 --> 00:04:24,411
have a ceiling

119
00:04:24,782 --> 00:04:26,580
Some statements that are clearly true

120
00:04:26,582 --> 00:04:28,270
would remain forever beyond our recognition

121
00:04:29,142 --> 00:04:31,830
Yet the history of mathematics repeatedly shows

122
00:04:31,861 --> 00:04:34,140
that given the right Gödel construction

123
00:04:34,142 --> 00:04:37,150
mathematicians can always recognize the statement as true

124
00:04:37,501 --> 00:04:40,150
This suggests there is something in the human mind

125
00:04:40,301 --> 00:04:42,229
that does not obey algorithmic limits

126
00:04:42,461 --> 00:04:42,810
Of course

127
00:04:43,101 --> 00:04:44,500
I must be honest with you

128
00:04:44,500 --> 00:04:46,950
This argument remains highly controversial in academia

129
00:04:47,101 --> 00:04:49,670
Many logicians and philosophers reject it

130
00:04:49,901 --> 00:04:51,130
One powerful criticism is

131
00:04:51,500 --> 00:04:53,420
that Penroses argument assumes

132
00:04:53,421 --> 00:04:55,010
human mathematicians are reliable

133
00:04:55,341 --> 00:04:56,620
meaning they do not make

134
00:04:56,620 --> 00:04:57,829
systematic errors

135
00:04:58,021 --> 00:05:00,930
But what if human reasoning itself can fail

136
00:05:01,341 --> 00:05:05,510
That weakens the claim that Gödel proves humans superior to machines

137
00:05:05,740 --> 00:05:07,139
Penrose answered this too

138
00:05:07,140 --> 00:05:09,510
He admits individual mathematicians make mistakes

139
00:05:09,580 --> 00:05:12,260
but the mathematical community as a whole

140
00:05:12,260 --> 00:05:14,310
displays a capacity for mathematical understanding

141
00:05:14,341 --> 00:05:17,751
that no known formal system can capture

142
00:05:18,022 --> 00:05:20,072
This debate remains unsettled

143
00:05:20,263 --> 00:05:21,301
But either way

144
00:05:21,303 --> 00:05:23,291
Penrose raised a question we cannot avoid

145
00:05:23,943 --> 00:05:26,032
If consciousness really is only computation

146
00:05:26,183 --> 00:05:27,332
then what is understanding

147
00:05:28,022 --> 00:05:29,991
When you understand a mathematical proof

148
00:05:30,142 --> 00:05:31,751
is what happens in your brain

149
00:05:31,823 --> 00:05:33,791
the same as what happens inside a computer

150
00:05:33,863 --> 00:05:35,212
that verifies the proof

151
00:05:35,582 --> 00:05:36,412
Are they the same thing

152
00:05:37,323 --> 00:05:39,451
In 1980 John Searle

153
00:05:39,482 --> 00:05:41,311
proposed a famous thought experiment

154
00:05:41,643 --> 00:05:42,691
the Chinese room

155
00:05:43,162 --> 00:05:45,732
A person who knows no Chinese is locked in a room

156
00:05:45,802 --> 00:05:47,811
holding a rule book

157
00:05:47,922 --> 00:05:52,152
that says which Chinese symbols to output for each input

158
00:05:52,722 --> 00:05:55,212
Chinese speakers outside communicate with the room

159
00:05:55,323 --> 00:05:57,611
and think the person inside fully understands Chinese

160
00:05:57,883 --> 00:06:00,131
But that person understands none of it

161
00:06:00,523 --> 00:06:01,491
Penrose says

162
00:06:01,602 --> 00:06:02,832
all AI today

163
00:06:03,203 --> 00:06:05,111
including the most advanced large language models

164
00:06:05,602 --> 00:06:07,691
is essentially a Chinese room

165
00:06:07,962 --> 00:06:10,251
It can produce flawless output

166
00:06:10,323 --> 00:06:12,571
but no understanding occurs

167
00:06:13,203 --> 00:06:15,131
This is not a matter of immature technology

168
00:06:15,323 --> 00:06:20,426
The Turing machine architecture cannot produce understanding in principle

169
00:06:20,777 --> 00:06:20,967
Now

170
00:06:21,297 --> 00:06:23,137
suppose we temporarily accept

171
00:06:23,138 --> 00:06:24,087
Penroses premise

172
00:06:24,537 --> 00:06:26,227
that consciousness is noncomputable

173
00:06:26,417 --> 00:06:27,446
The next question is

174
00:06:27,857 --> 00:06:29,587
if consciousness is not computation

175
00:06:29,618 --> 00:06:30,606
then what is it

176
00:06:30,977 --> 00:06:32,327
What is its physical basis

177
00:06:32,777 --> 00:06:35,027
This is Penroses boldest step

178
00:06:35,217 --> 00:06:39,066
He collaborated with American anesthesiologist Stuart Hameroff

179
00:06:39,217 --> 00:06:41,999
to propose orchestrated objective reduction

180
00:06:42,431 --> 00:06:44,190
The name is a mouthful

181
00:06:44,191 --> 00:06:45,720
so let me unpack it layer by layer

182
00:06:46,150 --> 00:06:48,159
First what does objective reduction mean

183
00:06:48,510 --> 00:06:50,150
In standard quantum mechanics

184
00:06:50,150 --> 00:06:52,270
a particle can simultaneously occupy

185
00:06:52,270 --> 00:06:53,460
a superposition of multiple states

186
00:06:53,910 --> 00:06:55,630
both here and there

187
00:06:55,631 --> 00:06:57,400
both this and that

188
00:06:57,470 --> 00:06:59,280
Only when you observe it

189
00:06:59,350 --> 00:07:01,679
does it collapse into a definite state

190
00:07:01,871 --> 00:07:04,659
This is the measurement problem who is observing

191
00:07:05,111 --> 00:07:07,220
What exactly happens at the moment of observation

192
00:07:07,551 --> 00:07:10,639
It is one of quantum mechanics deepest mysteries

193
00:07:10,870 --> 00:07:13,079
Penrose offered a unique answer

194
00:07:13,389 --> 00:07:14,149
He believes

195
00:07:14,149 --> 00:07:17,639
there is a collapse mechanism requiring no external observer

196
00:07:17,909 --> 00:07:19,918
When a quantum superposition causes

197
00:07:20,029 --> 00:07:22,199
the geometry of spacetime to split and curve

198
00:07:22,269 --> 00:07:23,459
beyond a critical threshold

199
00:07:23,949 --> 00:07:25,418
roughly at the Planck scale

200
00:07:25,870 --> 00:07:27,418
spacetime itself cannot sustain

201
00:07:27,709 --> 00:07:28,379
that separation

202
00:07:28,870 --> 00:07:31,158
so the superposition collapses spontaneously

203
00:07:31,509 --> 00:07:34,158
This collapse follows no algorithm

204
00:07:34,190 --> 00:07:36,798
It is a genuinely noncomputable physical process

205
00:07:37,269 --> 00:07:38,019
Penrose believes

206
00:07:38,269 --> 00:07:39,868
each such collapse

207
00:07:39,870 --> 00:07:44,480
is a proto conscious event a flash of subjective experience

208
00:07:44,751 --> 00:07:47,340
So where in the brain does this collapse occur

209
00:07:47,951 --> 00:07:49,840
That is Hameroffs contribution

210
00:07:50,112 --> 00:07:54,720
He noticed structures inside brain neurons called microtubules

211
00:07:55,071 --> 00:07:57,800
Microtubules are assembled from tubulin proteins

212
00:07:58,071 --> 00:08:01,480
Scientists once thought they were merely cellular scaffolding

213
00:08:01,951 --> 00:08:04,300
But Hameroff found a crucial clue

214
00:08:04,831 --> 00:08:06,641
All general anesthetics

215
00:08:06,672 --> 00:08:08,761
however different their chemistry

216
00:08:08,831 --> 00:08:10,920
act on tubulin proteins

217
00:08:11,232 --> 00:08:13,780
And what is the core effect of general anesthesia

218
00:08:14,232 --> 00:08:16,800
It precisely switches off your subjective experience

219
00:08:17,191 --> 00:08:19,021
It does not switch off electrical activity in the brain

220
00:08:19,511 --> 00:08:20,800
While you are anesthetized

221
00:08:20,872 --> 00:08:22,220
your EEG still shows signals

222
00:08:22,631 --> 00:08:24,360
Instead it switches off your consciousness

223
00:08:25,191 --> 00:08:27,141
If microtubules are only scaffolding

224
00:08:27,672 --> 00:08:28,821
why affect them

225
00:08:29,352 --> 00:08:30,701
Why does disturbing the scaffolding

226
00:08:31,151 --> 00:08:32,340
make consciousness disappear

227
00:08:33,363 --> 00:08:36,632
The core hypothesis of orchestrated objective reduction is

228
00:08:37,042 --> 00:08:40,211
that tubulin proteins can sustain quantum superposition

229
00:08:40,443 --> 00:08:44,451
When vast numbers of tubulin proteins in the brain enter superposition together

230
00:08:44,522 --> 00:08:49,201
and collapse through Penroses objective reduction mechanism

231
00:08:49,203 --> 00:08:50,811
consciousness arises

232
00:08:51,282 --> 00:08:52,392
The word orchestrated

233
00:08:52,802 --> 00:08:54,252
refers to how the brain

234
00:08:54,323 --> 00:08:56,722
coordinates the quantum states of hundreds of millions

235
00:08:56,723 --> 00:08:58,472
of tubulin proteins through cellular processes

236
00:08:58,762 --> 00:08:59,972
orchestrating them together

237
00:09:00,042 --> 00:09:02,772
into organized sequences of collapse

238
00:09:03,323 --> 00:09:06,012
Each collapse is one moment of now

239
00:09:06,323 --> 00:09:08,281
The flow of time you experience

240
00:09:08,282 --> 00:09:12,539
is a chain of discrete conscious events

241
00:09:12,930 --> 00:09:14,500
I want to stress something here

242
00:09:14,650 --> 00:09:18,899
Many imagine tubulin proteins are about the size of neurons

243
00:09:19,130 --> 00:09:20,419
They are not at all

244
00:09:20,650 --> 00:09:23,899
A single neuron contains roughly hundreds of millions of tubulin proteins

245
00:09:24,130 --> 00:09:27,059
The entire brain contains about one hundred billion neurons

246
00:09:27,290 --> 00:09:28,179
That means

247
00:09:28,290 --> 00:09:30,299
the total number of tubulin proteins in the brain

248
00:09:30,331 --> 00:09:33,090
is on the order of ten to the nineteenth or twentieth power

249
00:09:33,091 --> 00:09:34,020
in scale

250
00:09:34,491 --> 00:09:36,140
That is an astronomical number

251
00:09:36,410 --> 00:09:39,220
If consciousness truly involves quantum states in microtubules

252
00:09:39,290 --> 00:09:40,850
then the brains computing power

253
00:09:40,851 --> 00:09:42,699
far exceeds our previous estimates

254
00:09:42,731 --> 00:09:45,659
based on the neuron synapse model

255
00:09:46,371 --> 00:09:49,840
This may also explain something that has long puzzled neuroscientists

256
00:09:50,211 --> 00:09:51,740
Why does the human brain use only

257
00:09:51,770 --> 00:09:52,720
twenty watts of power

258
00:09:53,010 --> 00:09:54,399
about as much as a dim light bulb

259
00:09:54,691 --> 00:09:56,049
to carry out our

260
00:09:56,050 --> 00:09:57,079
entire cognitive activity

261
00:09:57,571 --> 00:10:00,179
If the brain worked only through synaptic transmission

262
00:10:00,211 --> 00:10:02,140
that efficiency would seem magical

263
00:10:02,451 --> 00:10:05,649
But if quantum processes also operate at the microtubule level

264
00:10:05,650 --> 00:10:07,502
the efficiency begins to make sense

265
00:10:07,772 --> 00:10:11,901
Since this theory was formally proposed in the mid 1990s

266
00:10:12,093 --> 00:10:16,642
mainstream science has called it extremely bold but poorly supported

267
00:10:17,213 --> 00:10:18,161
The biggest criticism is

268
00:10:18,493 --> 00:10:21,502
that the brain is too warm too wet and too noisy

269
00:10:21,693 --> 00:10:25,742
Quantum coherence could not survive there for even a few femtoseconds

270
00:10:25,892 --> 00:10:29,061
and could not support any meaningful quantum effect

271
00:10:29,252 --> 00:10:30,401
But in the last two or three years

272
00:10:30,973 --> 00:10:32,592
things have begun to change

273
00:10:33,103 --> 00:10:33,772
In 2022

274
00:10:34,142 --> 00:10:35,431
Hameroffs team

275
00:10:35,463 --> 00:10:37,711
published in Frontiers in Molecular Neuroscience

276
00:10:37,743 --> 00:10:39,551
an important paper

277
00:10:39,782 --> 00:10:40,411
They found

278
00:10:40,743 --> 00:10:42,982
that neuronal microtubules exhibit

279
00:10:42,983 --> 00:10:45,622
self similar resonant oscillations spanning

280
00:10:45,623 --> 00:10:47,431
frequencies from terahertz down to hertz

281
00:10:47,662 --> 00:10:50,472
like clocks within clocks nested layer by layer

282
00:10:50,583 --> 00:10:52,612
across terahertz gigahertz

283
00:10:52,983 --> 00:10:55,622
megahertz kilohertz and hertz

284
00:10:55,623 --> 00:10:56,951
these five scales

285
00:10:56,983 --> 00:10:58,211
repeat similar

286
00:10:58,463 --> 00:10:59,671
triplet patterns

287
00:11:00,022 --> 00:11:03,512
This fractal time crystal signature is extremely rare

288
00:11:03,623 --> 00:11:06,191
It means microtubules are not inert scaffolding

289
00:11:06,223 --> 00:11:09,232
but active multiscale oscillating systems

290
00:11:09,823 --> 00:11:10,691
More importantly

291
00:11:10,983 --> 00:11:12,102
among these oscillations

292
00:11:12,103 --> 00:11:13,831
signals in the megahertz range

293
00:11:13,863 --> 00:11:15,612
can be detected from the human scalp

294
00:11:16,022 --> 00:11:17,421
That is far beyond

295
00:11:17,422 --> 00:11:19,171
the frequency range of conventional EEG

296
00:11:19,583 --> 00:11:21,461
And general anesthetics

297
00:11:21,463 --> 00:11:23,711
precisely suppress these oscillations

298
00:11:23,902 --> 00:11:24,612
Consciousness switches off

299
00:11:24,902 --> 00:11:26,791
and microtubule oscillations switch off too

300
00:11:27,142 --> 00:11:32,173
This is among the first functional evidence directly linking microtubules to consciousness

301
00:11:32,684 --> 00:11:35,134
A 2024 study further found

302
00:11:35,324 --> 00:11:36,693
that networks of tryptophan

303
00:11:36,765 --> 00:11:38,313
within tubulin can

304
00:11:38,605 --> 00:11:40,494
under physiological conditions produce

305
00:11:40,564 --> 00:11:41,593
ultraviolet superradiance

306
00:11:42,044 --> 00:11:44,093
a collective quantum effect

307
00:11:44,324 --> 00:11:47,213
This directly challenges the brain is too warm objection

308
00:11:47,284 --> 00:11:48,404
because it proves

309
00:11:48,404 --> 00:11:50,124
protein structures themselves

310
00:11:50,125 --> 00:11:51,394
can at room temperature

311
00:11:51,644 --> 00:11:53,254
support quantum coherence

312
00:11:53,885 --> 00:11:54,994
Even more surprising

313
00:11:55,365 --> 00:11:56,534
in 2025

314
00:11:56,644 --> 00:11:57,894
a research team

315
00:11:57,924 --> 00:12:01,813
used IBMs 127 qubit processor

316
00:12:01,885 --> 00:12:06,134
to test Penroses objective reduction mechanism

317
00:12:06,725 --> 00:12:08,494
A measurement on one qubit

318
00:12:08,524 --> 00:12:10,134
was used to move

319
00:12:10,164 --> 00:12:12,313
roughly one trillionth of a kilogram

320
00:12:12,644 --> 00:12:13,293
of mass

321
00:12:13,365 --> 00:12:14,453
and then observed

322
00:12:14,524 --> 00:12:16,404
whether the quantum states evolution matched

323
00:12:16,404 --> 00:12:17,433
Penroses prediction of

324
00:12:17,684 --> 00:12:19,573
gravity induced collapse

325
00:12:19,804 --> 00:12:22,311
Preliminary results suggest that it did

326
00:12:22,782 --> 00:12:24,772
Is orchestrated objective reduction correct

327
00:12:25,263 --> 00:12:25,892
Honestly

328
00:12:26,302 --> 00:12:28,191
we cannot draw a final conclusion yet

329
00:12:28,343 --> 00:12:30,352
These experiments remain at an early stage

330
00:12:30,463 --> 00:12:33,311
and each has methodological issues open to debate

331
00:12:33,542 --> 00:12:35,372
But one thing is becoming increasingly clear

332
00:12:35,782 --> 00:12:37,632
Microtubules are more than scaffolding

333
00:12:37,743 --> 00:12:41,591
They are a severely underestimated layer of information processing in the brain

334
00:12:41,743 --> 00:12:43,662
The title of Hameroffs paper

335
00:12:43,662 --> 00:12:44,652
puts it plainly

336
00:12:45,103 --> 00:12:47,091
Neuroscience needs a new paradigm

337
00:12:47,503 --> 00:12:51,191
We cannot understand consciousness by looking only at synapses and neuronal firing

338
00:12:51,302 --> 00:12:53,902
We must look deeper inside the cell

339
00:12:53,902 --> 00:12:55,802
down to the level of microtubules

340
00:12:56,234 --> 00:12:56,423
Now

341
00:12:56,793 --> 00:12:58,923
let me bring the focus back to AI

342
00:12:59,234 --> 00:13:00,883
If Penrose is right

343
00:13:00,913 --> 00:13:03,023
what does that mean for the debate over AI consciousness

344
00:13:03,714 --> 00:13:06,342
First note one crucial subtlety

345
00:13:06,913 --> 00:13:09,722
Penrose denies that computation produces consciousness

346
00:13:09,793 --> 00:13:12,082
not that nonbiological systems can produce it

347
00:13:12,394 --> 00:13:16,243
His dividing line is not between carbon and silicon

348
00:13:16,313 --> 00:13:18,763
but between computation and understanding

349
00:13:19,193 --> 00:13:20,702
AI built on Turing machine architecture

350
00:13:21,033 --> 00:13:22,753
which includes all of todays

351
00:13:22,754 --> 00:13:23,623
large language models

352
00:13:24,033 --> 00:13:25,222
and all neural networks

353
00:13:25,594 --> 00:13:26,673
cannot in principle

354
00:13:26,673 --> 00:13:28,243
produce consciousness

355
00:13:28,953 --> 00:13:32,322
But if one day we build an entirely new physical system

356
00:13:32,433 --> 00:13:35,283
a device capable of exploiting quantum gravity effects

357
00:13:35,313 --> 00:13:37,802
it could theoretically become conscious

358
00:13:37,833 --> 00:13:40,214
because it would no longer be a Turing machine

359
00:13:40,565 --> 00:13:42,934
The title of a 2025 paper

360
00:13:43,004 --> 00:13:44,234
states the point directly

361
00:13:44,724 --> 00:13:46,484
Why current AI architectures

362
00:13:46,484 --> 00:13:47,633
cannot be conscious

363
00:13:48,404 --> 00:13:48,954
The authors use

364
00:13:49,364 --> 00:13:51,533
the framework of orchestrated objective reduction

365
00:13:51,685 --> 00:13:53,964
to analyze classical deep neural networks

366
00:13:53,965 --> 00:13:54,753
and their fundamental flaws

367
00:13:55,285 --> 00:13:58,373
They are several orders of magnitude less energy efficient than the human brain

368
00:13:58,524 --> 00:14:02,133
They lack an integration and binding mechanism for unified subjective experience

369
00:14:02,205 --> 00:14:06,734
And their current architecture offers no known path to consciousness

370
00:14:06,965 --> 00:14:11,053
This is not about needing a few more chip generations or more training data

371
00:14:11,165 --> 00:14:15,001
It says this road itself does not lead to consciousness

372
00:14:15,431 --> 00:14:16,100
Think about it

373
00:14:16,512 --> 00:14:17,870
If this conclusion holds

374
00:14:17,872 --> 00:14:18,501
what would it mean

375
00:14:18,992 --> 00:14:21,840
It would let us release half our anxiety about AI

376
00:14:22,232 --> 00:14:23,820
AI will grow ever smarter

377
00:14:24,032 --> 00:14:25,080
ever more capable

378
00:14:25,152 --> 00:14:27,240
and ever more like a conscious being

379
00:14:27,471 --> 00:14:31,281
But an uncrossable gulf separates seeming from being

380
00:14:31,671 --> 00:14:34,521
A mirror can perfectly reflect your smile

381
00:14:34,671 --> 00:14:36,400
but the mirror feels no joy

382
00:14:36,871 --> 00:14:38,801
This brings me to a deeper question

383
00:14:39,152 --> 00:14:40,270
On this channel

384
00:14:40,272 --> 00:14:41,341
we have discussed predictive coding

385
00:14:41,591 --> 00:14:43,061
the default mode network

386
00:14:43,351 --> 00:14:45,161
and the eight consciousnesses of Yogācāra

387
00:14:45,272 --> 00:14:49,480
Each framework approaches the nature of consciousness from a different angle

388
00:14:49,912 --> 00:14:51,990
What makes Penrose unique is

389
00:14:51,991 --> 00:14:53,740
that he does not say what consciousness resembles

390
00:14:53,991 --> 00:14:55,860
He says what consciousness is not

391
00:14:56,552 --> 00:14:58,041
Consciousness is not computation

392
00:14:58,512 --> 00:14:59,880
It is not information processing

393
00:15:00,272 --> 00:15:03,256
or a functional mapping from input to output

394
00:15:03,727 --> 00:15:04,636
Then what is consciousness

395
00:15:05,048 --> 00:15:07,817
Perhaps precisely because it is unlike anything we know

396
00:15:08,127 --> 00:15:09,297
it remains what

397
00:15:09,367 --> 00:15:11,416
philosopher David Chalmers

398
00:15:11,448 --> 00:15:12,236
called the

399
00:15:12,607 --> 00:15:13,156
hard problem

400
00:15:13,967 --> 00:15:17,156
Why do certain physical processes come with subjective experience

401
00:15:17,688 --> 00:15:19,616
Why when you see red

402
00:15:19,767 --> 00:15:21,116
is there a feeling of redness

403
00:15:21,607 --> 00:15:22,876
Where does that feeling come from

404
00:15:23,408 --> 00:15:25,577
Penrose did not finally answer this question

405
00:15:25,887 --> 00:15:27,197
He only pointed in a direction

406
00:15:27,688 --> 00:15:29,616
The answer is not in computer science

407
00:15:29,648 --> 00:15:31,177
nor in classical physics

408
00:15:31,328 --> 00:15:32,957
It may lie in quantum gravity

409
00:15:33,288 --> 00:15:35,166
in the deepest layer of physics

410
00:15:35,168 --> 00:15:36,931
that we do not yet fully understand

411
00:15:37,241 --> 00:15:38,880
A ninety five year old man

412
00:15:38,882 --> 00:15:42,311
has spent nearly forty years telling us the same thing

413
00:15:42,721 --> 00:15:44,411
You are looking in the wrong place

414
00:15:44,642 --> 00:15:46,411
Consciousness is not in the code

415
00:15:46,762 --> 00:15:48,953
It is in the structure of spacetime

416
00:15:49,263 --> 00:15:50,262
Personally I think

417
00:15:50,263 --> 00:15:54,793
Penroses deepest insight is not even orchestrated objective reduction itself

418
00:15:54,863 --> 00:15:57,593
but his obsession with understanding

419
00:15:58,064 --> 00:15:59,412
He keeps asking

420
00:15:59,704 --> 00:16:02,862
when a mind truly understands a proposition

421
00:16:02,863 --> 00:16:04,772
what exactly happens in the universe

422
00:16:05,184 --> 00:16:07,193
If that event is not computation

423
00:16:07,343 --> 00:16:09,992
then it is some new kind of physics

424
00:16:10,024 --> 00:16:12,793
a physics we have not yet discovered

425
00:16:13,263 --> 00:16:14,183
In other words

426
00:16:14,184 --> 00:16:16,752
consciousness is not merely a biological problem

427
00:16:16,784 --> 00:16:18,793
nor merely a philosophical one

428
00:16:18,983 --> 00:16:20,793
It is a problem in physics

429
00:16:21,824 --> 00:16:23,583
and perhaps

430
00:16:23,583 --> 00:16:24,573
the hardest one in physics

431
00:16:25,064 --> 00:16:26,112
because solving it

432
00:16:26,223 --> 00:16:27,313
requires us to understand

433
00:16:27,343 --> 00:16:30,152
the unification of quantum mechanics and general relativity

434
00:16:30,263 --> 00:16:33,898
the theory of quantum gravity that no one has yet completed

435
00:16:34,842 --> 00:16:36,057
Do you think he is right

436
00:16:36,661 --> 00:16:39,342
Do you think AI could ever truly understand anything

437
00:16:40,025 --> 00:16:40,528
Or

438
00:16:40,856 --> 00:16:42,082
however sophisticated it becomes

439
00:16:42,082 --> 00:16:44,090
will it always remain a Chinese room

440
00:16:44,575 --> 00:16:45,692
with a person inside

441
00:16:45,724 --> 00:16:47,433
following rules and consulting a dictionary

442
00:16:47,543 --> 00:16:49,649
to write a perfect reply in Chinese

443
00:16:49,720 --> 00:16:51,332
without understanding a single word

444
00:16:52,014 --> 00:16:53,963
Tell me what you think in the comments

445
00:16:54,072 --> 00:16:54,814
I am Leo Wang

446
00:16:55,101 --> 00:16:56,298
See you next time