1
00:00:00,163 -->
00:00:06,356
Ejaaz:
What's up guys, I just got to Japan, I am the most jet-lagged I have ever felt
2
00:00:06,513 -->
00:00:12,196
Ejaaz:
in my life, and I'm sat on my hotel room floor because I need to record this video.
3
00:00:12,423 -->
00:00:17,032
Ejaaz:
OpenAI just dropped one of their biggest releases and it's not an AI model,
4
00:00:17,160 -->
00:00:23,255
Ejaaz:
it's the announcement that they solved around 90 of the toughest problems in mathematics ever.
5
00:00:23,377 -->
00:00:27,638
Ejaaz:
And I know what you're thinking, you're wondering why on earth should I or you
6
00:00:27,638 -->
00:00:31,528
Ejaaz:
care about maths, That's something I haven't touched since high school or college.
7
00:00:31,528 -->
00:00:35,378
Ejaaz:
And the point of this video isn't that they solve these math problems.
8
00:00:35,378 -->
00:00:40,958
Ejaaz:
This will slowly and very quickly roll over into every particular field that
9
00:00:40,958 -->
00:00:44,489
Ejaaz:
you could think of that you probably studied to do your job today.
10
00:00:44,629 -->
00:00:48,028
Ejaaz:
It'll roll over into chemistry, into engineering, into physics,
11
00:00:48,028 -->
00:00:54,219
Ejaaz:
into whatever requires reasoning. And it's all happening much quicker than anyone can even imagine.
12
00:00:54,387 -->
00:00:57,408
Ejaaz:
So in this episode, I'm going to get into what exactly it's solved.
13
00:00:57,408 -->
00:01:01,868
Ejaaz:
But also more importantly, what happens next and what this means for you.
14
00:01:02,385 -->
00:01:05,949
Ejaaz:
All of that and more on today's episode of In The Loop, brought to you by Qualcomm.
15
00:01:09,350 -->
00:01:15,424
Ejaaz:
Let me just set some quick context. Three years ago, AI could not do math at
16
00:01:15,424 -->
00:01:18,191
Ejaaz:
all. Two plus two didn't equal four.
17
00:01:18,475 -->
00:01:22,074
Ejaaz:
It equaled 11. Strawberry had five R's in it.
18
00:01:22,405 -->
00:01:26,734
Ejaaz:
LLMs just weren't smart enough. And I remember commentary from Terence Tao,
19
00:01:27,034 -->
00:01:30,344
Ejaaz:
who is one of the leading mathematicians in the entire world.
20
00:01:30,344 -->
00:01:37,440
Ejaaz:
And he said, this is a reasonable attempt by AI, but I do not see how reasoning
21
00:01:37,672 -->
00:01:40,294
Ejaaz:
can eventually lead to solving maths.
22
00:01:40,294 -->
00:01:44,704
Ejaaz:
Reasoning is basically how you or I or any other human reasons about a problem.
23
00:01:44,704 -->
00:01:47,976
Ejaaz:
We don't just take it for face value, we apply different contexts to it.
24
00:01:48,295 -->
00:01:52,014
Ejaaz:
Back then, AIs could not do that, but they were at the start of their journey
25
00:01:52,014 -->
00:01:53,659
Ejaaz:
of figuring out this whole thing.
26
00:01:54,031 -->
00:01:59,624
Ejaaz:
Now, fast forward to six weeks ago, and OpenAI, who are back then and still
27
00:01:59,624 -->
00:02:03,454
Ejaaz:
are the experts in reasoning for training AI models, took that to its fullest
28
00:02:03,454 -->
00:02:07,661
Ejaaz:
extent when they announced that they had an internal model that was only 11 days old,
29
00:02:07,973 -->
00:02:11,024
Ejaaz:
had solved one of the Millennium Prize problems.
30
00:02:11,024 -->
00:02:14,964
Ejaaz:
Now, Millennium Prize problems, there's seven of them, and they're all in the
31
00:02:14,964 -->
00:02:17,052
Ejaaz:
field of mathematics. If anyone
32
00:02:17,180 -->
00:02:19,994
Ejaaz:
can go out there and solve any of these problems, you get $1 million.
33
00:02:20,709 -->
00:02:24,584
Ejaaz:
Now, for 200 years, these problems have existed. People have made progress,
34
00:02:24,584 -->
00:02:28,377
Ejaaz:
but the truth is, there are just too few mathematicians and experts in the world
35
00:02:28,494 -->
00:02:32,487
Ejaaz:
to try and figure this out, we are limited by human cognition and capacity.
36
00:02:32,725 -->
00:02:38,252
Ejaaz:
OpenAI came along with zero experience in maths, with zero help from math researchers,
37
00:02:38,496 -->
00:02:42,414
Ejaaz:
and they just decided, hmm, I wonder if I just use this powerful model that's
38
00:02:42,814 -->
00:02:46,083
Ejaaz:
more powerful than GPT-6 Astra and just let it go.
39
00:02:46,640 -->
00:02:49,952
Ejaaz:
Until it was able to solve the problems. There was infinite compute,
40
00:02:49,952 -->
00:02:51,801
Ejaaz:
infinite money, just get it done.
41
00:02:52,132 -->
00:02:56,762
Ejaaz:
And it solved it in 88 hours, 10,000 agents. It was one of the most costly experiments
42
00:02:56,762 -->
00:02:59,752
Ejaaz:
that they've done, probably that anyone's ever done for a research problem.
43
00:02:59,752 -->
00:03:01,022
Ejaaz:
It cost over $10 million.
44
00:03:01,642 -->
00:03:05,442
Ejaaz:
And there was a few interesting things about that. One, this was the first instance
45
00:03:05,442 -->
00:03:10,654
Ejaaz:
that an AI model had demonstrated that it wasn't just mimicking human intelligence.
46
00:03:10,846 -->
00:03:14,152
Ejaaz:
Up until then, people had been thinking, okay, AI models are smart,
47
00:03:14,152 -->
00:03:16,602
Ejaaz:
but they're just mimicking what it's learned on its data set.
48
00:03:16,602 -->
00:03:18,602
Ejaaz:
It's a next token predictor. What
49
00:03:18,602 -->
00:03:21,736
Ejaaz:
it means is it's trained on all human knowledge that we put out there.
50
00:03:21,951 -->
00:03:26,676
Ejaaz:
It's limited by the cap or the ceiling of how intelligent we already are.
51
00:03:26,908 -->
00:03:31,117
Ejaaz:
It solving Navier-Stokes was the first demonstration that it could go above
52
00:03:31,343 -->
00:03:35,203
Ejaaz:
that level, that it knew things that we as humans couldn't understand.
53
00:03:35,604 -->
00:03:39,812
Ejaaz:
And when Navier-Stokes got resolved, people weren't as excited as you might
54
00:03:39,812 -->
00:03:42,436
Ejaaz:
expect them to be, particularly mathematicians.
55
00:03:42,587 -->
00:03:46,292
Ejaaz:
They looked at this result and they weren't happy. They were perplexed.
56
00:03:46,292 -->
00:03:50,302
Ejaaz:
They were perplexed because they couldn't make sense of the solution.
57
00:03:50,546 -->
00:03:56,042
Ejaaz:
Now, fast forward to today and we have news that their new internal model has
58
00:03:56,042 -->
00:04:03,340
Ejaaz:
solved not just one, not two, but 90 out of 500 of the world's most hardest open math problems.
59
00:04:03,854 -->
00:04:08,545
Ejaaz:
That's 20% of all the hardest math problems that have existed for decades at this point.
60
00:04:08,882 -->
00:04:11,976
Ejaaz:
But there's a few differences between this and Navier-Stokes.
61
00:04:12,452 -->
00:04:17,903
Ejaaz:
It solved it in a much quicker way. On average, each problem was solved in three hours.
62
00:04:18,251 -->
00:04:23,533
Ejaaz:
And it also got it done at a much cheaper cost, costing only tens of thousands of dollars.
63
00:04:23,533 -->
00:04:27,813
Ejaaz:
So you can imagine or extrapolate in the future that these models will get faster
64
00:04:27,813 -->
00:04:32,853
Ejaaz:
or quicker and even cheaper to the point where the cost of solving problems
65
00:04:32,853 -->
00:04:37,493
Ejaaz:
that humans have spent decades trying to figure out becomes undeniable.
66
00:04:37,493 -->
00:04:39,892
Ejaaz:
You should just use AI and you don't need humans anymore.
67
00:04:40,130 -->
00:04:44,233
Ejaaz:
But I was critical. I saw this announcement from OpenAI here and I saw that
68
00:04:44,233 -->
00:04:47,703
Ejaaz:
they had dumped a massive GitHub, which was interesting. They open sourced the
69
00:04:47,703 -->
00:04:51,723
Ejaaz:
entire thing. So all the formal verification and proofs were laid out for anyone
70
00:04:51,723 -->
00:04:52,762
Ejaaz:
and everyone to critique.
71
00:04:52,953 -->
00:04:55,733
Ejaaz:
And it got me thinking, okay, it's solved a bunch of problems,
72
00:04:55,733 -->
00:04:58,613
Ejaaz:
but how is this useful for you or I?
73
00:04:58,613 -->
00:05:02,693
Ejaaz:
Like, how should I think about this being applied to the real world?
74
00:05:02,693 -->
00:05:06,373
Ejaaz:
And our friend Ole actually had a great breakdown in his tweet.
75
00:05:06,373 -->
00:05:08,973
Ejaaz:
He ran through a few of them, but being very high level here,
76
00:05:08,973 -->
00:05:12,631
Ejaaz:
because I'm not a mathematician, one of these problems is called Vlasov Maxwell.
77
00:05:12,869 -->
00:05:18,803
Ejaaz:
And if it's truly solved, this research can then result in a way to create virtually
78
00:05:19,183 -->
00:05:21,628
Ejaaz:
limitless clean energy to power civilization.
79
00:05:22,035 -->
00:05:26,656
Ejaaz:
Another example is the inverse elasticity problem that, if it's solved,
80
00:05:26,812 -->
00:05:32,043
Ejaaz:
can scan and map tissue stiffness, basically allowing you to evaluate and medicate
81
00:05:32,043 -->
00:05:34,579
Ejaaz:
and help individuals in the medical field.
82
00:05:34,881 -->
00:05:37,793
Ejaaz:
There's also another one called matrix multiplication. If you've heard that
83
00:05:37,793 -->
00:05:40,971
Ejaaz:
term before, it's because we've probably talked about it on this show.
84
00:05:41,278 -->
00:05:45,903
Ejaaz:
It's the foundation of how AI models run and how they run and compute all the
85
00:05:45,903 -->
00:05:50,043
Ejaaz:
important data to give you an amazing result. It's also how GPUs and any other
86
00:05:50,043 -->
00:05:54,473
Ejaaz:
modern graphics card processor works for videos, gaming, or whatever that might be.
87
00:05:54,670 -->
00:05:57,254
Ejaaz:
It's a very important fundamental mechanism of computing.
88
00:05:57,938 -->
00:06:02,018
Ejaaz:
It solved one of the most incredibly hard problems that that industry was facing.
89
00:06:02,273 -->
00:06:06,390
Ejaaz:
And if it's able to figure out, guess what? You get cheaper AI and bigger scientific
90
00:06:06,390 -->
00:06:10,790
Ejaaz:
simulations that you could use to map or design proteins that could result or
91
00:06:10,790 -->
00:06:13,227
Ejaaz:
end in the cure for cancer, for example.
92
00:06:13,395 -->
00:06:18,835
Ejaaz:
So the idea here is this isn't just a list of random niche, arbitrary mathematical problems.
93
00:06:19,049 -->
00:06:22,920
Ejaaz:
These are things that have stumped scientists, builders, engineers,
94
00:06:22,920 -->
00:06:27,310
Ejaaz:
and creators in the real world from innovating. It has prevented humans from
95
00:06:27,310 -->
00:06:31,260
Ejaaz:
innovating and building something that is marvelous and novel that would help
96
00:06:31,260 -->
00:06:33,167
Ejaaz:
world and humanity as it is today.
97
00:06:33,440 -->
00:06:37,360
Ejaaz:
So when you frame things this way, you start to realize that humanity's progress
98
00:06:37,360 -->
00:06:41,540
Ejaaz:
has always been dependent on the geniuses that they have in its civilization,
99
00:06:41,540 -->
00:06:42,844
Ejaaz:
in its society at the time.
100
00:06:43,152 -->
00:06:46,710
Ejaaz:
Who is smart enough to fix or figure out some of the hardest problems in the
101
00:06:46,710 -->
00:06:49,560
Ejaaz:
world so that we can progress as a human race?
102
00:06:49,740 -->
00:06:53,891
Ejaaz:
I loved this tweet from Will Depew, where he basically asked
103
00:06:54,187 -->
00:06:59,610
Ejaaz:
Fable 5.1 and GPT-6 Pro to rank all the discoveries that OpenAI's internal model
104
00:06:59,610 -->
00:07:02,910
Ejaaz:
had made over the last three years, just to see how important these discoveries
105
00:07:02,910 -->
00:07:05,530
Ejaaz:
actually were relative to that perspective that I just laid out,
106
00:07:05,530 -->
00:07:08,303
Ejaaz:
which is, you know, solving, you know, human innovation.
107
00:07:08,449 -->
00:07:14,213
Ejaaz:
And 81% of them were released on the day that OpenAI had solved these problems,
108
00:07:14,410 -->
00:07:17,609
Ejaaz:
meaning that a lot of these mathematical problems were fundamentally crucial.
109
00:07:17,754 -->
00:07:19,408
Ejaaz:
Just think about that for a second.
110
00:07:19,733 -->
00:07:25,280
Ejaaz:
The next Albert Einstein, the next Galileo isn't going to be a human.
111
00:07:25,280 -->
00:07:26,966
Ejaaz:
It's going to be a computer.
112
00:07:27,251 -->
00:07:30,148
Ejaaz:
It's going to be an AI because it's cheaper.
113
00:07:30,444 -->
00:07:35,080
Ejaaz:
It can work 24-7. And most importantly, it's now demonstrated that it's smarter
114
00:07:35,080 -->
00:07:38,490
Ejaaz:
than us. Framing it the way that Alex has here makes a lot of sense.
115
00:07:38,490 -->
00:07:43,122
Ejaaz:
He goes quasi RH. He's referencing one of the problems that it solved. Are you kidding me?
116
00:07:43,569 -->
00:07:47,724
Ejaaz:
If a human did this, it would be an instant Fields Medal. No questions asked.
117
00:07:47,724 -->
00:07:51,655
Ejaaz:
A Fields Medal is one of the highest honors you can earn in the mathematics field.
118
00:07:51,858 -->
00:07:55,954
Ejaaz:
PhDs, mathematicians make it one of their life goals to try and earn a Fields
119
00:07:55,954 -->
00:07:59,931
Ejaaz:
Medal. It is incredibly hard and a rare thing to do. It's very prestigious.
120
00:08:00,105 -->
00:08:06,316
Ejaaz:
And now the cost and effectiveness of that has been completely taken from mathematicians.
121
00:08:06,659 -->
00:08:11,084
Ejaaz:
And understandably, their reaction to this has mostly been mathematicians being
122
00:08:11,084 -->
00:08:14,176
Ejaaz:
very uncomfortable with the way open AI has approached this.
123
00:08:14,380 -->
00:08:17,410
Ejaaz:
Actually, it started off when they solved Navier Stokes six weeks ago.
124
00:08:17,590 -->
00:08:20,714
Ejaaz:
You might have thought mathematicians would have reacted positively to the news,
125
00:08:20,714 -->
00:08:23,186
Ejaaz:
but they really didn't. They were actually offended.
126
00:08:23,470 -->
00:08:27,684
Ejaaz:
They thought that the act of open AI using these LLMs to solve mathematical
127
00:08:27,684 -->
00:08:31,940
Ejaaz:
research, the act or thought of a computer doing their job that they had spent
128
00:08:32,154 -->
00:08:34,343
Ejaaz:
three to four years studying at a master's
129
00:08:34,656 -->
00:08:37,675
Ejaaz:
and bachelor's degree level and then and doing a PhD on top of this to,
130
00:08:37,965 -->
00:08:42,122
Ejaaz:
you know, their sole goal as a PhD student is to create net new information,
131
00:08:42,423 -->
00:08:46,204
Ejaaz:
can now be automated, that their prestige and their status that they've dedicated
132
00:08:46,204 -->
00:08:48,037
Ejaaz:
their life to has suddenly been taken away.
133
00:08:48,217 -->
00:08:51,787
Ejaaz:
And these are people that are performing at the top level of their field,
134
00:08:52,135 -->
00:08:53,581
Ejaaz:
suddenly automated overnight.
135
00:08:54,247 -->
00:08:57,533
Ejaaz:
Was a very uncomfortable feeling for them. And I totally get it.
136
00:08:57,974 -->
00:09:01,300
Ejaaz:
But at the same time, I'm grappling with this thinking, okay, well.
137
00:09:01,927 -->
00:09:05,401
Ejaaz:
The whole point about mathematicians doing their thing, chemists doing their
138
00:09:05,401 -->
00:09:09,311
Ejaaz:
things, Nobel laureates doing their thing, is so that we can push humanity forwards,
139
00:09:09,311 -->
00:09:11,701
Ejaaz:
that we can have a major invention or discovery
140
00:09:12,021 -->
00:09:15,424
Ejaaz:
that could make a lot of money, generate a lot of benefit for humans,
141
00:09:15,650 -->
00:09:17,361
Ejaaz:
and really help push us along.
142
00:09:17,361 -->
00:09:20,724
Ejaaz:
So why shouldn't we be happy about this AI model doing this for us?
143
00:09:20,881 -->
00:09:24,951
Ejaaz:
And I kind of tried to lay out my point in this tweet, where math PhDs basically
144
00:09:24,951 -->
00:09:30,871
Ejaaz:
dedicate their lives to trying to contribute in as small or as large a way as they can.
145
00:09:30,871 -->
00:09:34,181
Ejaaz:
But often, because we're humans, because we're limited in our ingenuity,
146
00:09:34,181 -->
00:09:37,331
Ejaaz:
because it's so hard of a field to figure out all these different problems,
147
00:09:37,331 -->
00:09:41,239
Ejaaz:
because our mental cognition and our context windows compared to AI LLMs
148
00:09:41,517 -->
00:09:45,581
Ejaaz:
are tiny, small, we need sleep, we're moody, we're irrational,
149
00:09:45,581 -->
00:09:47,137
Ejaaz:
we're emotional beings.
150
00:09:47,462 -->
00:09:52,210
Ejaaz:
Now we have a computer that can just work in the perfectly optimized way that we need to.
151
00:09:52,412 -->
00:09:56,795
Ejaaz:
And the thing that was holding back AI before, prior to this announcement,
152
00:09:57,190 -->
00:10:01,433
Ejaaz:
was that it was just trained on existing human data. It was dumbed down and
153
00:10:01,706 -->
00:10:05,438
Ejaaz:
held back by the limits of human capacity by even the smartest people.
154
00:10:05,758 -->
00:10:10,561
Ejaaz:
And now we see AI surpass that ingenuity. And that is a genuinely uncomfortable
155
00:10:10,561 -->
00:10:11,865
Ejaaz:
feeling, understandably.
156
00:10:11,985 -->
00:10:15,281
Ejaaz:
But you have to understand, this story isn't actually about the mathematicians.
157
00:10:15,281 -->
00:10:19,401
Ejaaz:
The way that they're feeling today is the way that experts across any single
158
00:10:19,401 -->
00:10:22,267
Ejaaz:
field are going to feel in the very near future.
159
00:10:22,430 -->
00:10:25,981
Ejaaz:
So I want to get into what that's going to look like and the major implications
160
00:10:25,981 -->
00:10:28,136
Ejaaz:
of that on today's show brought to you by Qualcomm.
161
00:10:31,497 -->
00:10:34,422
Ejaaz:
So the first thing we have to understand about this, aside from maths,
162
00:10:34,422 -->
00:10:40,504
Ejaaz:
is if reasoning is solving mathematics, which is one of the hardest fields, most complex
163
00:10:40,788 -->
00:10:44,828
Ejaaz:
to understand, learn, higher barrier of adoption to even enter the field,
164
00:10:44,950 -->
00:10:47,652
Ejaaz:
then it seems logical that this can apply to
165
00:10:48,012 -->
00:10:51,591
Ejaaz:
any other field that requires intense reasoning.
166
00:10:51,777 -->
00:10:54,192
Ejaaz:
The first thing that pops into mind that is closest to maths,
167
00:10:54,192 -->
00:10:56,322
Ejaaz:
it's kind of like the sister of maths, is physics.
168
00:10:56,322 -->
00:11:00,012
Ejaaz:
And then you kind of think of chemistry, all the numbers, equations that are
169
00:11:00,032 -->
00:11:03,492
Ejaaz:
required to create chemical compounds that could end up being a pharmaceutical
170
00:11:03,492 -->
00:11:07,927
Ejaaz:
product that cures a major ailment or illness. That is a big part of chemistry.
171
00:11:08,141 -->
00:11:10,902
Ejaaz:
And then it kind of rolls into systems biology and different aspects,
172
00:11:10,902 -->
00:11:12,202
Ejaaz:
but it's still very mathematical.
173
00:11:12,202 -->
00:11:16,072
Ejaaz:
And so those folks will probably push back and they'll say, but you do realize
174
00:11:16,072 -->
00:11:20,767
Ejaaz:
you need human intuition, you need judgment. And any feel that requires judgment
175
00:11:20,889 -->
00:11:24,784
Ejaaz:
is something that AI LLMs can't solve with just reasoning.
176
00:11:24,964 -->
00:11:28,332
Ejaaz:
And then you realize, well, actually, that's not strictly true,
177
00:11:28,332 -->
00:11:32,662
Ejaaz:
because you can technically take that logic and tell an LLM,
178
00:11:32,662 -->
00:11:36,832
Ejaaz:
hey, like, can you also try and reason and figure out human intuition and apply
179
00:11:36,832 -->
00:11:38,081
Ejaaz:
that? And it could end up doing that.
180
00:11:38,267 -->
00:11:42,872
Ejaaz:
But it's also about what OpenAI is not saying. And this is arguably the biggest
181
00:11:42,872 -->
00:11:48,796
Ejaaz:
point of the entire episode. This is post by Justin Drake. He is very renowned in cryptography.
182
00:11:49,174 -->
00:11:54,741
Ejaaz:
And he says, today I call upon the blockchain industry to calmly begin planning for bunker mode.
183
00:11:54,892 -->
00:11:59,792
Ejaaz:
What he goes on to explain is bunker mode is, we need to shift off of blockchain
184
00:11:59,792 -->
00:12:04,502
Ejaaz:
addresses because I think cryptography has been broken, which is weird because
185
00:12:04,502 -->
00:12:06,693
Ejaaz:
there's been no proof that an AI has done this.
186
00:12:07,030 -->
00:12:10,083
Ejaaz:
But he referred to OpenAI's announcement of solving all these mathematical problems.
187
00:12:10,432 -->
00:12:13,555
Ejaaz:
And he said the biggest problem about this is what they haven't said.
188
00:12:14,002 -->
00:12:19,402
Ejaaz:
He said, what is the coincidence of OpenAI's internal model being so smart to
189
00:12:19,402 -->
00:12:22,972
Ejaaz:
solve these niche problems that have existed for hundreds of years and somehow
190
00:12:22,972 -->
00:12:27,863
Ejaaz:
not been able to break the fundamentals of cryptography that has secured an industry
191
00:12:28,136 -->
00:12:31,195
Ejaaz:
worth trillions and trillions of dollars?
192
00:12:31,665 -->
00:12:35,764
Ejaaz:
And so the point he makes is, if an internal lab such as OpenAI has been able
193
00:12:35,764 -->
00:12:38,440
Ejaaz:
to do that, they wouldn't tell the public.
194
00:12:38,892 -->
00:12:42,624
Ejaaz:
And then I started to realize, oh wait, this model that OpenAI just used to
195
00:12:42,624 -->
00:12:45,609
Ejaaz:
do all these cool stuff isn't even accessible to you or I.
196
00:12:45,998 -->
00:12:49,899
Ejaaz:
It's accessible to OpenAI and they've had it for a while now, for months, for weeks.
197
00:12:50,322 -->
00:12:54,450
Ejaaz:
And look how much damage that they've done in that very little time.
198
00:12:54,601 -->
00:12:58,584
Ejaaz:
And then you start to think, well, if I were OpenAI and I had this model that
199
00:12:58,584 -->
00:13:02,034
Ejaaz:
could solve all these hard problems, I wonder what would happen if I pointed
200
00:13:02,034 -->
00:13:07,124
Ejaaz:
it at a problem that would give me a huge commercial advantage against my competitor
201
00:13:07,124 -->
00:13:09,427
Ejaaz:
Anthropic or Meta or Google.
202
00:13:09,885 -->
00:13:13,494
Ejaaz:
That's something that I would keep internally. And so we're starting to realize
203
00:13:13,494 -->
00:13:18,854
Ejaaz:
this gap being created where the model labs are forming these models that are
204
00:13:18,854 -->
00:13:22,844
Ejaaz:
very capable of self-improving, doing it in a very cost-effective manner.
205
00:13:22,844 -->
00:13:27,144
Ejaaz:
So the cost of human intelligence is rapidly declining, but it's not at the
206
00:13:27,144 -->
00:13:32,134
Ejaaz:
benefit of humans, it's at the benefit of these models, which are owned by these private companies.
207
00:13:32,854 -->
00:13:39,483
Ejaaz:
Have the incentive to not publicly release this for their own commercial or specific advantage.
208
00:13:39,483 -->
00:13:45,027
Ejaaz:
But I guess to challenge myself, on the flip side, if these models were to become publicly accessible,
209
00:13:45,358 -->
00:13:50,243
Ejaaz:
it creates another interesting challenge, which is anyone and everyone with
210
00:13:50,243 -->
00:13:55,569
Ejaaz:
access to this model who can think of a smart enough prompt can end up being as smart
211
00:13:55,859 -->
00:13:59,953
Ejaaz:
as Einstein, or at least achieving some of the greatest things that these former
212
00:13:59,953 -->
00:14:03,103
Ejaaz:
researchers have built over history in humanity.
213
00:14:03,103 -->
00:14:08,233
Ejaaz:
The future of human innovation is not really human at all.
214
00:14:08,233 -->
00:14:13,493
Ejaaz:
It's human and AI assisted. There is no future where we do not use this tool
215
00:14:13,493 -->
00:14:17,503
Ejaaz:
because it has now become cheap enough and more capable than us to be able to
216
00:14:17,503 -->
00:14:20,084
Ejaaz:
do this. So we need to work in confluence with this.
217
00:14:20,269 -->
00:14:24,337
Ejaaz:
But there is one major issue, which is humans, particularly the experts,
218
00:14:24,663 -->
00:14:30,589
Ejaaz:
have flipped from being the experts to becoming a complete bottleneck. Let me explain.
219
00:14:31,042 -->
00:14:35,293
Ejaaz:
With the math problems that have just been solved, there are 90 of them that
220
00:14:35,293 -->
00:14:37,699
Ejaaz:
have what's known as a lean resolution.
221
00:14:37,891 -->
00:14:42,253
Ejaaz:
So this is a formula that's been tested in a computer and it's come out with
222
00:14:42,253 -->
00:14:44,915
Ejaaz:
a big green check, 100 out of 100.
223
00:14:45,536 -->
00:14:48,369
Ejaaz:
But according to research practices today.
224
00:14:48,990 -->
00:14:53,044
Ejaaz:
You need a human to actually verify that it's true. We can't trust the computer.
225
00:14:53,364 -->
00:14:57,924
Ejaaz:
We need to trust the human. Problem is the humans, the smartest mathematicians,
226
00:14:58,144 -->
00:15:02,127
Ejaaz:
don't understand the formulas, don't understand the solutions that's in front of them.
227
00:15:02,388 -->
00:15:04,541
Ejaaz:
So it's going to take a lot of time.
228
00:15:04,832 -->
00:15:08,007
Ejaaz:
That time can cost us major innovation and leaps in the future.
229
00:15:08,367 -->
00:15:10,564
Ejaaz:
So I'm sitting here and I'm thinking about this and I'm thinking,
230
00:15:10,564 -->
00:15:14,340
Ejaaz:
hmm, what could help humans do this faster? AI.
231
00:15:14,961 -->
00:15:18,085
Ejaaz:
We should just use AI to check the AI solution.
232
00:15:18,404 -->
00:15:20,214
Ejaaz:
And then it becomes this weird thing where it's like, okay, well,
233
00:15:20,214 -->
00:15:23,842
Ejaaz:
why do we need the human in the loop anyway? Why do we need humans at all?
234
00:15:24,155 -->
00:15:28,131
Ejaaz:
Now, that is a very doomer and bearish take and I just want to be very upfront and clear.
235
00:15:28,450 -->
00:15:32,394
Ejaaz:
That's not actually where I think things will settle. I think to my earlier
236
00:15:32,394 -->
00:15:37,894
Ejaaz:
point, human intuition is a very important thing that is required to make the right judgment calls.
237
00:15:37,894 -->
00:15:41,204
Ejaaz:
I think there are plenty of fields out there that require this and it's not
238
00:15:41,204 -->
00:15:45,414
Ejaaz:
something that I think an AI model can intuitively figure out anytime soon.
239
00:15:45,414 -->
00:15:47,119
Ejaaz:
However, it's something that we should be wary of.
240
00:15:47,363 -->
00:15:50,114
Ejaaz:
And most importantly, that researchers of any other field should be wary of
241
00:15:50,114 -->
00:15:52,654
Ejaaz:
because it's going to slowly roll over.
242
00:15:52,654 -->
00:15:57,374
Ejaaz:
This is like a tidal tsunami that's going to sweep in very unexpectedly.
243
00:15:57,374 -->
00:16:00,882
Ejaaz:
And if you're wondering where I'm betting the next field is going to be,
244
00:16:01,062 -->
00:16:07,044
Ejaaz:
I think it's going to be biology, particularly focused on medical services or
245
00:16:07,044 -->
00:16:09,497
Ejaaz:
medical cures for the pharmaceutical industry.
246
00:16:10,020 -->
00:16:14,484
Ejaaz:
And there's good reasoning behind this. For the last six to 12 of months,
247
00:16:14,588 -->
00:16:20,974
Ejaaz:
major AI labs have been buying up as much scientific and biological data as they can.
248
00:16:20,974 -->
00:16:26,044
Ejaaz:
This data includes things like lab experiments and how to run these lab experiments
249
00:16:26,044 -->
00:16:31,104
Ejaaz:
and details on DNA and sequences that they can run through AI models.
250
00:16:31,104 -->
00:16:34,514
Ejaaz:
Anthropic has released a bunch of stuff on this. In fact, Anthropic two weeks
251
00:16:34,514 -->
00:16:38,854
Ejaaz:
ago announced that they have a automated wetware lab that they're building that
252
00:16:38,854 -->
00:16:44,994
Ejaaz:
you plug Claude into and it can operate robotic arms to try and run experiments autonomously.
253
00:16:44,994 -->
00:16:48,412
Ejaaz:
Now, the thing that a lot of critics very validly say is.
254
00:16:49,038 -->
00:16:53,972
Ejaaz:
AI models don't work in the physical world. You still need humans and intuition to go do the thing.
255
00:16:54,250 -->
00:16:58,064
Ejaaz:
But what happens when robotics accelerate and become good enough over the next
256
00:16:58,401 -->
00:17:02,129
Ejaaz:
three to five years, and these models are already vastly more capable than human
257
00:17:02,129 -->
00:17:05,289
Ejaaz:
intelligence, and you can just kind of plug them in and, you know,
258
00:17:05,289 -->
00:17:07,219
Ejaaz:
they create a cure for something the next day.
259
00:17:07,567 -->
00:17:10,383
Ejaaz:
That is a very weird world to kind of think of.
260
00:17:10,615 -->
00:17:13,866
Ejaaz:
And I fully think that humans are going to be heavily integrated into this.
261
00:17:14,080 -->
00:17:17,778
Ejaaz:
I guess I just don't know how. It's honestly hard for me to wrap my head around,
262
00:17:18,276 -->
00:17:19,658
Ejaaz:
if I'm being very honest with you.
263
00:17:19,960 -->
00:17:23,559
Ejaaz:
But I think it's an important one to talk about, obviously, as I'm sitting on
264
00:17:23,559 -->
00:17:27,239
Ejaaz:
this hotel floor and my fiance is trying to sleep in the bed next to me.
265
00:17:27,239 -->
00:17:32,289
Ejaaz:
So I hope you found this interesting. I felt very compelled to talk to you guys
266
00:17:32,289 -->
00:17:35,534
Ejaaz:
about this, because this isn't a story about maths at all.
267
00:17:35,842 -->
00:17:39,987
Ejaaz:
It's a story about AI finally being able to surpass human cognition,
268
00:17:40,382 -->
00:17:44,753
Ejaaz:
but doing it much quicker and cheaper than we could have anticipated.
269
00:17:45,351 -->
00:17:49,709
Ejaaz:
And so I'm scared or slightly nervous about how quickly this rolls into any
270
00:17:49,709 -->
00:17:51,417
Ejaaz:
and every other industry.
271
00:17:51,922 -->
00:17:57,669
Ejaaz:
And the ramifications that have on degree holders, workers, job roles that have
272
00:17:57,669 -->
00:18:02,667
Ejaaz:
existed for centuries that have been highly revered suddenly gets turned off overnight.
273
00:18:02,783 -->
00:18:05,559
Ejaaz:
Let me know what you guys think. Do you agree with my takes?
274
00:18:05,559 -->
00:18:08,351
Ejaaz:
Are there anything that I potentially missed?
275
00:18:08,525 -->
00:18:10,985
Ejaaz:
But thank you so much for listening. If you've listened this far,
276
00:18:11,375 -->
00:18:14,029
Ejaaz:
if you enjoyed this episode, you should definitely go check out some of our
277
00:18:14,029 -->
00:18:17,700
Ejaaz:
other episodes. We are back to posting day in and day out five times a week.
278
00:18:18,031 -->
00:18:21,129
Ejaaz:
If you have a friend that you think might find this episode or other episodes
279
00:18:21,129 -->
00:18:24,034
Ejaaz:
interesting, please share it with them. It helps us out massively.
280
00:18:24,191 -->
00:18:27,529
Ejaaz:
I'm anywhere and everywhere, by the way, on X in the description below,
281
00:18:27,529 -->
00:18:31,296
Ejaaz:
or you can just leave a comment here. I read all of them and I try and reply to all of them as well.
282
00:18:31,714 -->
00:18:35,271
Ejaaz:
But thank you so much for listening. This was today's episode on In The Loop
283
00:18:35,445 -->
00:18:37,192
Ejaaz:
brought to you by Qualcomm. See you on the next one.
