WEBVTT

00:00.920 --> 00:03.420
[speaker_0] Hello, my name
is Samuel Shadrach.

00:04.420 --> 00:08.400
I am looking for a co-founder to leak
information about the

00:08.460 --> 00:11.200
US AI companies and the US government.

00:12.040 --> 00:16.020
This plan is based on a mix of Eliezer
Yudkowsky's worldview

00:16.260 --> 00:19.860
and Julian Assange's worldview.
You can find their worldviews both,

00:19.940 --> 00:20.060
you know,

00:20.250 --> 00:21.840
online. There's a lot of writing about it.

00:23.260 --> 00:27.200
I want a political leader in the United
States to

00:27.280 --> 00:30.860
obtain enough power to shut down the AI
companies and

00:30.960 --> 00:32.119
pause AI research.

00:32.980 --> 00:36.280
In order to do this,
they will have to oppose a lot of people

00:36.340 --> 00:36.760
who want to

00:36.820 --> 00:40.640
accelerate AI research.
This includes people in the intelligence

00:40.720 --> 00:44.160
agencies, people in the executive branch,
including, you know, the current

00:44.240 --> 00:47.240
president,
people running various media houses,

00:47.380 --> 00:47.980
people running

00:48.020 --> 00:51.520
various lobby groups,
people funding all of the people I listed

00:51.560 --> 00:51.920
up till

00:51.930 --> 00:54.860
now, you know, including the media,
the lobby groups, intelligence and so on.

00:55.580 --> 00:59.160
I want to leak information about all of
these enemies to the

00:59.200 --> 00:59.700
public.

01:01.920 --> 01:04.780
This rest of the video
is just going to talk about, you know,

01:04.840 --> 01:05.900
like roughly

01:06.000 --> 01:07.040
maybe twenty minutes

01:07.900 --> 01:10.020
describing this plan in more detail.

01:10.180 --> 01:13.180
Before I describe the plan,
I will give you some disclaimers.

01:13.780 --> 01:16.620
The biggest disclaimer is
that all plans are fake.

01:17.960 --> 01:18.320
Like,

01:20.840 --> 01:24.000
over the next few months,
I am going to do more stuff in the real

01:24.040 --> 01:24.700
world and figure

01:24.760 --> 01:26.640
out, you know,
which parts of this plan make sense

01:26.680 --> 01:27.920
and which parts do not.

01:28.200 --> 01:31.460
I do not believe in, you know,
figuring out a perfect plan from day one,

01:32.100 --> 01:35.800
and it just executes perfectly.
I'm publishing incomplete hot

01:35.860 --> 01:39.600
takes and as, you know, we do stuff,
we are going to

01:39.680 --> 01:41.580
find out, you know,
what the actual plan is going to be.

01:42.960 --> 01:45.460
Uh, second disclaimer,
so far I have been working, you know,

01:45.480 --> 01:46.680
all by myself, solo on

01:46.720 --> 01:50.560
this. If hypothetically you
were a co-founder, you chose to work with

01:50.600 --> 01:53.380
me, then obviously your input
is also valued in the plan.

01:53.580 --> 01:56.540
We will have to make the plan together.
It won't just be me making it alone.

01:57.420 --> 01:58.960
So yeah, that's my other disclaimer.

02:00.180 --> 02:00.420
Cool.

02:02.720 --> 02:04.820
First,
I will start with like my broader

02:04.920 --> 02:06.600
worldview,

02:06.660 --> 02:09.160
like not specific to this plan,
but just what is my worldview,

02:09.180 --> 02:09.660
what do I think

02:09.699 --> 02:11.500
about politics, about life,
about whatever.

02:11.650 --> 02:15.460
And from there you will get a sense of,
okay, you know, why am I working on this?

02:16.500 --> 02:17.299
Yeah. So

02:18.720 --> 02:21.540
first of all,
I agree with Eliezer Yudkowsky to a

02:21.600 --> 02:22.120
significant

02:22.180 --> 02:25.880
extent. I think there
is a significant chance we will

02:26.020 --> 02:29.820
soon build an AI that
is better than the best human experts at

02:30.200 --> 02:30.920
everything.

02:31.880 --> 02:34.600
Like every single task you can imagine,
if you can imagine it,

02:34.640 --> 02:35.560
you can also imagine

02:35.600 --> 02:37.760
an AI that's better than the best humans
on that task.

02:38.960 --> 02:41.620
Now, if we do this,
it will be the most important event of

02:41.640 --> 02:42.340
human history.

02:42.540 --> 02:46.460
We will have literally, you know,
a hard disk this big that will

02:46.500 --> 02:49.280
be capable of, you know,
spawning an entire civilization from

02:49.400 --> 02:50.260
scratch with no

02:50.320 --> 02:51.100
human input.

02:53.220 --> 02:56.960
I think there's maybe a twenty-five
percent chance we will get this by 2030.

02:57.400 --> 02:59.100
Uh,
maybe forty percent chance we will get

02:59.140 --> 03:00.360
this by 2035.

03:01.580 --> 03:05.080
If we do build such a superintelligence,
I think the most likely

03:05.160 --> 03:09.040
outcome is we will lose control of the
superintelligence, as in we, the human

03:09.080 --> 03:12.360
race, will lose control of it,
and literally everybody on Earth will die.

03:12.560 --> 03:14.540
I will die, you will die,
your loved ones will die,

03:14.620 --> 03:15.360
everyone will die.

03:16.700 --> 03:20.280
Even if we did figure out how to keep a
superintelligence in control,

03:20.660 --> 03:24.380
I think the most likely outcome is,
you know, whoever was that person who

03:24.420 --> 03:27.860
managed to keep the superintelligence in
their control, they will use it to rule

03:27.900 --> 03:31.260
the world forever. They will become,
you know, permanent global dictator.

03:34.500 --> 03:34.860
Uh,

03:35.700 --> 03:38.920
in spite of all this,
like as we know today, there

03:38.940 --> 03:39.600
are many people

03:39.640 --> 03:41.540
who want to accelerate AI research.

03:42.920 --> 03:45.840
There are many people in the US
intelligence who want to accelerate AI

03:45.880 --> 03:46.520
research.

03:47.280 --> 03:50.520
I think people in US intelligence have had
a lot of

03:50.580 --> 03:54.480
unaccountable power, like forever,
at least on certain key

03:54.560 --> 03:56.850
domains, like whether to start wars,
whether to, you know,

03:56.940 --> 03:58.060
race on key technology,

03:58.160 --> 03:59.460
whether to do surveillance, so on.

04:00.440 --> 04:04.300
So even if you could get fifty-one percent

04:04.520 --> 04:07.700
of the public, in this case the US public,
to, you know,

04:08.580 --> 04:11.040
say, "Okay,
we don't want AI to be accelerated at

04:11.080 --> 04:12.300
maximum pace," even if you could

04:12.360 --> 04:13.580
get that fifty-one percent,

04:14.760 --> 04:18.599
that may not actually get you a pause
because, you know, the

04:18.700 --> 04:20.660
institutional structures are such that,
you know,

04:21.660 --> 04:25.440
fifty-one percent is not enough.
Like the people who run,

04:26.800 --> 04:28.780
uh, who are,
are accelerating AI research are not

04:28.820 --> 04:30.300
that ac-acc-accountable to the

04:30.360 --> 04:30.880
public.

04:35.480 --> 04:39.300
Yeah. And ultimately, uh, now I think,
okay, if you could leak information

04:39.340 --> 04:42.000
about these people,
you can make them more accountable to the

04:42.080 --> 04:42.820
public, more

04:42.860 --> 04:46.440
ensure that, you know,
the public will actually gets translated

04:46.520 --> 04:46.820
into,

04:46.860 --> 04:48.539
you know,
decision-making in the real world.

04:51.300 --> 04:54.500
Okay. Some other plans
which I have chosen not to work on.

04:55.240 --> 04:58.880
At some point, I did think about, okay,
what if we could leak the information of

04:59.020 --> 05:02.420
everybody on the planet, not just,
you know, people building AI?

05:03.180 --> 05:06.400
I now think that plan has a number of
moral issues.

05:06.480 --> 05:10.040
I am not working on that plan.
I'm now working specifically

05:10.180 --> 05:13.440
on leaking specific information about,
you know, specific people who are

05:13.500 --> 05:15.020
accelerating AI research. That's it.

05:17.540 --> 05:20.100
I don't think I should try to become a
billionaire.

05:20.340 --> 05:23.080
If I become a billionaire,
I don't see how I'll be able to pause AI

05:23.120 --> 05:23.520
research.

05:23.560 --> 05:25.120
Like sure, there are things I can try, but

05:26.620 --> 05:28.440
like there's a very long road even from
there.

05:28.480 --> 05:31.880
It's not like, okay, you get,
you get power as a billionaire and,

05:31.900 --> 05:32.240
you know, then

05:32.280 --> 05:33.400
the world gets saved or something.

05:34.340 --> 05:38.130
I don't think I can become a politician
because I fucking hate talking to

05:38.180 --> 05:39.550
people. I'm sorry. [chuckles] That
is true.

05:40.340 --> 05:43.380
And I have learned this the hard way over
the past one year.

05:43.520 --> 05:47.110
Uh, I still think somebody should try to,
you know, persuade

05:47.440 --> 05:50.400
the public around risks from
superintelligence.

05:50.540 --> 05:53.070
I just don't think I can become
that political leader who will, you know,

05:53.100 --> 05:53.200
persuade

05:53.460 --> 05:54.240
millions of people.

05:56.780 --> 06:00.300
I think there are people in the past who
have, you know, opposed, you know,

06:00.360 --> 06:02.920
billionaires, politicians,
powerful people.

06:03.000 --> 06:05.280
In particular, there
are like certain software developers who

06:05.300 --> 06:05.540
have, you know,

06:05.600 --> 06:06.920
become activists and

06:08.304 --> 06:11.784
And, you know, oppose these people,
some of these software developers who have

06:11.884 --> 06:13.804
succeeded,
they have succeeded because they

06:13.844 --> 06:15.584
were living in a different country from

06:15.644 --> 06:17.304
the country whose leaders they want to
oppose.

06:18.004 --> 06:19.964
So, I mean, there
are many examples you can take.

06:19.984 --> 06:22.774
For example, you know,
like Sandra Albakyan who runs, you know,

06:22.784 --> 06:23.523
LibGen.

06:25.444 --> 06:25.644
Uh,

06:27.244 --> 06:29.024
sorry, uh, who runs SkyHub,

06:29.844 --> 06:33.164
who lives in, you know, Eastern Europe,
which makes her more immune to, you know,

06:33.664 --> 06:37.044
uh, lawsuits from, you know, US, uh,

06:37.564 --> 06:40.284
academic publishers.
I can give you many examples like this.

06:40.323 --> 06:41.624
You can go check them on my website.

06:42.504 --> 06:45.704
Yeah, so I think this plan of, okay,
live in a different country from the

06:45.744 --> 06:46.004
country

06:46.044 --> 06:48.784
whose leaders you want to oppose,
I think this plan still has some weight in

06:48.824 --> 06:49.024
it,

06:49.584 --> 06:51.384
and this seems like what I'll be doing.

06:53.264 --> 06:56.864
Lastly, I have typically,
at least until recently,

06:56.984 --> 06:59.304
been, you know,
kind of a believer in capitalism,

06:59.484 --> 07:00.324
in democracy.

07:01.564 --> 07:05.424
Uh, I believed, okay, like, you know,
self-interested

07:05.444 --> 07:08.784
people run the world,
but you can use structures to constrain

07:08.804 --> 07:08.984
their

07:09.024 --> 07:12.004
self-interest,
and this can lead to good outcomes for

07:12.044 --> 07:12.704
society.

07:14.024 --> 07:16.724
However,
now I'm not advocating just capitalism as

07:16.784 --> 07:17.724
usual, democracy as

07:17.764 --> 07:19.764
usual. I'm saying, okay,
we need a mass movement.

07:19.824 --> 07:21.284
We need to, you know,
overthrow these people.

07:21.384 --> 07:22.584
We need to pause AI research.

07:22.664 --> 07:25.104
Like,
this is like radical shift from my earlier

07:25.164 --> 07:25.664
worldview.

07:26.784 --> 07:28.484
And like, yes, my opinions have changed.

07:29.804 --> 07:30.724
Uh, I...

07:31.544 --> 07:34.004
In theory,
you can have a mass movement within,

07:34.084 --> 07:35.334
you know, pure democratic

07:35.444 --> 07:38.384
framework. In practice,
I don't know what a mass movement will

07:38.444 --> 07:39.224
actually do.

07:39.324 --> 07:42.204
Mass movements can become violent.

07:42.264 --> 07:44.414
Mass movements can, you know,
change an old form,

07:44.564 --> 07:45.404
structure of government and

07:45.444 --> 07:49.204
replace it with a new structure.
Exactly what will happen once there is a

07:49.264 --> 07:50.904
mass movement, even I don't actually know.

07:51.724 --> 07:54.584
Uh, my opinion on this is this
is not actually my problem.

07:54.924 --> 07:57.344
Uh, the people who
are in fact leading the mass movement,

07:57.424 --> 07:58.264
this is their problem to

07:58.324 --> 08:01.184
deal with it, okay?
If you did have millions of people who

08:01.244 --> 08:01.564
are, you know,

08:01.584 --> 08:05.484
protesting on the streets to pause AI
research, what these people will do,

08:05.584 --> 08:07.084
what will happen after that, that is

08:07.944 --> 08:11.463
the problem of the peop- the leaders.
That's not my problem personally.

08:13.964 --> 08:17.704
Cool. Next thing I want to talk about
is the theory of change.

08:19.144 --> 08:21.034
I did briefly mention it.
I will again mention it,

08:22.064 --> 08:25.774
uh, more clearly. The theory of change
is simply that,

08:25.804 --> 08:29.444
okay,
if the public knew everything about the

08:29.484 --> 08:30.884
leaders who are governing them,

08:32.084 --> 08:35.404
it becomes more likely
that the leaders will do what the majority

08:35.424 --> 08:35.924
of the public

08:35.984 --> 08:36.864
actually wants.

08:38.504 --> 08:42.244
Uh, otherwise, typically what happens,
leaders face multiple

08:42.484 --> 08:44.844
competing pressures.
They face pressures to make money.

08:45.544 --> 08:48.584
They face pressures to win wars
and defend their territory.

08:49.084 --> 08:52.804
They face pressures to win votes
and get elected, which is quite

08:52.864 --> 08:55.484
closely related to,
but not the same thing as, you know,

08:55.684 --> 08:56.404
doing what the public

08:56.444 --> 08:57.244
actually wants.

08:58.224 --> 09:01.424
Uh, right now,
I just want to strengthen this pressure

09:01.544 --> 09:02.124
that, okay,

09:02.264 --> 09:06.144
leaders should have to actually get votes,
win people's

09:06.184 --> 09:08.664
minds and hearts,
and only then be able to do the thing.

09:09.664 --> 09:09.944
Uh,

09:11.044 --> 09:15.004
so basically, I want to make the,
whatever, democratic pressure

09:15.164 --> 09:19.124
stronger as compared to, you know,
market pressure or military pressure or

09:19.164 --> 09:19.524
whatever.

09:21.204 --> 09:24.224
Uh,
there are many people in effectual circles

09:24.264 --> 09:25.164
in particular who have

09:25.184 --> 09:28.824
various theories of change. Uh,
some of the theories of change

09:29.124 --> 09:32.984
often include, okay, identify who
is powerful and then we will go side with

09:33.044 --> 09:33.944
them, ally with them.

09:35.044 --> 09:35.404
Uh,

09:36.444 --> 09:38.254
my theory of change is not this.
My theory of change

09:38.284 --> 09:39.344
is not trying to be like,

09:39.404 --> 09:42.724
okay, let's find the specific, you know,
powerful person and go persuade that guy

09:42.844 --> 09:46.643
or woman, whatever.
My theory of change is, okay, if

09:46.744 --> 09:50.524
all the leaders' information was public,
then those leaders actually don't have

09:50.564 --> 09:52.244
freedom to take independent decisions.

09:52.284 --> 09:55.984
They will have to do whatever the public
wants out of them, and I think this is

09:56.004 --> 09:57.344
good in this case.

10:00.364 --> 10:03.064
Uh,
maybe leaking information about them

10:03.104 --> 10:04.204
is enough to make them

10:04.244 --> 10:06.464
accountable. Maybe that
is not actually enough.

10:06.484 --> 10:09.584
You need actual mass movement, actual,
you know, people protesting.

10:10.644 --> 10:14.024
I do not actually know. I'm not sure.
I think both are worth trying.

10:14.104 --> 10:16.984
Someone should try leaking information on
these people and someone should try

10:17.044 --> 10:18.064
building a mass movement.

10:22.524 --> 10:22.704
Cool.

10:23.884 --> 10:25.424
Next section of this video.

10:26.244 --> 10:28.234
What is my motivation for working on this?

10:28.444 --> 10:31.354
So like, in my head, motivation
is different, theory of change

10:31.364 --> 10:31.814
is different.

10:31.964 --> 10:35.424
Theory of change is, okay, if I do X,
then Y will happen in the world,

10:35.504 --> 10:35.824
then A will

10:35.884 --> 10:37.894
happen, then B will happen.
You can have your whole, you know,

10:38.224 --> 10:39.204
galaxy brain plan.

10:39.844 --> 10:42.824
Motivation is about what you feel,
what are your emotions around this whole

10:42.864 --> 10:43.044
thing,

10:43.664 --> 10:46.724
and feelings don't have to be rational
or justified by facts, whatever.

10:47.724 --> 10:49.524
So yeah, what is my motivation?
What do I feel?

10:49.564 --> 10:52.024
Like, what are my feelings that
are driving towards like, okay,

10:52.064 --> 10:52.764
you have to go work

10:52.784 --> 10:53.204
on this?

10:54.184 --> 10:57.544
I think my biggest motivation here,
and I'm not actually sure if, you know, my

10:57.564 --> 11:01.504
motivations change with time or whatever,
but I think my biggest motivation

11:01.564 --> 11:02.464
here is

11:03.444 --> 11:07.324
I don't want to die. I don't want,
you know, everyone I care about

11:07.364 --> 11:08.484
to die. I

11:09.724 --> 11:13.224
don't want these people who
are building superintelligence to kill me

11:13.264 --> 11:13.364
and

11:13.404 --> 11:17.264
everyone I care about. And
if someone comes and, you know,

11:17.584 --> 11:19.994
tries to kill me,
I'm willing to go pretty far to, you know,

11:20.024 --> 11:20.684
fucking fight that

11:20.764 --> 11:21.184
person.

11:22.064 --> 11:25.964
Uh, like take a hypothetical. Imagine,
imagine this was not, you

11:25.984 --> 11:26.684
know, some random

11:27.664 --> 11:30.024
AIC or whatever. Imagine it
was one of my friends.

11:30.364 --> 11:33.224
Tomorrow they decided, okay,
they will build a bomb that bo-

11:33.364 --> 11:34.224
blows up everyone on

11:34.264 --> 11:37.864
the planet. Uh, they had some, you know,
deranged beliefs.

11:37.924 --> 11:39.464
Oh, this will get us utopia, whatever.

11:40.124 --> 11:42.834
Uh, I don't buy those beliefs,
but what in practice they're doing,

11:42.904 --> 11:43.044
they're

11:43.084 --> 11:46.064
literally spending day
and night working on a bomb to blow up

11:46.084 --> 11:46.524
everyone

11:47.524 --> 11:51.504
and the plan to set it off. Now,
I might like my friend, like I

11:51.564 --> 11:54.104
might like my friend for their honesty,
for their courage.

11:54.204 --> 11:57.524
They might be, you know,
high on openness to experience like me.

11:57.564 --> 12:01.144
They might be atheist. They might,
you know, support technological progress.

12:01.184 --> 12:02.744
There may be various reasons I like my
friend.

12:03.684 --> 12:06.024
But the fact that I like my friend
is less important.

12:06.084 --> 12:08.544
The fact that they plan to fucking kill me
is more important.

12:09.184 --> 12:11.533
And therefore, you know,
if I have to go kill my friend, then,

12:11.604 --> 12:12.144
you know, I will do

12:12.224 --> 12:12.544
it.

12:13.484 --> 12:13.904
Uh-

12:15.732 --> 12:17.852
Now I have to caution a bit here, like,

12:19.172 --> 12:22.512
I am not saying, okay, we should just go,
you know, kill everyone who's going to

12:22.532 --> 12:24.492
build ASI and that will fix the problem.

12:24.632 --> 12:26.752
Assassinations don't actually work.

12:26.952 --> 12:27.332
Like,

12:28.272 --> 12:30.072
I am not trying to assassinate people.

12:30.152 --> 12:33.892
I am not actively asking anyone to attempt
an assassination either.

12:34.472 --> 12:38.092
However, emotionally in my head, I'm like,
yeah, if we have to go that far, of

12:38.132 --> 12:40.552
course, we're going to go that far. Like,
there's no question about it.

12:43.632 --> 12:46.992
Uh, this is my strongest motivation,
but there are other motivations also.

12:47.052 --> 12:50.772
You can go look into the document I wrote
for some of them.

12:50.892 --> 12:53.502
Uh,
I often remind myself about the fact that,

12:53.552 --> 12:54.562
you know, if all this

12:54.632 --> 12:55.372
happens, if

12:56.252 --> 13:00.202
I am not crazy and this, you know,
ASI gets built, you know, I am going to

13:00.272 --> 13:02.672
die.
I have taken various steps to keep

13:02.732 --> 13:03.892
reminding myself I'm going to

13:03.952 --> 13:04.292
die.

13:05.912 --> 13:09.072
If I have to sacrifice myself
and the world gets saved, you know,

13:09.112 --> 13:09.682
I am fine with

13:09.772 --> 13:10.142
that.

13:11.292 --> 13:14.672
I generally hate authority.
That has always been a thing.

13:14.712 --> 13:18.652
This could be authority of parents
or family, of religion,

13:18.752 --> 13:22.592
of governments, of educational systems,
like any sort of authority.

13:23.552 --> 13:27.492
And I want the truth about, you know,
what all these forms of authority

13:27.512 --> 13:30.712
are actually doing. Like,
how does society function as a result of,

13:30.752 --> 13:31.002
you know, all

13:31.072 --> 13:32.972
this authority,
all this pressure being exerted on people?

13:33.612 --> 13:36.972
I want the truth of all of this to come
out publicly, and I have some hope, you

13:37.012 --> 13:37.112
know,

13:37.992 --> 13:41.892
if people are still alive, you know,
in 10 years or 50 years, 100

13:41.932 --> 13:43.892
years from now,
they can use all this information

13:43.932 --> 13:45.432
and they can design some better

13:45.492 --> 13:48.952
society for themselves, better than,
you know, the current whatever we live in.

13:51.792 --> 13:55.692
Yeah, so I want the truth to come out.
I am a very strong believer in the truth.

14:00.172 --> 14:02.712
Okay, some more motivations. Uh,

14:04.532 --> 14:08.112
I am not pretending to represent,
you know, all of

14:08.132 --> 14:10.092
humanity's interests like some...

14:11.052 --> 14:14.012
I am some benevolent leader who, you know,
accounts everyone's interests and

14:14.092 --> 14:16.892
calculates, you know,
like in Yudkowsky's worldview, this

14:16.932 --> 14:18.012
is like a CEV.

14:18.932 --> 14:19.352
Uh,

14:20.932 --> 14:23.292
like there are different philosophies
or different names for it,

14:23.352 --> 14:24.092
but like I'm not

14:24.152 --> 14:26.532
trying to say, "Oh,
I am representing everyone's interests." I

14:26.551 --> 14:27.352
am here to represent

14:27.392 --> 14:28.912
my interests, which
is I don't want to die

14:28.952 --> 14:30.072
and I don't want to live in a dead

14:30.132 --> 14:32.832
world.
And you might also benefit from that.

14:32.892 --> 14:35.012
Like, if I save the world,
you will also not die,

14:35.042 --> 14:36.272
and that's also in your benefit.

14:36.332 --> 14:40.092
But I am not, like, saying, "Oh,
I will just

14:40.132 --> 14:43.972
defer to you on things." Like,
I'm doing this because I want the world to

14:44.012 --> 14:44.092
be

14:44.112 --> 14:45.892
a certain way and not another way.

14:47.312 --> 14:50.332
Yeah, and many people say they
are fine with human extinction.

14:50.912 --> 14:54.072
I think that some of them are wrong,
but also it doesn't matter.

14:54.332 --> 14:56.632
You may be fine with it.
I am not fine with it and, you know,

14:56.672 --> 14:57.312
I'm going to fight

14:57.352 --> 14:57.572
this.

15:01.212 --> 15:05.192
I am a huge fan of Naval Ravikant's quote
that I

15:05.232 --> 15:07.852
should not just keep blaming other people
that the world is ending.

15:07.932 --> 15:10.252
I should also, you know,
take ownership of this problem

15:10.292 --> 15:10.982
and be like, you know, if

15:10.992 --> 15:12.432
the world ends, this is my fault as well.

15:12.572 --> 15:16.412
Like,
I also should have done my best to prevent

15:16.452 --> 15:19.812
this from happening. And if I failed to,
you know, do my best, I should also blame

15:19.852 --> 15:20.412
myself.

15:21.412 --> 15:23.732
And I, yeah, I buy that.

15:26.192 --> 15:26.692
Uh,

15:27.552 --> 15:31.392
a more self-interest, again,
another reason, like,

15:32.752 --> 15:35.472
working alone on problems is, you know,

15:36.072 --> 15:37.172
demotivating.

15:38.012 --> 15:40.972
I don't want to become depressed.
I think there is a small chance,

15:41.052 --> 15:41.592
not very big,

15:41.652 --> 15:44.972
but a small chance I will become depressed
if I don't find anyone to work with

15:45.012 --> 15:48.112
this. So again,
I would really like to work with someone

15:48.152 --> 15:48.672
on this.

15:49.892 --> 15:53.492
Uh, don't join just because, oh,
Samuel's mental health or whatever.

15:53.572 --> 15:56.472
That is a bad reason to join.
If you actually believe in this cause

15:56.512 --> 15:56.692
and you

15:56.732 --> 15:59.732
actually want to go all in,
then you can come work with me.

16:02.232 --> 16:02.532
Uh,

16:03.732 --> 16:07.612
if I am going to die for whatever reason,
be it due to ASI, be it for any other

16:07.621 --> 16:10.312
reason, like,
I want people to remember that, okay,

16:10.392 --> 16:11.432
here are things I

16:11.452 --> 16:15.132
did that were beneficial for the

16:15.192 --> 16:18.612
world. Uh,
I want people to remember my work after I

16:18.672 --> 16:18.952
die.

16:19.072 --> 16:21.032
I don't want to just die and be forgotten.

16:21.252 --> 16:22.892
I think that's another motivation I do
have.

16:25.092 --> 16:28.892
So that I think concludes the section on,
you know, what are my motivations, why am

16:28.932 --> 16:29.832
I working on this.

16:32.252 --> 16:35.612
Okay, next section,
what will I not be working on?

16:36.512 --> 16:39.412
I am, like I said,
I am not working on persuading people.

16:39.712 --> 16:41.492
My focus is on leaking information.

16:41.632 --> 16:45.012
Other people can work on persuading people
that, you know, superintelligence is

16:45.052 --> 16:48.012
coming, the world is ending,
you need to do something about it today.

16:50.132 --> 16:54.062
I am not working on sabotage.
Sabotage means basically destroying other

16:54.112 --> 16:57.512
people's, uh, equipment, property,

16:57.852 --> 16:59.952
occasionally livelihood. But yeah,

17:00.892 --> 17:03.212
like,
I don't think sabotaging AI companies

17:03.252 --> 17:04.372
is actually going to work.

17:04.481 --> 17:06.152
Again,
there are too many people trying to build

17:06.212 --> 17:06.532
AI.

17:06.872 --> 17:07.092
Uh,

17:08.232 --> 17:09.592
if you just go, you know,

17:10.632 --> 17:14.492
destroy a few computers or something,
be it by hacking, be it any means, like,

17:14.832 --> 17:17.952
this doesn't actually stop the AI race.
There are too many people.

17:17.992 --> 17:19.892
There's too much money in this space.

17:20.752 --> 17:24.352
Like, if you actually, let's say,
you know, blew up TSMC, I can see how that

17:24.412 --> 17:27.432
actually helps. But apart from that,
I don't see how sabotage actually helps.

17:27.932 --> 17:29.692
Therefore, I am not working on sabotage.

17:30.312 --> 17:32.811
I am not working on assassinating anyone.

17:32.932 --> 17:34.772
Uh, I want to again make this clear.

17:34.832 --> 17:38.612
I don't think assassinating any one person
is actually going to stop the AI race.

17:41.952 --> 17:42.252
Cool.

17:43.092 --> 17:46.152
Now I'm going to come to an important
section on morality.

17:48.472 --> 17:51.152
Yeah,
why am I talking about morality so much?

17:51.392 --> 17:55.272
Uh, well, firstly, because a lot of,
you know, people in

17:55.452 --> 17:58.612
EA strong circles had issues with my moral
worldview or whatever.

17:59.232 --> 18:03.172
But also because I think morality
is one of the core things

18:03.252 --> 18:04.872
about a person,
one of the core things

18:04.912 --> 18:06.552
that helps people to coordinate.

18:06.952 --> 18:07.212
Uh,

18:08.252 --> 18:12.072
if I'm trying to find more people to
coordinate with me, it makes at least some

18:12.132 --> 18:13.932
sense for me to talk about morality.

18:14.132 --> 18:17.922
Also, the kind of plan we're working on,
we will have to cross some moral lines, I

18:18.012 --> 18:20.002
think. Especially if we
are crossing moral lines,

18:20.092 --> 18:21.152
we need to be careful about,

18:21.212 --> 18:23.232
okay, which lines we're crossing
and why and so on.

18:26.242 --> 18:30.182
Yeah, my worldview on morality is I think,
first of all, most writing on

18:30.222 --> 18:32.582
morality is kind of useless. It's just,

18:33.462 --> 18:35.082
yeah,
like if you see like writings on moral

18:35.122 --> 18:36.462
philosophy and then what those people

18:36.482 --> 18:39.002
actually go do in the real life,
like these are like two completely

18:39.062 --> 18:39.362
separate

18:39.422 --> 18:43.082
things. And therefore,
I think most writing on morality is

18:43.182 --> 18:46.602
useless. Well, maybe useless
is going too far.

18:46.622 --> 18:50.382
I would say not that useful.
Like morality is something

18:50.422 --> 18:53.722
that's tested in practice, not in theory.
I very strongly believe this.

18:54.662 --> 18:57.222
Uh,
one thing I do believe about morality

18:57.322 --> 18:58.032
is that,

18:59.342 --> 19:01.762
and again, I'm not sure, but this
is something I believe, like

19:01.842 --> 19:02.922
if you have had like

19:02.962 --> 19:06.522
a very nice life,
like you've had a comfortable life where--

19:06.562 --> 19:06.742
with

19:06.762 --> 19:08.742
people who you trust,
with people you're happy with,

19:08.802 --> 19:10.722
like you will by default want

19:10.742 --> 19:12.101
to be nice to people all the time.

19:13.002 --> 19:16.311
I don't think being nice to people all the
time is going to fix the problem of risk

19:16.322 --> 19:19.942
from superintelligence.
I think some people need to be less nice

19:20.022 --> 19:20.222
to the

19:20.242 --> 19:22.662
people building superintelligence and,
you know, go fight them.

19:23.802 --> 19:26.542
Uh, again, not physically fight them,
but fight them in other means.

19:26.922 --> 19:27.042
Uh,

19:28.402 --> 19:32.142
so if you have had a very nice life,
your default approach may or may not

19:32.182 --> 19:32.802
actually work.

19:33.602 --> 19:35.942
If you have suffered a lot in life,
now you have a choice.

19:36.022 --> 19:39.842
Like you understand that, okay,
when a situation comes, there is a

19:39.902 --> 19:42.542
more nice way of dealing with it
and a less nice way of dealing with it.

19:43.482 --> 19:47.122
And I personally, I am not everyone,
I am a specific person, I am

19:47.142 --> 19:51.092
sympathetic to the people who have
suffered in the past and then therefore

19:51.142 --> 19:53.542
they understand that sometimes you have to
take the less nice route.

19:53.582 --> 19:56.062
You can't always be nice
and expect situations to get solved.

19:57.862 --> 20:00.922
I am sympathetic those-- to those people.
I am one of those people.

20:01.222 --> 20:01.822
I might

20:02.622 --> 20:06.422
prefer working with someone like
that who understands this

20:06.862 --> 20:08.042
fact about human nature.

20:10.302 --> 20:13.062
Uh,
I think it's possible to go too far with

20:13.142 --> 20:13.322
this.

20:13.402 --> 20:15.102
Like,
I don't want to become someone who

20:15.182 --> 20:17.202
full-time enjoys, you know, inflicting

20:17.242 --> 20:18.962
pain and suffering on my enemies. But

20:19.942 --> 20:22.052
like we need to understand our job
is to inflict pain

20:22.062 --> 20:23.182
and suffering on our enemies.

20:23.262 --> 20:25.842
Like they need to lose power,
we need to gain power.

20:29.702 --> 20:32.162
Uh,
I don't think I'm alone in having

20:32.222 --> 20:33.502
understanding of this, you know,

20:33.542 --> 20:36.672
fact about human nature.
I think many political leaders also have

20:36.702 --> 20:37.342
understood that

20:37.362 --> 20:39.862
sometimes you have to take, you know,
the less nice route.

20:41.022 --> 20:44.682
Uh, examples, Winston Churchill,
in my view,

20:44.742 --> 20:47.942
basically spent his entire career
brainwashing teenagers to, you know,

20:47.982 --> 20:48.502
walk into

20:48.542 --> 20:50.162
their own deaths. And I think this
is fine.

20:50.202 --> 20:52.322
He's still on the right side of history
for me.

20:52.442 --> 20:54.742
Uh,
Nelson Mandela's wife basically joined a

20:54.762 --> 20:56.092
terrorist organization to, you know,

20:56.122 --> 20:58.082
do attacks against the British. I

20:58.942 --> 21:01.342
think that was fine.
She's on the right side of history.

21:01.522 --> 21:04.862
Was it all this most maximally effective
and all that? I don't know.

21:04.942 --> 21:07.582
But like I think morally all this is okay.

21:08.282 --> 21:09.682
Uh,
there were people who tried in

21:09.742 --> 21:10.722
assassinating Hitler.

21:10.822 --> 21:12.942
I support the people who tried
assassinating Hitler.

21:13.562 --> 21:15.202
Uh,
I respect all these people for

21:15.242 --> 21:16.982
understanding that sometimes you do need
to

21:17.042 --> 21:17.941
actually go that far.

21:20.762 --> 21:22.342
Okay,
so this is my thing on morality in

21:22.382 --> 21:22.742
general.

21:22.812 --> 21:25.422
Now applying it to the specific plan of
leaking information.

21:25.482 --> 21:29.142
Like when it comes to leaking information,
what are the specific moral

21:29.162 --> 21:33.072
decisions we have to make? Uh, first,
who are we targeting?

21:33.302 --> 21:36.122
Second,
what are we going to leak to the public?

21:36.182 --> 21:38.262
And third, what methods
are we going to use?

21:39.442 --> 21:42.682
Okay, so the first one is easy. Uh,
who are we targeting?

21:42.722 --> 21:46.102
For me, it's like, for me, the answer
is everyone. It's a very broad umbrella.

21:46.302 --> 21:50.182
Like we can target anyone working at any
of the USA companies,

21:50.242 --> 21:53.922
anyone working in the US intelligence
community, anyone working in any of the

21:53.942 --> 21:57.642
political campaigns working to accelerate
AI, anyone being

21:57.702 --> 22:00.062
funded,
anyone who's funding any of these people.

22:00.502 --> 22:03.062
Like this could be senior people,
junior people,

22:04.062 --> 22:06.702
people currently working,
people previously working.

22:06.762 --> 22:09.682
Like the umbrella is very broad.
Like for who are we targeting?

22:09.762 --> 22:13.242
I'm targeting everyone.
Like I am fine with acquiring

22:13.302 --> 22:17.122
information on everyone,
but I want to be selective about

22:17.162 --> 22:19.932
what information we leak.
Like we can acquire a lot of information,

22:20.002 --> 22:20.422
but then what

22:20.442 --> 22:21.942
we leak to the public is a separate thing.

22:21.971 --> 22:23.162
There,
I would like this to be a bit more

22:23.202 --> 22:23.622
selective.

22:25.502 --> 22:29.102
Now what-- Now comes question two,
what information am I okay leaking to the

22:29.142 --> 22:31.722
public as opposed to, you know,
we just acquire it,

22:31.782 --> 22:33.082
we kept it for ourselves, we

22:33.102 --> 22:33.942
don't do much with it?

22:35.502 --> 22:39.442
The short answer is I don't actually know
what information we should be

22:39.482 --> 22:42.102
leaking to the public.
Like the end goal is, okay,

22:42.162 --> 22:43.022
make all these leaders

22:43.042 --> 22:46.462
accountable to the public and, you know,
stop the AI race.

22:47.542 --> 22:50.642
Now, in order to do that,
what information should we leak such

22:50.722 --> 22:51.452
that this happens?

22:51.562 --> 22:54.442
I don't actually know.
I have not yet done surveys.

22:54.662 --> 22:55.722
I just have some guesses

22:56.642 --> 22:59.242
based on, you know,
my reading of previous leaks and so on.

23:00.322 --> 23:04.082
Uh, guess number one, I want, want to

23:04.142 --> 23:07.622
find out like why don't we have anti-AI

23:07.702 --> 23:09.862
political leaders already? Like

23:11.302 --> 23:14.722
already if you survey, you know,
the public, it's like, you know, 50%

23:14.942 --> 23:18.542
pro-AI, 50% anti.
So like half the public is anti-AI.

23:18.582 --> 23:20.602
Like why are there no leaders representing
these people?

23:20.642 --> 23:24.282
Like, and why do they not have, you know,
50% of the political power if, you know,

23:24.462 --> 23:26.402
50% of the people are voting against AI?

23:27.242 --> 23:29.182
So I want to find out, you know, what
are the pressures that

23:29.242 --> 23:30.722
are being put on all

23:30.762 --> 23:33.142
these anti-AI political leaders?

23:33.202 --> 23:37.102
Like why are these people not already in
power? Like what is stopping them?

23:37.162 --> 23:37.362
Like

23:39.022 --> 23:42.222
are they being, you know,
cut out from the relevant networks of

23:42.242 --> 23:42.882
elites?

23:43.102 --> 23:47.062
Do they not get funding?
Do media houses refuse to work with them?

23:47.702 --> 23:50.212
Are they just being demoralized?
Are they being, you know,

23:50.262 --> 23:51.142
physically threatened?

23:51.262 --> 23:54.122
Like what is stopping them? This
is something I want to understand

23:54.202 --> 23:55.002
and also leak to

23:55.042 --> 23:55.662
the public.

23:56.502 --> 24:00.422
So a concrete example, in the 2016 US

24:00.502 --> 24:04.442
presidential election, uh, it
was found out that people

24:04.482 --> 24:08.262
in the DNC were conspiring against Bernie
Sanders in

24:08.322 --> 24:11.902
favor of Hillary Clinton. Uh,
a hacker group basically got

24:12.022 --> 24:15.362
emails within people in the DNC,
and they leaked these emails.

24:15.562 --> 24:18.402
And as a result,
I think Hillary Clinton actually became

24:18.442 --> 24:19.122
less popular.

24:19.462 --> 24:21.862
Uh, Bernie Sanders, you know,
became slightly more popular.

24:22.802 --> 24:24.842
Uh,
I think this is an example of the type of

24:24.902 --> 24:26.582
thing I would want to leak.

24:26.642 --> 24:29.182
I would want to leak, you know, what
are the pressures all these people

24:29.222 --> 24:29.762
are facing

24:29.982 --> 24:33.874
and why are they not able To come to power
if they are actually, you

24:33.914 --> 24:35.374
know, representing people's interests.

24:37.334 --> 24:39.754
Second,
I want to leak information about personal

24:39.814 --> 24:40.314
abuse.

24:40.814 --> 24:43.734
For example,
Sam Altman has been accused of raping his

24:43.774 --> 24:44.254
sister.

24:44.514 --> 24:45.854
I don't know if that actually happened.

24:45.954 --> 24:49.874
However, if this actually happened,
I want evidence about this to be leaked to

24:49.914 --> 24:50.474
the public.

24:52.094 --> 24:56.064
Uh, again, if you had enough--
if you had actual concrete evidence, you

24:56.074 --> 24:59.404
had enough people rallying around this,
I think you actually can put powerful

24:59.494 --> 25:03.214
people in prison, even if, like,
if they are, you know, rapists or,

25:04.314 --> 25:05.854
uh, similarly abusive.

25:06.694 --> 25:09.674
I think this can be done. Historic-- Like,
doing this is hard.

25:09.734 --> 25:12.934
Like, you know, for example,
you obviously know about the Epstein

25:12.994 --> 25:13.424
files,

25:13.514 --> 25:15.874
where there have been a lot of
speculations,

25:15.974 --> 25:17.494
but not enough concrete evidence

25:17.554 --> 25:19.333
as to, you know, who is a rapist, who
is not.

25:19.474 --> 25:22.974
If you had concrete evidence and you had,
you know, public pressure, I still think

25:23.014 --> 25:24.454
you can put these people in prison.

25:25.834 --> 25:28.194
Third,
I want to leak information about

25:28.254 --> 25:29.574
democratic safeguards being

25:29.614 --> 25:33.354
broken. So an example here, again,
famous example

25:33.574 --> 25:35.254
would be when Edward Snowden,

25:36.074 --> 25:37.774
uh,
talked about the NSA surveillance

25:37.834 --> 25:38.494
programs.

25:38.754 --> 25:42.474
He particularly talked about how various
congressmen, various

25:42.494 --> 25:44.814
senators were not informed about these
programs.

25:44.874 --> 25:47.934
He talked about how various Supreme Court
judges were not informed about these

25:47.954 --> 25:51.074
programs.
He talked about how secret courts were

25:51.084 --> 25:55.064
established, uh, so
that you did not have to go through

25:55.064 --> 25:58.644
the official courts to, you know, get,
you know, your warrant or whatever for the

25:58.734 --> 25:59.354
surveillance.

26:00.614 --> 26:02.644
By default,
I'm expecting similar things

26:02.674 --> 26:03.714
are going to happen again.

26:03.814 --> 26:07.174
You know, the US AI companies
and the US intelligence are going to make

26:07.194 --> 26:10.274
alliances and have--
and they could end up having some secret

26:10.284 --> 26:11.034
programs running.

26:11.074 --> 26:13.914
I would want to leak information about
those secret programs.

26:16.054 --> 26:19.994
Uh,
then I want to leak information about who

26:20.074 --> 26:22.794
in the AI companies
is actually fine with human extinction.

26:23.614 --> 26:23.934
Like,

26:25.954 --> 26:29.754
supporting human extinction actively
is very unpopular,

26:29.894 --> 26:31.894
and I want it to remain very unpopular.

26:32.274 --> 26:35.914
I want to leak information about, like,
who's, you know, building

26:36.234 --> 26:40.213
ASI and secretly believes, okay,
human extinction due to

26:40.294 --> 26:44.254
ASI is a good thing. Again,
I don't want to punish free

26:44.394 --> 26:46.694
speech here. Like,
you have an unpopular opinion, fine.

26:47.114 --> 26:50.434
But if you use that unpopular opinion to,
you know, literally risk everyone's lives

26:50.474 --> 26:51.943
on Earth, then I want that to be leaked.

26:51.994 --> 26:54.034
That, okay, this is,
you had this unpopular opinion,

26:54.074 --> 26:54.854
therefore you did it.

26:55.814 --> 26:59.704
So concrete example here, uh,
Elon Musk has accused

26:59.834 --> 27:02.294
Larry Page of supporting human extinction.

27:03.374 --> 27:05.834
Uh,
Larry Page has a controlling interest in

27:05.874 --> 27:06.734
Google and Google

27:06.774 --> 27:07.374
DeepMind.

27:08.574 --> 27:12.394
Jürgen Schmidhuber, uh, more
or less supports human

27:12.434 --> 27:14.694
extinction. Again,
he's waffly about his language,

27:14.794 --> 27:16.054
but he basically is fine with,

27:16.134 --> 27:18.014
you know,
an AI takes over the civilization and,

27:18.094 --> 27:18.704
you know, humans are

27:19.014 --> 27:20.274
ir-irrelevant, blah, blah, blah.

27:21.134 --> 27:24.674
Uh, many of DeepMind's early employees,
such as, you know, Shane Legg, such as

27:24.714 --> 27:28.154
Marcus Hutter, such as Seth Hochreiter,
they were trained under Jürgen

27:28.194 --> 27:32.074
Schmidhuber. So if all of these people,
you know, such as Shane

27:32.114 --> 27:35.114
Legg, such as Marcus Hutter, and so on,
if they are all actually fine with human

27:35.174 --> 27:37.974
extinction, then I also want
that information to reach the public.

27:38.474 --> 27:39.313
I want to leak that.

27:41.534 --> 27:43.174
Lastly, uh,

27:44.294 --> 27:47.194
if there are any, like,
rogue groups who are trying to, you know,

27:47.494 --> 27:47.694
build

27:47.774 --> 27:51.714
superintelligence in secret, like, again,
I would want to leak that also.

27:51.754 --> 27:54.754
Like, to pause AI,
it's not enough to just have, like, a

27:54.774 --> 27:58.494
treaty signed by some leaders.
You also need to enforce

27:58.574 --> 28:00.414
it.
If there are people in secret trying to

28:00.454 --> 28:02.254
build, you know, ASI, I don't want them

28:02.294 --> 28:04.094
to build it in secret. I want to leak,
okay,

28:04.174 --> 28:05.694
who's trying to build this in secret.

28:06.494 --> 28:08.484
Yeah,
these are my guesses for what information

28:08.514 --> 28:09.164
we have to leak.

28:09.174 --> 28:11.214
But again, I will reiterate,
I don't actually know.

28:11.274 --> 28:12.974
We are going to find this out as we go
along.

28:13.874 --> 28:17.374
Uh,
never in human history have we lived under

28:17.474 --> 28:17.714
fully

28:17.794 --> 28:21.094
transparent leadership.
Leadership has always been secretive by

28:21.134 --> 28:21.674
default

28:22.594 --> 28:26.514
and unaccountable by default. Uh,
today we have

28:26.574 --> 28:29.654
tools that enable, you know, less secrecy,
more

28:29.704 --> 28:33.174
transparency, and maybe also more

28:33.214 --> 28:35.134
accountability,
although that's kind of debated.

28:35.854 --> 28:39.194
So we are going to find this out as we go
along, what leaks are actually the most

28:39.234 --> 28:39.684
useful.

28:41.634 --> 28:44.134
Okay,
so I initially started with we need to

28:44.194 --> 28:45.614
make some moral decisions, like what

28:45.694 --> 28:48.554
information can we acquire?
What do we leak?

28:48.814 --> 28:52.114
And what methods to be used?
So what information we acquire, again,

28:52.194 --> 28:52.634
it's like we

28:52.674 --> 28:55.794
can acquire everything. What do we leak?

28:56.754 --> 28:59.884
Uh, my answer is I don't know,
but here are my specific guesses of some

28:59.914 --> 29:00.354
information

29:00.394 --> 29:04.084
we, I think we should try to leak.
Third comes what methods are we allowed to

29:04.154 --> 29:07.174
use? [coughs] Now,
when it comes to what methods

29:07.214 --> 29:08.114
are we allowed to use,

29:08.214 --> 29:09.874
now,
if you're trying to get information on

29:09.894 --> 29:11.194
someone, you have to break their

29:11.234 --> 29:14.234
security in some way. Now,
typically this means either breaking

29:14.274 --> 29:15.194
physical security

29:15.434 --> 29:18.854
or cybersecurity, uh, more,
more broadly information security.

29:18.874 --> 29:20.804
So physical security,
information security,

29:21.034 --> 29:22.354
or you break their social circle.

29:23.494 --> 29:27.394
So example of breaking physical security
would be, you know, the Pentagon Pizza

29:27.414 --> 29:27.914
Index,

29:29.014 --> 29:31.434
uh, [coughs] which was basically

29:32.574 --> 29:36.314
people found out that, okay, when a war
is about to be declared, you know,

29:36.414 --> 29:40.174
people near the Pentagon order a lot of
pizzas, and this shows up on Google

29:40.234 --> 29:43.154
Maps or whatever. So you can predict,
okay, if a lot of pizzas

29:43.174 --> 29:43.664
are being ordered,

29:43.714 --> 29:45.814
probably a war is going to be, you know,
ordered.

29:47.134 --> 29:49.674
Uh, another example is this famous

29:50.614 --> 29:53.514
Substack writer, Dominic Cummings,
who worked under,

29:54.554 --> 29:58.194
like, worked, uh,
during the Brexit political campaigns in

29:58.254 --> 29:58.314
the

29:58.394 --> 30:01.054
UK.
He wrote about how he basically literally

30:01.064 --> 30:02.314
searched through the trash cans of

30:02.354 --> 30:04.614
his opponent,
and he found the entire political campaign

30:04.674 --> 30:05.594
of his, you know, opponent

30:05.654 --> 30:08.614
in those trash cans. So these
are examples of, you know,

30:08.674 --> 30:09.294
breaking physical

30:09.354 --> 30:11.354
security, where your

30:12.174 --> 30:15.774
enemy has left some, you know, clues,
some trails, some information in physical

30:15.874 --> 30:18.894
space. Uh,
there are more violent ways of this

30:18.914 --> 30:19.794
that the

30:19.854 --> 30:22.614
most violent is something like, you know,
literally kidnap someone and get

30:22.654 --> 30:25.594
information out of them.
I'm not very keen on doing that,

30:25.654 --> 30:26.594
at least to begin with.

30:26.834 --> 30:26.974
Uh,

30:27.954 --> 30:28.234
that's

30:29.054 --> 30:32.974
physical security.
Next comes cybersecurity, information

30:33.034 --> 30:36.194
security. Uh,
this means basically we hack into their

30:36.254 --> 30:36.934
servers, and we get

30:36.974 --> 30:38.034
information out of that

30:39.554 --> 30:43.294
Third,
and what I'm actually have most hope about

30:43.354 --> 30:43.534
right

30:43.574 --> 30:45.534
now is breaking their social circle.

30:46.194 --> 30:49.434
Breaking social circle basically means
find someone in their social circle who

30:49.454 --> 30:51.774
will, you know, turn against them
and give us the information.

30:54.354 --> 30:57.474
Now,
this has already historically been done by

30:57.534 --> 30:58.114
various, you know,

30:58.174 --> 31:01.734
journalists and, you know, OSINT, uh,
open-source intelligence

31:01.794 --> 31:02.634
specialists.

31:03.934 --> 31:06.844
Uh, there are, at least today, there
are now various media houses

31:06.874 --> 31:07.444
that work with,

31:07.494 --> 31:08.864
you know,
OSINT specialists because they've

31:08.874 --> 31:10.064
understood this is how, you know,

31:10.114 --> 31:11.054
journalism works today.

31:11.974 --> 31:15.654
So OSINT includes everything from,
you know, analyzing social media to,

31:15.674 --> 31:15.794
you know,

31:15.874 --> 31:19.774
building tools to dox people online, to,
you know, analyzing flight data of,

31:19.834 --> 31:23.694
you know, flights, flights are public,
to analyzing satellite data, to

31:23.794 --> 31:26.854
analyzing, you know,
camera footage that's uploaded publicly.

31:27.474 --> 31:29.694
Like,
just any information available online,

31:29.774 --> 31:31.434
try to analyze that, identify, you

31:31.474 --> 31:35.414
know, like who knows who,
who did what when, who did what

31:35.474 --> 31:36.834
why, just get information.

31:38.314 --> 31:41.834
So this I want to work on. Uh,
once we have this

31:41.854 --> 31:45.754
information, then you can do things like,
okay, interview specific people

31:45.794 --> 31:48.234
and ask them to, you know,
turn to your side.

31:48.354 --> 31:52.134
Now, to turn to y- turn them to your side,
you need, y- typically, you

31:52.174 --> 31:54.614
either need to pay them money or,
you know,

31:54.654 --> 31:56.034
they should have some ideological

31:56.194 --> 31:58.944
or personal motivation for, you know,
giving you that information.

31:59.764 --> 32:03.404
So we can figure that stuff out. You know,
can we build, you know,

32:03.454 --> 32:05.974
infrastructure to, you know,
pay these people anonymously?

32:06.574 --> 32:09.574
Can we figure out, okay,
whose writings are persuasive and then go,

32:09.614 --> 32:10.154
you know, DM

32:10.214 --> 32:14.084
that persuasive message to
that specific insider and be like,

32:14.104 --> 32:17.274
"Okay, this is why you should, you know,
give us the information." We can figure

32:17.294 --> 32:18.174
out stuff like this.

32:19.514 --> 32:21.974
And lastly,
once we have an insider who's willing to,

32:22.034 --> 32:22.994
you know, turn on their

32:23.734 --> 32:25.414
circle and, you know, give us information,

32:26.394 --> 32:29.584
now how do they actually safely give this
to us? Like, what is the tooling?

32:29.624 --> 32:31.274
What's the infrastructure required?

32:32.254 --> 32:35.014
Uh, for example,
I have written a guide for whistleblowers.

32:35.074 --> 32:37.174
Like, if you are, you know,
US government whistleblower,

32:37.234 --> 32:37.874
how do you safely

32:37.894 --> 32:41.054
whistleblow?
I have written about how you can pay

32:41.134 --> 32:41.874
whistleblowers using

32:41.914 --> 32:45.754
cryptocurrency. I have written about,
you know, what are the tools

32:45.814 --> 32:48.714
that you can use to redact information
before publishing it.

32:48.804 --> 32:50.954
So like,
once a whistleblower has some information,

32:51.614 --> 32:51.904
[coughs]

32:52.514 --> 32:54.374
they can't immediately just leak it.

32:54.434 --> 32:57.134
They need to sit and, like,
redact certain pieces which are sensitive,

32:57.254 --> 32:57.514
and then

32:57.554 --> 33:00.414
only they can leak it.
I've written about tools to make

33:00.434 --> 33:01.094
that easy.

33:01.494 --> 33:04.034
There's also stuff around, like,
who will actually host this information?

33:04.094 --> 33:07.203
Like, whoever tries to host it, you know,
puts it on their website or, uh,

33:07.294 --> 33:10.994
whatever, they again become a target.
So make hosting easy.

33:12.014 --> 33:15.094
I think this all comes broadly around the
social circle umbrella.

33:15.154 --> 33:18.814
How do we break, you know, security of,
uh, people

33:18.834 --> 33:22.514
accelerating AI by, you know,
getting someone in their social circle to

33:22.554 --> 33:22.694
turn

33:22.734 --> 33:23.274
against them.

33:25.294 --> 33:27.074
These are all the methods we have roughly.

33:27.154 --> 33:28.734
Now,
which of these methods am I okay with,

33:28.814 --> 33:29.554
morally speaking?

33:31.454 --> 33:34.414
Uh, again,
I don't have very strong hard lines here,

33:34.514 --> 33:35.254
but I

33:35.314 --> 33:38.744
generally lean towards approaches that
are honest and not deceptive.

33:39.134 --> 33:39.494
Like,

33:40.734 --> 33:44.054
I care about honesty.
I care about maintaining promises to our

33:44.094 --> 33:48.034
allies, which requires honesty. Like,
if you're dishonest, why will people trust

33:48.094 --> 33:50.014
your promises? I care about honesty.

33:50.434 --> 33:53.074
Uh,
I want the truth to come out in the long

33:53.134 --> 33:53.254
run.

33:53.314 --> 33:54.774
Like,
the whole reason I'm doing this entire

33:54.834 --> 33:56.234
leaking thing in the first place is I

33:56.274 --> 33:57.834
want the truth to come out eventually.

33:58.634 --> 33:59.034
Uh,

34:00.374 --> 34:03.334
so it makes sense, like, okay, even
if we are going to, you know, attack these

34:03.374 --> 34:05.134
people,
I want to be honest about how we do it.

34:06.314 --> 34:09.874
I am less keen on things like social
engineering, for example.

34:10.034 --> 34:12.734
Like, social engineering works,
but I'm also not very keen on it for this

34:12.814 --> 34:13.154
reason.

34:14.114 --> 34:14.454
So I,

34:15.414 --> 34:19.214
I care about honesty. However,
I am also okay with some

34:19.254 --> 34:23.014
amount of violence.
I am okay with violating people's consent.

34:23.214 --> 34:25.074
I am okay with resisting authority.

34:25.434 --> 34:29.174
Like, we are not here to be nice. We
are here to fight these people and win.

34:29.934 --> 34:33.734
So for example, if this means, you know,
we hack into their servers, and this

34:33.994 --> 34:36.514
means we break their consent,
and this means we get-- leak

34:36.554 --> 34:37.434
that information anyway,

34:37.494 --> 34:40.554
you know, I am fine with this.
That's my short version of, okay,

34:40.574 --> 34:41.274
which methods am I

34:41.314 --> 34:41.914
fine with. Like,

34:43.134 --> 34:46.914
I am-- I want to be honest
and keep promises, but I

34:46.934 --> 34:50.504
am okay with some amount of, you know,
violence, some amount of, you know,

34:50.534 --> 34:51.834
violating people's consent.

34:52.974 --> 34:53.174
Yeah.

34:56.453 --> 34:58.854
And like, again, morality is personal.

34:58.953 --> 35:01.354
If you want to work with me,
it's possible you have a different answer

35:01.374 --> 35:01.764
to me than,

35:01.974 --> 35:04.634
um, on this question.
You can share your answer with me.

35:04.714 --> 35:06.474
We can, you know, debate about it,
whatever.

35:06.494 --> 35:06.834
Like,

35:07.974 --> 35:09.884
this is all right now my personal
opinions.

35:09.954 --> 35:12.634
Obviously, once we have, you know,
more people working on such stuff,

35:12.714 --> 35:13.074
they will all

35:13.094 --> 35:14.854
have to take their own personal calls on,
you know,

35:15.714 --> 35:16.774
how far are they willing to go.

35:23.054 --> 35:26.434
Cool.
I think now I'll come broadly to the last

35:26.474 --> 35:29.274
section of this video,
which roughly talks, okay, on

35:30.494 --> 35:34.294
who do I want to ally with? Like,
more narrowly, okay, who

35:34.334 --> 35:37.153
should my co-founder be? More broadly,
who should our allies be?

35:40.054 --> 35:43.054
Okay, so the first
and most important allies are obviously,

35:43.154 --> 35:43.954
okay, who do I

35:44.014 --> 35:47.314
work with and where do they get funding
from? So our funder has to be an ally.

35:47.654 --> 35:51.494
Now, I am not sure

35:51.554 --> 35:55.014
if there's any way to run this operation
as a for-profit. Like, I have tried.

35:55.214 --> 35:55.254
I

35:56.394 --> 35:59.484
wish we could do this as a for-profit,
but I'm not sure where we actually get the

35:59.534 --> 36:01.754
profit from. Uh, at least as of right now.

36:02.874 --> 36:05.014
Like,
there are many ways of profiting from

36:05.054 --> 36:05.734
cyberattacks.

36:05.774 --> 36:07.474
However, most of those seem unethical.

36:07.554 --> 36:07.814
Like,

36:09.454 --> 36:11.774
like,
I am okay targeting people who build

36:11.834 --> 36:13.374
superintelligence because, like I said,

36:13.434 --> 36:15.154
these people literally plan to kill me.

36:15.214 --> 36:18.034
You know, if they're planning to kill me,
I'm fine with, okay, of course, we attack

36:18.074 --> 36:21.954
them in return. However,
I'm not okay with, okay, we go,

36:22.034 --> 36:24.314
you know,
attack some random strangers in some

36:24.374 --> 36:25.734
random country, like

36:25.774 --> 36:28.514
cyberattack them, and then, you know,
like lock their computers

36:28.534 --> 36:29.354
and demand a crypto

36:29.394 --> 36:32.694
ransom or whatever.
I'm not keen with that, on with that,

36:32.794 --> 36:33.134
that we go

36:33.194 --> 36:36.314
at-attack some random innocent people who
have done something wrong. Who have...

36:36.374 --> 36:38.814
Sorry,
that we go attack some random innocent

36:38.854 --> 36:39.874
people who have done nothing wrong.

36:39.974 --> 36:40.114
Like-

36:41.586 --> 36:44.206
I--
that seems to go a bit too far even for

36:44.266 --> 36:44.406
me.

36:44.866 --> 36:48.186
Uh, therefore,
I'm not keen on profiting out of

36:48.206 --> 36:49.566
cyberattacks. Uh,

36:51.006 --> 36:53.046
otherwise you can profit out of, like,
defense.

36:53.126 --> 36:56.026
Like, you can sell, you know,
cyber defensive services to, you know,

36:56.226 --> 36:56.506
various

36:56.566 --> 36:58.226
corporates, to your own government.

36:59.046 --> 37:00.346
That I am okay with.

37:01.546 --> 37:03.146
So that could be a way we get profit.

37:03.266 --> 37:05.386
Otherwise,
we just have to do this as a nonprofit.

37:05.626 --> 37:08.966
I think we can get funding as a nonprofit
if we are actually competent.

37:09.086 --> 37:13.066
Like WikiLeaks at some point got over,
you know, $25 million in

37:13.086 --> 37:13.995
crypto donations,

37:14.946 --> 37:16.366
which is

37:17.866 --> 37:20.246
not as much as, you know,
like a billion-dollar tech company

37:20.266 --> 37:21.006
or whatever, but also

37:21.046 --> 37:24.126
that's a lot of money.
So I think we can get funding even

37:24.146 --> 37:24.986
if we operate as a

37:25.046 --> 37:28.446
nonprofit, if we are actually competent,
we are actually getting results.

37:30.386 --> 37:32.026
So yeah, the first person...

37:34.135 --> 37:37.786
[phone ringing] Sorry. Yeah. Yeah,
so the first person I

37:37.826 --> 37:41.686
need, and, uh,
I need to necessarily ally with is,

37:41.726 --> 37:44.406
okay, who are the people funding us?

37:44.496 --> 37:48.326
Another set of people who I think it will
be important to ally with is people

37:48.366 --> 37:50.786
from our own country.
At least they should give us some

37:50.826 --> 37:51.656
assurance they are not

37:51.686 --> 37:54.046
going to put us in prison for the things
we are going to do.

37:55.126 --> 37:57.506
Now, when I say our country,
which country do I mean?

37:57.746 --> 38:01.466
Now, as long as I'm with, living in India,
by default, this means, you know, the

38:01.486 --> 38:03.046
Indian government
and the Indian intelligence.

38:03.146 --> 38:06.926
If I can get some assurance from them
that, okay, if I attack any

38:07.006 --> 38:10.246
US AI company, leak whatever,
like Indian government won't put me in

38:10.286 --> 38:10.626
prison.

38:10.686 --> 38:12.726
Like if I can get that assurance,
that would be nice.

38:16.066 --> 38:16.386
Uh,

38:17.686 --> 38:21.366
also, I'm not choosy. I'm not saying,
"Okay, I will only ally with India here."

38:21.446 --> 38:21.726
Like,

38:22.806 --> 38:26.666
if tomorrow any of these countries,
could be Russia, could be

38:26.706 --> 38:30.406
Israel, could be whatever, they
were willing to

38:30.446 --> 38:33.326
work with me on this sort of stuff.
They were

38:34.266 --> 38:36.406
willing to take, let's say,
a strong stance.

38:36.456 --> 38:37.986
Okay, you know, superintelligence is bad.

38:38.066 --> 38:40.146
We will, you know, go against

38:41.006 --> 38:42.326
the people building superintelligence.

38:42.386 --> 38:44.426
If other countries
are willing to take that stance,

38:44.506 --> 38:46.126
I can also go live in some

38:46.166 --> 38:47.266
other country. Like,

38:48.426 --> 38:50.446
there are some details here.
They're kind of sensitive.

38:50.766 --> 38:50.946
Uh,

38:52.926 --> 38:54.086
but yeah, like

38:55.186 --> 38:59.026
the short answer is, yeah,
I want to be living outside the US

38:59.046 --> 39:02.486
and like figure out a way to, you know,
attack the US government, US AI companies,

39:02.626 --> 39:05.226
because they are the ones who
are right now accelerating this race to

39:05.286 --> 39:05.446
build

39:05.486 --> 39:06.326
superintelligence.

39:09.046 --> 39:11.666
Uh, again, like for example,
if I again go back to, you know,

39:11.706 --> 39:12.236
the WikiLeaks

39:12.306 --> 39:14.706
donation.
Now WikiLeaks got $25 million in

39:14.726 --> 39:15.246
donations.

39:15.306 --> 39:17.426
Like,
who did they actually get those donations

39:17.466 --> 39:18.666
from? We don't really know.

39:18.706 --> 39:21.686
Like some donors' names got leaked,
but mostly we don't actually know now.

39:22.226 --> 39:24.366
This may have been some, you know,
random American person.

39:24.426 --> 39:26.626
It may have been a Chinese person.
It may have been a Russian person.

39:26.646 --> 39:28.046
We actually don't know.

39:30.986 --> 39:34.886
Uh, if I am working in India,
I think I have more freedom to, you

39:34.906 --> 39:38.686
know, do things publicly. Like,
I can be more public about what I'm

39:38.706 --> 39:42.566
doing. I can also be more independent of,
you know, Indian formal,

39:42.806 --> 39:44.866
whatever, military chain of command,
et cetera.

39:45.756 --> 39:48.586
And I think this is a good thing because
people will trust us more.

39:48.726 --> 39:49.746
I will get feedback.

39:51.006 --> 39:54.226
But the downside is, yeah,
so far I've not been able to find people

39:54.306 --> 39:54.906
to work with.

39:55.086 --> 39:57.725
If I can find people to work with in
India, that would be great.

39:59.766 --> 40:03.386
If I hypothetically, again, this
is strictly hypothetical,

40:03.746 --> 40:07.586
went to Russia to work on this,
like I will be forced to be a lot

40:07.606 --> 40:10.526
more private about it,
and I will be forced to be a lot less

40:10.586 --> 40:11.566
independent about

40:11.626 --> 40:11.866
it.

40:13.046 --> 40:16.726
Uh, if I hypothetically
were to go live in Israel or something,

40:16.786 --> 40:18.526
obviously that has its own trade-offs
again.

40:20.146 --> 40:24.116
What I do strongly feel though is
that people who are living in the

40:24.126 --> 40:27.846
US, in the, in UK, in Western Europe,
if any of you

40:27.906 --> 40:31.186
think you have protection
when fighting against the US government,

40:31.366 --> 40:31.766
I think you are

40:31.786 --> 40:33.826
sorely mistaken, and you
are going to regret it.

40:33.906 --> 40:36.966
Like worst case,
you're going to pay for it with many years

40:37.146 --> 40:37.666
in prison.

40:38.626 --> 40:38.926
Uh,

40:40.206 --> 40:44.026
if you are living in, you know, Europe
and you're planning these sorts of

40:44.086 --> 40:45.126
offensive actions,

40:46.046 --> 40:49.666
you should probably leave Europe
is my guess. But there are details there.

40:49.846 --> 40:52.046
I don't want to get into that.

40:56.166 --> 40:57.966
Yeah, so the, again,
coming back to allies.

40:58.106 --> 41:00.176
The first ally is, okay,
people who can give us funding.

41:00.265 --> 41:02.736
Second ally is, okay,
if our own country can protect us,

41:02.806 --> 41:03.606
that would be nice.

41:08.086 --> 41:11.886
Uh, a third set of allies,
it would be nice if we, there are some

41:11.926 --> 41:15.026
insiders in the US AI companies
and the US government who will work with

41:15.106 --> 41:15.306
us.

41:15.366 --> 41:18.626
Like this is not like absolutely
essential, but it would be nice

41:18.686 --> 41:19.306
if we had people

41:19.326 --> 41:19.806
like this.

41:21.646 --> 41:24.966
Uh, it would be nice if there
are people outside our organization who

41:25.026 --> 41:25.486
are giving us

41:25.506 --> 41:27.566
good feedback. Like most feedback

41:28.446 --> 41:31.526
tends to suck.
Like most people have no context shared

41:31.566 --> 41:31.866
with you.

41:31.926 --> 41:35.835
They don't understand the constraints
you're operating with, but at least

41:35.835 --> 41:39.246
some feedback would be useful, and
that feedback is important,

41:39.826 --> 41:41.905
and I want at least some people to be
giving us feedback.

41:41.966 --> 41:44.525
I don't want to be just completely
isolated from the rest of society.

41:46.486 --> 41:46.866
Uh,

41:47.806 --> 41:50.646
example of feedback.
I previously said I don't actually know

41:50.666 --> 41:51.506
which leaks will help

41:51.526 --> 41:54.965
versus will not help. Like I have guesses,
but I'm not sure.

41:55.006 --> 41:58.866
If someone is very skilled at politics,
they know, okay, these leaks

41:58.886 --> 42:01.786
will specifically help. You know,
I want that sort of feedback.

42:01.846 --> 42:02.526
Like I want to know.

42:03.426 --> 42:05.525
So I also want that sort of ally.

42:06.946 --> 42:10.806
Okay, allies who I don't want.
I don't want to ally with any media house.

42:11.306 --> 42:13.746
Uh,
I don't want to ally with people who will

42:13.786 --> 42:14.886
help us with recruiting.

42:15.986 --> 42:17.186
I have reasons, but yeah.

42:18.066 --> 42:20.746
Like I want a small team
that actually has a mission

42:20.786 --> 42:21.806
and can actually do what it

42:21.846 --> 42:25.786
takes rather than, okay,
broad group of alliance of people who have

42:25.826 --> 42:28.446
network and blah, blah, blah,
but don't actually do anything.

42:32.366 --> 42:36.106
Yeah. Also, who will respect us?
Like we are planning to, or at least I

42:36.206 --> 42:40.086
am and people I want to work with,
are planning to do something

42:40.126 --> 42:43.506
at least somewhat radical.
Like not everyone is going to respect us.

42:43.586 --> 42:45.946
Not everyone is going to like us for the
fact that we're doing this.

42:47.110 --> 42:51.050
Some people will respect us. In general,
again, this is a guess, I

42:51.110 --> 42:53.330
don't know,
but my guess from limited number of

42:53.370 --> 42:54.090
conversations,

42:54.750 --> 42:58.450
people who value power more
are going to respect us more in general.

42:58.970 --> 43:02.390
So like some tech founders will respect
us, some billionaires will respect us,

43:02.510 --> 43:02.649
some

43:02.690 --> 43:06.190
politicians will respect us,
some social media influencers will respect

43:06.270 --> 43:06.630
us, some

43:06.830 --> 43:10.170
actors will, like, respect us,
some libertarians will respect us, some,

43:10.250 --> 43:10.370
you know,

43:10.430 --> 43:13.110
more radical-leaning leftists will respect
us.

43:14.710 --> 43:17.990
People who value power less will tend to
respect us less.

43:18.690 --> 43:21.630
Uh, or people who value, you know,
knowledge and truth less,

43:21.750 --> 43:22.250
they will tend to

43:22.270 --> 43:25.730
respect us less. Now,
I think religious people will respect us

43:25.790 --> 43:26.080
less,

43:26.270 --> 43:29.790
uh, liberals will respect us less,
people without any strong political

43:29.830 --> 43:30.170
opinion

43:30.190 --> 43:31.370
will respect us less.

43:32.490 --> 43:32.790
Uh,

43:36.010 --> 43:38.570
and like that's just a thing. This
is not something I can choose.

43:38.650 --> 43:41.070
This is just a fact about reality.
Some people will like us more

43:41.090 --> 43:41.580
and some people

43:41.590 --> 43:44.299
will hate us for what we're doing,
and that's just it.

43:46.890 --> 43:50.590
Uh, if we do succeed,
I think we will inspire more people to do

43:50.630 --> 43:50.869
similar

43:50.910 --> 43:54.850
things like this. Like, I am inspired by,
you know, long chain of people,

43:54.950 --> 43:58.750
many of which were broadly part of the
cypherpunk activist movement.

43:59.650 --> 43:59.970
Uh,

44:01.050 --> 44:03.110
if I succeed,
I will go inspire more people.

44:03.450 --> 44:04.530
That will happen.

44:04.610 --> 44:04.850
Like,

44:05.690 --> 44:08.010
I will prove that, like,
you can just go do things.

44:08.190 --> 44:08.510
Like,

44:10.470 --> 44:13.510
a lot of people are pessimistic about
their ability to, you know,

44:13.570 --> 44:14.370
change society in

44:14.390 --> 44:17.440
some very big way. Like,
I have talked to many people like this.

44:17.510 --> 44:20.810
This is, like, not some-- a very rare
or unique thing.

44:20.890 --> 44:23.440
Most people are pessimistic about their
ability to inf--

44:23.470 --> 44:24.810
affect large scale society

44:25.050 --> 44:28.390
changes. So if we can succeed at all,
this, you know,

44:29.270 --> 44:32.610
can help make them more optimistic
that they can also go do something.

44:38.590 --> 44:42.250
If you're still watching this video by
now, you already probably have a decent

44:42.330 --> 44:43.910
idea,
do you actually want to work with me on

44:43.970 --> 44:44.530
this or not?

44:44.610 --> 44:47.850
Like, but yeah, again,
I will just reiterate if this

44:47.870 --> 44:48.510
is not already

44:48.570 --> 44:49.370
clear, like,

44:50.590 --> 44:54.470
yeah, I prefer honesty. I prefer, like, we

44:54.510 --> 44:56.810
are okay with little bit violent approach,
too.

44:57.170 --> 45:00.590
So if you are okay with little bit violent
approach, too, then you should work with

45:00.630 --> 45:00.830
me.

45:01.970 --> 45:05.730
If you believe superintelligence
is coming, the

45:05.770 --> 45:09.610
world may be ending, especially
if you have emotionally internalized this,

45:09.710 --> 45:12.190
again, that's a good signal
that you should be working with me.

45:13.110 --> 45:16.390
If you value power,
that's a signal you should work with me.

45:17.090 --> 45:19.870
If you understand that the world
is run by self-interest,

45:21.090 --> 45:24.450
uh,
and systems that constrain self-interest,

45:24.610 --> 45:25.070
that is again

45:25.150 --> 45:26.210
a signal you should work with me.

45:27.550 --> 45:30.090
In terms of, like, competence, uh,

45:32.490 --> 45:36.070
I am personally good at generalist
research and creating new

45:36.130 --> 45:38.710
ideas. Like, again,
I'm not saying I'm some, you know,

45:38.720 --> 45:39.650
world expert or anything.

45:39.710 --> 45:42.150
I'm not.
But I think that's what I'm good at.

45:43.090 --> 45:46.750
I am also decent at software development.
Again, I'm not an expert at that.

45:47.030 --> 45:50.950
Like, I have built multiple apps.
I've worked on multiple scraping projects.

45:51.230 --> 45:53.990
I have some decent understanding of how
OpSec works.

45:56.730 --> 46:00.500
I am willing to upskill in cybersecurity,
so I'll become

46:00.530 --> 46:02.990
better at, you know, security research,
hacking, and so on.

46:03.470 --> 46:06.770
Again, I don't think I can become, like,
the world's best hacker or something, but

46:06.850 --> 46:09.470
I think if I put in the hours,
I put in the work, I can get, like,

46:09.550 --> 46:09.970
decent.

46:11.110 --> 46:12.710
So that's what I am good at.

46:13.730 --> 46:17.150
What I am not good at is networking,
fundraising,

46:17.310 --> 46:18.390
persuading people.

46:19.530 --> 46:21.730
Like, if you are good at those things,
you know,

46:21.850 --> 46:23.230
you will have a complementary fit.

46:23.710 --> 46:27.130
Uh, if you can help me raise funding,
for example, like, that would be a very

46:27.190 --> 46:28.990
strong signal that you should be working
with me.

46:31.110 --> 46:34.370
If you are skilled as a hacker, again,
you are welcome to work with me.

46:34.570 --> 46:37.130
Like,
anyone who's willing to work on this sort

46:37.170 --> 46:38.550
of radical project is

46:38.630 --> 46:42.410
going to be useful to me and also, well,
ideally to you.

46:42.450 --> 46:43.770
Like,
you should also be doing this for

46:43.810 --> 46:46.140
yourself, not just to appease

46:46.370 --> 46:49.610
Samuel. Like,
you should believe in leaking information.

46:49.670 --> 46:51.450
You should believe that, okay,
the world is ending.

46:51.750 --> 46:53.830
Like, this has to come from within you.

46:53.910 --> 46:54.510
I can't, you know,

46:55.830 --> 46:58.550
convince... Well, maybe I can convince,
but ultimately, like,

46:58.560 --> 46:59.510
the drive to this has

46:59.550 --> 47:01.450
to come from within.
I want to work with people who

47:01.490 --> 47:02.670
are self-motivated

47:04.110 --> 47:06.150
and agentic and so on.

47:09.450 --> 47:12.550
Uh,
maybe a minor aside on how should our

47:12.570 --> 47:13.430
OpSec work.

47:13.510 --> 47:13.830
Like...

47:18.850 --> 47:22.250
Yeah, this might seem like a tangent,
but I think it also needs to be covered.

47:22.630 --> 47:25.370
Uh,
I think we need to have the ability to

47:25.430 --> 47:26.510
keep at least something

47:26.610 --> 47:29.310
secret. Like,
we are getting information on our enemies.

47:29.410 --> 47:31.510
We don't want to just leak everything we
get.

47:31.550 --> 47:33.090
Like, some of those things
are actually private,

47:33.130 --> 47:34.370
and we want to keep them private.

47:34.450 --> 47:36.930
So some things we have to keep secret,
and the rest we will leak.

47:38.290 --> 47:40.790
Also, some of our methods,
some of our capabilities,

47:40.870 --> 47:42.050
how did we attack, that

47:42.090 --> 47:43.330
also needs to be kept secret.

47:44.510 --> 47:46.370
So these two things need to be secret a
bit.

47:46.710 --> 47:48.990
Rest remaining apart from this,
I think it's fine if we

47:49.010 --> 47:49.970
are public about a bunch of

47:50.010 --> 47:53.590
stuff. Like, again, you can debate, argue,
whatever with me, like,

47:54.850 --> 47:56.410
uh, opinions vary on that. Uh,

47:58.170 --> 48:02.090
now just as I think it is important we
are able to keep some secrets, I

48:02.110 --> 48:04.510
also think it will be difficult for us to
keep secrets.

48:04.550 --> 48:04.810
Like,

48:06.670 --> 48:09.270
in general, if you go read about,
you know, all the activists

48:09.310 --> 48:09.860
and so on who have

48:09.870 --> 48:12.610
worked on similar things,
generally they have found it hard to keep

48:12.670 --> 48:13.330
secrets even

48:13.370 --> 48:14.150
though they had to.

48:15.850 --> 48:18.550
And, like,
that is generally my whole worldview.

48:18.570 --> 48:20.950
Like, attacking people,
getting their information is easy,

48:21.290 --> 48:22.150
and defending your own

48:22.210 --> 48:24.830
secrets is hard. That
is why I'm saying let's go attack them,

48:24.870 --> 48:25.610
get their secrets.

48:25.730 --> 48:28.770
But on the flip side,
that also means they can also attack us

48:28.790 --> 48:29.330
and get all our

48:29.370 --> 48:30.450
secrets. So,

48:32.410 --> 48:34.590
like, that is just a fact about reality.

48:34.670 --> 48:37.610
This is where the playing field looks like
right now, where attack is easier and

48:37.630 --> 48:40.490
defense is hard. Uh,

48:42.870 --> 48:46.590
if we had, like, a large group of people,
and then we actually, you know, somewhat

48:46.670 --> 48:50.350
isolated from society,
then I think we can keep some

48:50.410 --> 48:53.270
secrets. We can, you know,
remain psychologically healthy.

48:53.350 --> 48:55.970
Like, the hard part of secrets
is not just technical stuff.

48:56.050 --> 48:58.230
The hard part is the effects of secrets on
your psychology

48:59.278 --> 49:00.598
Many people have written about this.

49:00.758 --> 49:03.318
This is again a whole separate thesis I
can get into.

49:04.018 --> 49:07.638
But like, yeah, keeping secrets
is hard because it's psychologically

49:07.698 --> 49:10.458
costly. Like, especially
if you think you're morally wrong in

49:10.518 --> 49:11.218
keeping the secret,

49:11.288 --> 49:13.078
then you are like done for.

49:13.978 --> 49:14.198
Uh,

49:15.898 --> 49:17.878
so we need to have strong moral
justifications.

49:17.938 --> 49:19.538
Okay, these are the things we
are keeping secret.

49:19.598 --> 49:21.238
These are the things we are leaking.
This is why.

49:21.298 --> 49:22.938
Like,
we need to be clear in our own heads about

49:22.998 --> 49:23.958
that so that we can remain

49:23.978 --> 49:26.858
psychologically healthy. That
is important.

49:28.578 --> 49:32.428
And yeah, like at least at the start,
once it's just me or one,

49:32.738 --> 49:36.278
like two people, three people, like we
are just not going to be able to keep

49:36.478 --> 49:39.998
any secrets. Like that's unfortunate,
but that's the reality.

49:40.038 --> 49:43.538
Like once we have a lot more people,
once we have money, competence, et cetera,

49:43.658 --> 49:47.278
then we may be able to pull it off.
But at the start, like keeping secrets

49:47.307 --> 49:47.538
is going

49:47.558 --> 49:49.858
to be hard even for us.
Like not just for our enemies,

49:49.918 --> 49:50.598
but also for us.

49:53.238 --> 49:53.998
Okay. So

49:55.078 --> 49:57.998
if you want to work with me,
you can contact me on Signal.

49:58.098 --> 49:59.738
My Signal username is available.

50:00.918 --> 50:04.598
I do not think the risks are very high
if you just have like a quick call with

50:04.638 --> 50:04.818
me.

50:05.158 --> 50:07.718
But if you actually work with me
full-time, then yes, there

50:07.758 --> 50:08.738
are significant risks

50:08.778 --> 50:10.378
to you for sure, 100%.

50:12.598 --> 50:15.038
Uh,
if you want to actually work with me

50:15.098 --> 50:16.538
full-time, like beyond just, you know,

50:16.598 --> 50:19.688
first one or two calls,
like I think the actual answer is,

50:19.758 --> 50:20.238
you know, you have to

50:20.278 --> 50:21.718
take a flight,
you have to meet me in person.

50:21.818 --> 50:23.298
I don't think there's any way to avoid
that.

50:23.398 --> 50:23.718
Like,

50:24.878 --> 50:28.818
and once we are working together,
we will also have to figure out a plan to

50:28.838 --> 50:30.858
keep your loved ones safe. Like,

50:32.638 --> 50:34.038
again,
those are things we'll have to figure out

50:34.078 --> 50:35.198
later, but they're important.

50:35.348 --> 50:35.738
And

50:37.478 --> 50:40.338
yeah,
like we have to live in a country outside,

50:40.398 --> 50:41.218
you know, US, UK,

50:41.298 --> 50:45.078
Europe, and we have to stay outside
and your social

50:45.118 --> 50:47.668
circle has to also stay outside.
And then we have to, you know,

50:47.678 --> 50:48.098
work on this

50:48.118 --> 50:50.658
full-time.
Like that seems like the only way I think

50:50.698 --> 50:51.858
we can actually pull this off.

50:54.858 --> 50:58.618
Okay. So that was a 50-minute

50:58.658 --> 51:01.038
video apparently. That
was way longer than I thought.

51:01.178 --> 51:03.778
Uh, this is my first take of the video,
by the way.

51:03.878 --> 51:07.818
I have not, you know,
made multiple takes and I don't think I

51:07.898 --> 51:09.987
will. Like I said, I am busy and,
you know,

51:10.017 --> 51:11.878
this plan will be outdated in six

51:11.958 --> 51:14.598
months anyway. Like
if you ask Samuel six months from now what

51:14.638 --> 51:15.318
his plan is, it will

51:15.358 --> 51:17.948
have been modified a fair bit anyway,
you know.

51:18.078 --> 51:19.618
So this is just my video on

51:20.578 --> 51:23.818
as of what, 15 July 2026, you know,

51:24.378 --> 51:27.958
here is my galaxy brain incomplete,
possibly in-inaccurate

51:28.218 --> 51:31.598
plan for what to do things.
If you want to work with me on this

51:31.658 --> 51:32.058
or something

51:32.098 --> 51:32.908
similar to this,

51:33.978 --> 51:35.918
like do reach out. Like,

51:38.458 --> 51:40.518
yeah, just do reach out. Like

51:41.598 --> 51:45.478
I am somewhat competent,
I am very motivated, and I think we

51:45.538 --> 51:47.398
can figure it out as we go along.